├── .difyignore ├── .env.example ├── .gitignore ├── GUIDE.md ├── LICENSE ├── PRIVACY.md ├── README.md ├── _assets ├── arxiv_mcp_server_test.png ├── asAgentStrategiesNode.png ├── everything_mcp_server_test_resource.png ├── icon.svg ├── initial_mcp_plugin_settings.png ├── plugin_install_offline.png └── plugin_install_online.png ├── main.py ├── manifest.yaml ├── output_parser └── cot_output_parser.py ├── prompt └── template.py ├── provider ├── agent.py └── agent.yaml ├── requirements.txt ├── strategies ├── ReAct.py ├── ReAct.yaml ├── function_calling.py ├── function_calling.yaml ├── mcpReAct.py └── mcpReAct.yaml └── test ├── chatflow ├── README.md ├── mcp_multi_sse_chatflow_v0.0.3.yml ├── mcp_sse_chatflow_v0.0.2.yml └── mcp_stdio_chatflow_v0.0.1.yml └── result ├── everything_mcp_server_prompt_log.txt ├── everything_mcp_server_resource_log.txt └── everything_mcp_server_tool_log.txt /.difyignore: -------------------------------------------------------------------------------- 1 | # Byte-compiled / optimized / DLL files 2 | __pycache__/ 3 | *.py[cod] 4 | *$py.class 5 | 6 | # Distribution / packaging 7 | .Python 8 | build/ 9 | develop-eggs/ 10 | dist/ 11 | downloads/ 12 | eggs/ 13 | .eggs/ 14 | lib/ 15 | lib64/ 16 | parts/ 17 | sdist/ 18 | var/ 19 | wheels/ 20 | share/python-wheels/ 21 | *.egg-info/ 22 | .installed.cfg 23 | *.egg 24 | MANIFEST 25 | 26 | # PyInstaller 27 | # Usually these files are written by a python script from a template 28 | # before PyInstaller builds the exe, so as to inject date/other infos into it. 29 | *.manifest 30 | *.spec 31 | 32 | # Installer logs 33 | pip-log.txt 34 | pip-delete-this-directory.txt 35 | 36 | # Unit test / coverage reports 37 | htmlcov/ 38 | .tox/ 39 | .nox/ 40 | .coverage 41 | .coverage.* 42 | .cache 43 | nosetests.xml 44 | coverage.xml 45 | *.cover 46 | *.py,cover 47 | .hypothesis/ 48 | .pytest_cache/ 49 | cover/ 50 | 51 | # Translations 52 | *.mo 53 | *.pot 54 | 55 | # Django stuff: 56 | *.log 57 | local_settings.py 58 | db.sqlite3 59 | db.sqlite3-journal 60 | 61 | # Flask stuff: 62 | instance/ 63 | .webassets-cache 64 | 65 | # Scrapy stuff: 66 | .scrapy 67 | 68 | # Sphinx documentation 69 | docs/_build/ 70 | 71 | # PyBuilder 72 | .pybuilder/ 73 | target/ 74 | 75 | # Jupyter Notebook 76 | .ipynb_checkpoints 77 | 78 | # IPython 79 | profile_default/ 80 | ipython_config.py 81 | 82 | # pyenv 83 | # For a library or package, you might want to ignore these files since the code is 84 | # intended to run in multiple environments; otherwise, check them in: 85 | .python-version 86 | 87 | # pipenv 88 | # According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control. 89 | # However, in case of collaboration, if having platform-specific dependencies or dependencies 90 | # having no cross-platform support, pipenv may install dependencies that don't work, or not 91 | # install all needed dependencies. 92 | Pipfile.lock 93 | 94 | # UV 95 | # Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control. 96 | # This is especially recommended for binary packages to ensure reproducibility, and is more 97 | # commonly ignored for libraries. 98 | uv.lock 99 | 100 | # poetry 101 | # Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control. 102 | # This is especially recommended for binary packages to ensure reproducibility, and is more 103 | # commonly ignored for libraries. 104 | # https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control 105 | poetry.lock 106 | 107 | # pdm 108 | # Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control. 109 | #pdm.lock 110 | # pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it 111 | # in version control. 112 | # https://pdm.fming.dev/latest/usage/project/#working-with-version-control 113 | .pdm.toml 114 | .pdm-python 115 | .pdm-build/ 116 | 117 | # PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm 118 | __pypackages__/ 119 | 120 | # Celery stuff 121 | celerybeat-schedule 122 | celerybeat.pid 123 | 124 | # SageMath parsed files 125 | *.sage.py 126 | 127 | # Environments 128 | .env 129 | .venv 130 | env/ 131 | venv/ 132 | ENV/ 133 | env.bak/ 134 | venv.bak/ 135 | 136 | # Spyder project settings 137 | .spyderproject 138 | .spyproject 139 | 140 | # Rope project settings 141 | .ropeproject 142 | 143 | # mkdocs documentation 144 | /site 145 | 146 | # mypy 147 | .mypy_cache/ 148 | .dmypy.json 149 | dmypy.json 150 | 151 | # Pyre type checker 152 | .pyre/ 153 | 154 | # pytype static type analyzer 155 | .pytype/ 156 | 157 | # Cython debug symbols 158 | cython_debug/ 159 | 160 | # PyCharm 161 | # JetBrains specific template is maintained in a separate JetBrains.gitignore that can 162 | # be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore 163 | # and can be added to the global gitignore or merged into this file. For a more nuclear 164 | # option (not recommended) you can uncomment the following to ignore the entire idea folder. 165 | .idea/ 166 | 167 | # Vscode 168 | .vscode/ 169 | 170 | # Git 171 | .git/ 172 | .gitignore 173 | 174 | # Mac 175 | .DS_Store 176 | 177 | # Windows 178 | Thumbs.db 179 | -------------------------------------------------------------------------------- /.env.example: -------------------------------------------------------------------------------- 1 | INSTALL_METHOD=remote 2 | REMOTE_INSTALL_HOST=debug.dify.ai 3 | REMOTE_INSTALL_PORT=5003 4 | REMOTE_INSTALL_KEY=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx 5 | -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- 1 | .DS_Store 2 | *.pyc 3 | venv/ 4 | .venv/ 5 | .vscode/ 6 | mhelper.py 7 | logs 8 | .env 9 | .idea 10 | .venv/ 11 | .scripts/ 12 | -------------------------------------------------------------------------------- /GUIDE.md: -------------------------------------------------------------------------------- 1 | ## User Guide of how to develop a Dify Plugin 2 | 3 | Hi there, looks like you have already created a Plugin, now let's get you started with the development! 4 | 5 | ### Choose a Plugin type you want to develop 6 | 7 | Before start, you need some basic knowledge about the Plugin types, Plugin supports to extend the following abilities in Dify: 8 | - **Tool**: Tool Providers like Google Search, Stable Diffusion, etc. it can be used to perform a specific task. 9 | - **Model**: Model Providers like OpenAI, Anthropic, etc. you can use their models to enhance the AI capabilities. 10 | - **Endpoint**: Like Service API in Dify and Ingress in Kubernetes, you can extend a http service as an endpoint and control its logics using your own code. 11 | 12 | Based on the ability you want to extend, we have divided the Plugin into three types: **Tool**, **Model**, and **Extension**. 13 | 14 | - **Tool**: It's a tool provider, but not only limited to tools, you can implement an endpoint there, for example, you need both `Sending Message` and `Receiving Message` if you are building a Discord Bot, **Tool** and **Endpoint** are both required. 15 | - **Model**: Just a model provider, extending others is not allowed. 16 | - **Extension**: Other times, you may only need a simple http service to extend the functionalities, **Extension** is the right choice for you. 17 | 18 | I believe you have chosen the right type for your Plugin while creating it, if not, you can change it later by modifying the `manifest.yaml` file. 19 | 20 | ### Manifest 21 | 22 | Now you can edit the `manifest.yaml` file to describe your Plugin, here is the basic structure of it: 23 | 24 | - version(version, required)：Plugin's version 25 | - type(type, required)：Plugin's type, currently only supports `plugin`, future support `bundle` 26 | - author(string, required)：Author, it's the organization name in Marketplace and should also equals to the owner of the repository 27 | - label(label, required)：Multi-language name 28 | - created_at(RFC3339, required)：Creation time, Marketplace requires that the creation time must be less than the current time 29 | - icon(asset, required)：Icon path 30 | - resource (object)：Resources to be applied 31 | - memory (int64)：Maximum memory usage, mainly related to resource application on SaaS for serverless, unit bytes 32 | - permission(object)：Permission application 33 | - tool(object)：Reverse call tool permission 34 | - enabled (bool) 35 | - model(object)：Reverse call model permission 36 | - enabled(bool) 37 | - llm(bool) 38 | - text_embedding(bool) 39 | - rerank(bool) 40 | - tts(bool) 41 | - speech2text(bool) 42 | - moderation(bool) 43 | - node(object)：Reverse call node permission 44 | - enabled(bool) 45 | - endpoint(object)：Allow to register endpoint permission 46 | - enabled(bool) 47 | - app(object)：Reverse call app permission 48 | - enabled(bool) 49 | - storage(object)：Apply for persistent storage permission 50 | - enabled(bool) 51 | - size(int64)：Maximum allowed persistent memory, unit bytes 52 | - plugins(object, required)：Plugin extension specific ability yaml file list, absolute path in the plugin package, if you need to extend the model, you need to define a file like openai.yaml, and fill in the path here, and the file on the path must exist, otherwise the packaging will fail. 53 | - Format 54 | - tools(list[string]): Extended tool suppliers, as for the detailed format, please refer to [Tool Guide](https://docs.dify.ai/docs/plugins/standard/tool_provider) 55 | - models(list[string])：Extended model suppliers, as for the detailed format, please refer to [Model Guide](https://docs.dify.ai/docs/plugins/standard/model_provider) 56 | - endpoints(list[string])：Extended Endpoints suppliers, as for the detailed format, please refer to [Endpoint Guide](https://docs.dify.ai/docs/plugins/standard/endpoint_group) 57 | - Restrictions 58 | - Not allowed to extend both tools and models 59 | - Not allowed to have no extension 60 | - Not allowed to extend both models and endpoints 61 | - Currently only supports up to one supplier of each type of extension 62 | - meta(object) 63 | - version(version, required)：manifest format version, initial version 0.0.1 64 | - arch(list[string], required)：Supported architectures, currently only supports amd64 arm64 65 | - runner(object, required)：Runtime configuration 66 | - language(string)：Currently only supports python 67 | - version(string)：Language version, currently only supports 3.12 68 | - entrypoint(string)：Program entry, in python it should be main 69 | 70 | ### Install Dependencies 71 | 72 | - First of all, you need a Python 3.11+ environment, as our SDK requires that. 73 | - Then, install the dependencies: 74 | ```bash 75 | pip install -r requirements.txt 76 | ``` 77 | - If you want to add more dependencies, you can add them to the `requirements.txt` file, once you have set the runner to python in the `manifest.yaml` file, `requirements.txt` will be automatically generated and used for packaging and deployment. 78 | 79 | ### Implement the Plugin 80 | 81 | Now you can start to implement your Plugin, by following these examples, you can quickly understand how to implement your own Plugin: 82 | 83 | - [OpenAI](https://github.com/langgenius/dify-plugin-sdks/tree/main/python/examples/openai): best practice for model provider 84 | - [Google Search](https://github.com/langgenius/dify-plugin-sdks/tree/main/python/examples/google): a simple example for tool provider 85 | - [Neko](https://github.com/langgenius/dify-plugin-sdks/tree/main/python/examples/neko): a funny example for endpoint group 86 | 87 | ### Test and Debug the Plugin 88 | 89 | You may already noticed that a `.env.example` file in the root directory of your Plugin, just copy it to `.env` and fill in the corresponding values, there are some environment variables you need to set if you want to debug your Plugin locally. 90 | 91 | - `INSTALL_METHOD`: Set this to `remote`, your plugin will connect to a Dify instance through the network. 92 | - `REMOTE_INSTALL_HOST`: The host of your Dify instance, you can use our SaaS instance `https://debug.dify.ai`, or self-hosted Dify instance. 93 | - `REMOTE_INSTALL_PORT`: The port of your Dify instance, default is 5003 94 | - `REMOTE_INSTALL_KEY`: You should get your debugging key from the Dify instance you used, at the right top of the plugin management page, you can see a button with a `debug` icon, click it and you will get the key. 95 | 96 | Run the following command to start your Plugin: 97 | 98 | ```bash 99 | python -m main 100 | ``` 101 | 102 | Refresh the page of your Dify instance, you should be able to see your Plugin in the list now, but it will be marked as `debugging`, you can use it normally, but not recommended for production. 103 | 104 | ### Package the Plugin 105 | 106 | After all, just package your Plugin by running the following command: 107 | 108 | ```bash 109 | dify-plugin plugin package ./ROOT_DIRECTORY_OF_YOUR_PLUGIN 110 | ``` 111 | 112 | you will get a `plugin.difypkg` file, that's all, you can submit it to the Marketplace now, look forward to your Plugin being listed! 113 | 114 | 115 | ## User Privacy Policy 116 | 117 | Please fill in the privacy policy of the plugin if you want to make it published on the Marketplace, refer to [PRIVACY.md](PRIVACY.md) for more details. -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | Apache License 2 | Version 2.0, January 2004 3 | http://www.apache.org/licenses/ 4 | 5 | TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION 6 | 7 | 1. Definitions. 8 | 9 | "License" shall mean the terms and conditions for use, reproduction, 10 | and distribution as defined by Sections 1 through 9 of this document. 11 | 12 | "Licensor" shall mean the copyright owner or entity authorized by 13 | the copyright owner that is granting the License. 14 | 15 | "Legal Entity" shall mean the union of the acting entity and all 16 | other entities that control, are controlled by, or are under common 17 | control with that entity. For the purposes of this definition, 18 | "control" means (i) the power, direct or indirect, to cause the 19 | direction or management of such entity, whether by contract or 20 | otherwise, or (ii) ownership of fifty percent (50%) or more of the 21 | outstanding shares, or (iii) beneficial ownership of such entity. 22 | 23 | "You" (or "Your") shall mean an individual or Legal Entity 24 | exercising permissions granted by this License. 25 | 26 | "Source" form shall mean the preferred form for making modifications, 27 | including but not limited to software source code, documentation 28 | source, and configuration files. 29 | 30 | "Object" form shall mean any form resulting from mechanical 31 | transformation or translation of a Source form, including but 32 | not limited to compiled object code, generated documentation, 33 | and conversions to other media types. 34 | 35 | "Work" shall mean the work of authorship, whether in Source or 36 | Object form, made available under the License, as indicated by a 37 | copyright notice that is included in or attached to the work 38 | (an example is provided in the Appendix below). 39 | 40 | "Derivative Works" shall mean any work, whether in Source or Object 41 | form, that is based on (or derived from) the Work and for which the 42 | editorial revisions, annotations, elaborations, or other modifications 43 | represent, as a whole, an original work of authorship. For the purposes 44 | of this License, Derivative Works shall not include works that remain 45 | separable from, or merely link (or bind by name) to the interfaces of, 46 | the Work and Derivative Works thereof. 47 | 48 | "Contribution" shall mean any work of authorship, including 49 | the original version of the Work and any modifications or additions 50 | to that Work or Derivative Works thereof, that is intentionally 51 | submitted to Licensor for inclusion in the Work by the copyright owner 52 | or by an individual or Legal Entity authorized to submit on behalf of 53 | the copyright owner. For the purposes of this definition, "submitted" 54 | means any form of electronic, verbal, or written communication sent 55 | to the Licensor or its representatives, including but not limited to 56 | communication on electronic mailing lists, source code control systems, 57 | and issue tracking systems that are managed by, or on behalf of, the 58 | Licensor for the purpose of discussing and improving the Work, but 59 | excluding communication that is conspicuously marked or otherwise 60 | designated in writing by the copyright owner as "Not a Contribution." 61 | 62 | "Contributor" shall mean Licensor and any individual or Legal Entity 63 | on behalf of whom a Contribution has been received by Licensor and 64 | subsequently incorporated within the Work. 65 | 66 | 2. Grant of Copyright License. Subject to the terms and conditions of 67 | this License, each Contributor hereby grants to You a perpetual, 68 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 69 | copyright license to reproduce, prepare Derivative Works of, 70 | publicly display, publicly perform, sublicense, and distribute the 71 | Work and such Derivative Works in Source or Object form. 72 | 73 | 3. Grant of Patent License. Subject to the terms and conditions of 74 | this License, each Contributor hereby grants to You a perpetual, 75 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 76 | (except as stated in this section) patent license to make, have made, 77 | use, offer to sell, sell, import, and otherwise transfer the Work, 78 | where such license applies only to those patent claims licensable 79 | by such Contributor that are necessarily infringed by their 80 | Contribution(s) alone or by combination of their Contribution(s) 81 | with the Work to which such Contribution(s) was submitted. If You 82 | institute patent litigation against any entity (including a 83 | cross-claim or counterclaim in a lawsuit) alleging that the Work 84 | or a Contribution incorporated within the Work constitutes direct 85 | or contributory patent infringement, then any patent licenses 86 | granted to You under this License for that Work shall terminate 87 | as of the date such litigation is filed. 88 | 89 | 4. Redistribution. You may reproduce and distribute copies of the 90 | Work or Derivative Works thereof in any medium, with or without 91 | modifications, and in Source or Object form, provided that You 92 | meet the following conditions: 93 | 94 | (a) You must give any other recipients of the Work or 95 | Derivative Works a copy of this License; and 96 | 97 | (b) You must cause any modified files to carry prominent notices 98 | stating that You changed the files; and 99 | 100 | (c) You must retain, in the Source form of any Derivative Works 101 | that You distribute, all copyright, patent, trademark, and 102 | attribution notices from the Source form of the Work, 103 | excluding those notices that do not pertain to any part of 104 | the Derivative Works; and 105 | 106 | (d) If the Work includes a "NOTICE" text file as part of its 107 | distribution, then any Derivative Works that You distribute must 108 | include a readable copy of the attribution notices contained 109 | within such NOTICE file, excluding those notices that do not 110 | pertain to any part of the Derivative Works, in at least one 111 | of the following places: within a NOTICE text file distributed 112 | as part of the Derivative Works; within the Source form or 113 | documentation, if provided along with the Derivative Works; or, 114 | within a display generated by the Derivative Works, if and 115 | wherever such third-party notices normally appear. The contents 116 | of the NOTICE file are for informational purposes only and 117 | do not modify the License. You may add Your own attribution 118 | notices within Derivative Works that You distribute, alongside 119 | or as an addendum to the NOTICE text from the Work, provided 120 | that such additional attribution notices cannot be construed 121 | as modifying the License. 122 | 123 | You may add Your own copyright statement to Your modifications and 124 | may provide additional or different license terms and conditions 125 | for use, reproduction, or distribution of Your modifications, or 126 | for any such Derivative Works as a whole, provided Your use, 127 | reproduction, and distribution of the Work otherwise complies with 128 | the conditions stated in this License. 129 | 130 | 5. Submission of Contributions. Unless You explicitly state otherwise, 131 | any Contribution intentionally submitted for inclusion in the Work 132 | by You to the Licensor shall be under the terms and conditions of 133 | this License, without any additional terms or conditions. 134 | Notwithstanding the above, nothing herein shall supersede or modify 135 | the terms of any separate license agreement you may have executed 136 | with Licensor regarding such Contributions. 137 | 138 | 6. Trademarks. This License does not grant permission to use the trade 139 | names, trademarks, service marks, or product names of the Licensor, 140 | except as required for reasonable and customary use in describing the 141 | origin of the Work and reproducing the content of the NOTICE file. 142 | 143 | 7. Disclaimer of Warranty. Unless required by applicable law or 144 | agreed to in writing, Licensor provides the Work (and each 145 | Contributor provides its Contributions) on an "AS IS" BASIS, 146 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or 147 | implied, including, without limitation, any warranties or conditions 148 | of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A 149 | PARTICULAR PURPOSE. You are solely responsible for determining the 150 | appropriateness of using or redistributing the Work and assume any 151 | risks associated with Your exercise of permissions under this License. 152 | 153 | 8. Limitation of Liability. In no event and under no legal theory, 154 | whether in tort (including negligence), contract, or otherwise, 155 | unless required by applicable law (such as deliberate and grossly 156 | negligent acts) or agreed to in writing, shall any Contributor be 157 | liable to You for damages, including any direct, indirect, special, 158 | incidental, or consequential damages of any character arising as a 159 | result of this License or out of the use or inability to use the 160 | Work (including but not limited to damages for loss of goodwill, 161 | work stoppage, computer failure or malfunction, or any and all 162 | other commercial damages or losses), even if such Contributor 163 | has been advised of the possibility of such damages. 164 | 165 | 9. Accepting Warranty or Additional Liability. While redistributing 166 | the Work or Derivative Works thereof, You may choose to offer, 167 | and charge a fee for, acceptance of support, warranty, indemnity, 168 | or other liability obligations and/or rights consistent with this 169 | License. However, in accepting such obligations, You may act only 170 | on Your own behalf and on Your sole responsibility, not on behalf 171 | of any other Contributor, and only if You agree to indemnify, 172 | defend, and hold each Contributor harmless for any liability 173 | incurred by, or claims asserted against, such Contributor by reason 174 | of your accepting any such warranty or additional liability. 175 | 176 | END OF TERMS AND CONDITIONS 177 | 178 | APPENDIX: How to apply the Apache License to your work. 179 | 180 | To apply the Apache License to your work, attach the following 181 | boilerplate notice, with the fields enclosed by brackets "[]" 182 | replaced with your own identifying information. (Don't include 183 | the brackets!) The text should be enclosed in the appropriate 184 | comment syntax for the file format. We also recommend that a 185 | file or class name and description of purpose be included on the 186 | same "printed page" as the copyright notice for easier 187 | identification within third-party archives. 188 | 189 | Copyright [yyyy] [name of copyright owner] 190 | 191 | Licensed under the Apache License, Version 2.0 (the "License"); 192 | you may not use this file except in compliance with the License. 193 | You may obtain a copy of the License at 194 | 195 | http://www.apache.org/licenses/LICENSE-2.0 196 | 197 | Unless required by applicable law or agreed to in writing, software 198 | distributed under the License is distributed on an "AS IS" BASIS, 199 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 200 | See the License for the specific language governing permissions and 201 | limitations under the License. 202 | -------------------------------------------------------------------------------- /PRIVACY.md: -------------------------------------------------------------------------------- 1 | ## Privacy 2 | This plugin is released under the Apache License 2.0. Users are responsible for protecting their own privacy and securing their data according to the guidelines of their chosen LLM API platform. Please read the README for correct usage, as the default implementation does not include a Human-In-The-Loop mechanism. 3 | 4 | ### Q1: What data does the plugin access? 5 | ### A1: 6 | The plugin itself does not actively collect personal information. However, the data it accesses depends on the connected MCP server. Depending on the server’s configuration, it may temporarily access user-related data (e.g., user names, file paths, etc.). Once the execution of the mcpReAct agent strategy node is complete, all such data is discarded. 7 | 8 | ### Q2: Does the plugin collect or store personal information? 9 | ### A2: 10 | No. As an individual developer, I do not implement any persistent collection of personal data. The plugin is designed to run transiently, and any accessed information is not stored beyond the lifetime of a single operation. 11 | 12 | ### Q3: What about privacy in relation to LLM APIs? 13 | ### A3: 14 | Since users choose their own LLM API platform, the privacy and security of data exchanged with those APIs are determined by the respective platforms. Please review and comply with the privacy policies of your chosen LLM provider. 15 | 16 | ### Q4: What is the user’s responsibility? 17 | ### A4: 18 | Read the Documentation: Ensure you follow the usage guidelines provided in the README. 19 | Trusted MCP Servers: Only connect to MCP servers you trust. 20 | Privacy Management: Under Apache License 2.0, this plugin comes with no warranty regarding privacy or security. You are solely responsible for protecting your own data. -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # dify-mcp-client 2 | `MCP Client` as Agent Strategy Plugin. 3 | > [!IMPORTANT] 4 | > Dify is not `MCP Server` but `MCP Host`. 5 | 6 | ![showcase1](./_assets/arxiv_mcp_server_test.png) 7 | 8 | ## How it works 9 | Each `MCP client` (ReAct Agent) node can connect `MCP servers`. 10 | 1. `Tool`, `Resource`, `Prompt` lists are converted into Dify Tools. 11 | 2. Your selected LLM can see their `name`, `description`, `argument type` 12 | 3. The LLM calls Tools based on the ReAct loop (Reason → Act → Observe). 13 | 14 | > [!NOTE] 15 | > Most of the code in this repository contains the following files. 16 | > #### Dify Official Plugins / Agent Strategies 17 | > https://github.com/langgenius/dify-official-plugins/tree/main/agent-strategies/cot_agent 18 | 19 | ## ✅ What I did 20 | - Copied `ReAct.py` and renamed file as `mcpReAct.py` 21 | - Added `config_json` GUI input field by editing `mcpReAct.yaml` and `class mcpReActParams()` 22 | 23 | ### in mcpReAct.py, I added 24 | - New 12 functions for MCP 25 | - `__init__()` for initializing `AsyncExitStack` and `event loop` 26 | - Some codes in `_handle_invoke_action()` for MCP 27 | - MCP setup and cleanup in `_invoke()` 28 | 29 | > [!IMPORTANT] 30 | > ReAct while loop is as they are 31 | 32 | ## 🔄 Update history 33 | - Add SSE MCP client (v0.0.2) 34 | - Support multi SSE servers (v0.0.3) 35 | - Update python module and simplify its dependency (v0.0.4) 36 | - mcp(v1.1.2→v1.6.0+) 37 | - dify_plugin(0.0.1b72→v0.1.0) 38 | 39 | ## ⚠️ Caution and Limitation 40 | > [!CAUTION] 41 | > This plugin does **not** implement a **human-in-the-loop** mechanism by default, so connect **reliable mcp server only**.
42 | > To avoid it, decrease `max itereations`(default:`3`) to `1`, and use this Agent node repeatedly in Chatflow.
43 | > However, agent memory is reset by the end of Workflow.
44 | > Use `Conversaton Variable` to save history and pass it to QUERY. 45 | > Don't forget to add a phrase such as 46 | > *"ask for user's permission when calling tools"* in INSTRUCTION. 47 | 48 | # How to use this plugin 49 | 50 | ## 🛜Install the plugin from GitHub 51 | - Enter the following GitHub repository name 52 | ``` 53 | https://github.com/3dify-project/dify-mcp-client/ 54 | ``` 55 | - Dify > PLUGINS > + Install plugin > INSTALL FROM > GitHub 56 | ![difyUI1](./_assets/plugin_install_online.png) 57 | 58 | ## ⬇️Install the plugin from .difypkg file 59 | - Go to Releases https://github.com/3dify-project/dify-mcp-client/releases 60 | - Select suitable version of `.difypkg` 61 | - Dify > PLUGINS > + Install plugin > INSTALL FROM > Local Package File 62 | ![difyUI2](./_assets/plugin_install_offline.png) 63 | 64 | ## How to handle errors when installing plugins? 65 | 66 | **Issue**: If you encounter the error message: `plugin verification has been enabled, and the plugin you want to install has a bad signature`, how to handle the issue?
67 | **Solution**: Open `/docker/.env` and change from `true` to `false`: 68 | ``` 69 | FORCE_VERIFYING_SIGNATURE=false 70 | ``` 71 | Run the following commands to restart the Dify service: 72 | ```bash 73 | cd docker 74 | docker compose down 75 | docker compose up -d 76 | ``` 77 | Once this field is added, the Dify platform will allow the installation of all plugins that are not listed (and thus not verified) in the Dify Marketplace. 78 | > [!TIP] 79 | > Marketplace need Approval. If stars⭐ reach 100, I'll consider to make PR for them. 80 | 81 | ## Where does this plugin show up? 82 | - It takes few minutes to install 83 | - Once installed, you can use it any workflows as Agent node 84 | - Select "mcpReAct" strategy (otherwise no MCP) 85 | ![asAgentStrategiesNode](./_assets/asAgentStrategiesNode.png) 86 | 87 | ## Config 88 | MCP Agent Plugin node require config_json like this to command or URL to connect MCP servers 89 | ``` 90 | { 91 | "mcpServers":{ 92 | "name_of_server1":{ 93 | "url": "http://host.docker.internal:8080/sse" 94 | }, 95 | "name_of_server2":{ 96 | "url": "http://host.docker.internal:8008/sse" 97 | } 98 | } 99 | } 100 | ``` 101 | > [!WARNING] 102 | > - Each server's port number should be different, like 8080, 8008, ... 103 | > - If you want to use stdio mcp server, there are 3 ways. 104 | > 1. Convert it to SSE mcp server https://github.com/3dify-project/dify-mcp-client/edit/main/README.md#how-to-convert-stdio-mcp-server-into-sse-mcp-server 105 | > 2. Deploy with source code (**NOT** by .difypkg or GitHub reposity name install) https://github.com/3dify-project/dify-mcp-client/edit/main/README.md#-how-to-develop-and-deploy-plugin 106 | > 3. Pre-install Node.js inside dify-plugin docker (issue:https://github.com/3dify-project/dify-mcp-client/issues/10) guide: https://github.com/tangyoha/tangyoha-bili/tree/master/dify/mcp/map_mcp 107 | 108 | ## Chatflow Example 109 | ![showcase2](./_assets/everything_mcp_server_test_resource.png) 110 | > [!WARNING] 111 | > - The Tools field should not be left blank. so **select Dify tools** like "current time". 112 | #### I provide this Dify ChatFlow `.yml` for testing this plugin. 113 | https://github.com/3dify-project/dify-mcp-client/tree/main/test/chatflow 114 | #### After download DSL(yml) file, import it in Dify and you can test MCP using "Everything MCP server" 115 | https://github.com/modelcontextprotocol/servers/tree/main/src/everything 116 | 117 | # How to convert `stdio` MCP server into SSE MCP server 118 | ## option1️⃣: Edit MCP server's code 119 | If fastMCP server, change like this 120 | ```diff 121 | if __name__ == "__main__": 122 | - mcp.run(transport="stdio") 123 | + mcp.run(transport="sse") 124 | ``` 125 | 126 | ## option2️⃣: via mcp-proxy 127 | ``` 128 | \mcp-proxy>uv venv -p 3.12 129 | .venv\Scripts\activate 130 | uv tool install mcp-proxy 131 | ``` 132 | ### Check Node.js has installed and npx(.cmd) Path 133 | (Mac/Linux) 134 | ``` 135 | which npx 136 | ``` 137 | (Windows) 138 | ``` 139 | where npx 140 | ``` 141 | result 142 | ``` 143 | C:\Program Files\nodejs\npx 144 | C:\Program Files\nodejs\npx.cmd 145 | C:\Users\USER_NAME\AppData\Roaming\npm\npx 146 | C:\Users\USER_NAME\AppData\Roaming\npm\npx.cmd 147 | ``` 148 | 149 | If claude_desktop_config.json is following schema, 150 | ``` 151 | { 152 | "mcpServers": { 153 | "SERVER_NAME": { 154 | "command": CMD_NAME_OR_PATH 155 | "args": {VALUE1, VALUE2} 156 | } 157 | } 158 | } 159 | ``` 160 | ### Wake up stdio MCP server by this command 161 | ``` 162 | mcp-proxy --sse-port=8080 --pass-environment -- CMD_NAME_OR_PATH --arg1 VALUE1 --arg2 VALUE2 ... 163 | ``` 164 | If your OS is Windows, use npx.cmd instead of npx. Following is example command to convert stdio "everything MCP server" to SSE via mcp-proxy. 165 | ``` 166 | mcp-proxy --sse-port=8080 --pass-environment -- C:\Program Files\nodejs\npx.cmd --arg1 -y --arg2 @modelcontextprotocol/server-everything 167 | ``` 168 | 169 | Similarly, on another command line (If you use sample Chatflow for v0.0.3) 170 | ``` 171 | pip install mcp-simple-arxiv 172 | mcp-proxy --sse-port=8008 --pass-environment -- C:\Users\USER_NAME\AppData\Local\Programs\Python\Python310\python.exe -m -mcp_simple_arxiv 173 | ``` 174 | 175 | Following is a mcp-proxy setup log. 176 | ``` 177 | (mcp_proxy) C:\User\USER_NAME\mcp-proxy>mcp-proxy --sse-port=8080 --pass-environment -- C:\Program Files\nodejs\npx.cmd --arg1 -y --arg2 @modelcontextprotocol/server-everything 178 | DEBUG:root:Starting stdio client and SSE server 179 | DEBUG:asyncio:Using proactor: IocpProactor 180 | DEBUG:mcp.server.lowlevel.server:Initializing server 'example-servers/everything' 181 | DEBUG:mcp.server.sse:SseServerTransport initialized with endpoint: /messages/ 182 | INFO: Started server process [53104] 183 | INFO: Waiting for application startup. 184 | INFO: Application startup complete. 185 | INFO: Uvicorn running on http://127.0.0.1:8080 (Press CTRL+C to quit) 186 | ``` 187 | 188 | # 🔨 How to develop and deploy plugin 189 | 190 | ### Official plugin dev guide 191 | https://github.com/3dify-project/dify-mcp-client/blob/main/GUIDE.md 192 | 193 | ### Dify plugin SDK daemon 194 | If your OS is Windows and CPU is Intel or AMD, you need to download `dify-plugin-windows-amd64.exe` (v0.0.7)
195 | Choose your OS-compatible verson here:
196 | https://github.com/langgenius/dify-plugin-daemon/releases
197 | 1. Rename it as dify.exe for convinence 198 | 2. mkdir "C\User\user\\.local\bin" (Windows) and register it as system path. 199 | 3. Copy `dify.exe` to under dify-mcp-client/ 200 | > [!TIP] 201 | > Following guide is helpful. 202 | > https://docs.dify.ai/plugins/quick-start/develop-plugins/initialize-development-tools 203 | 204 | ### Reference 205 | https://docs.dify.ai/plugins/quick-start/develop-plugins/initialize-development-tools 206 | 207 | > [!NOTE] 208 | > You can skip this stage if you pull or download codes of this repo 209 | > ``` 210 | > dify plugin init 211 | > ``` 212 | > Initial settings are as follow 213 | > ![InitialDifyPluginSetting](./_assets/initial_mcp_plugin_settings.png) 214 | 215 | ### Change directory 216 | ``` 217 | cd dify-mcp-client 218 | ``` 219 | 220 | ### Install python module 221 | Python3.12+ is compatible. The `venv` and `uv` are not necessary, but recommended. 222 | ``` 223 | uv venv -p 3.12 224 | .venv\Scripts\activate 225 | ``` 226 | Install python modules for plugin development 227 | ``` 228 | uv pip install -r requirements.txt 229 | ``` 230 | 231 | ### Duplicate `env.example` and rename one to `.env` 232 | I changed `REMOTE_INSTALL_HOST` from `debug.dify.ai` to `localhost` 233 | (Docker Compose environment) 234 | click 🪲bug icon button to see these information 235 | 236 | ### Activate Dify plugin 237 | ``` 238 | python -m main 239 | ``` 240 | (ctrl+C to stop) 241 | > [!TIP] 242 | > REMOTE_INSTALL_KEY of .env often changes. 243 | > If you encounter error messages like `handshake failed, invalid key`, renew it. 244 | 245 | ### Package into .difypkg 246 | `./dify-mcp-client` is my default root name 247 | ``` 248 | dify plugin package ./ROOT_OF_YOUR_PROJECT 249 | ``` 250 | 251 | ## Useful GitHub repositories for developers 252 | 253 | #### Dify Plugin SDKs 254 | https://github.com/langgenius/dify-plugin-sdks 255 | 256 | #### MCP Python SDK 257 | https://github.com/modelcontextprotocol/python-sdk 258 |
259 | 260 | > [!TIP] 261 | > MCP client example
262 | > https://github.com/modelcontextprotocol/python-sdk/blob/main/examples/clients/simple-chatbot/mcp_simple_chatbot/main.py
263 | 264 | > [!NOTE] 265 | > Dify plugin has `requirements.txt` which automatically installs python modules.
266 | > I include latest `mcp` in it, so you don't need to download the MCP SDK separately. 267 | -------------------------------------------------------------------------------- /_assets/arxiv_mcp_server_test.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/3dify-project/dify-mcp-client/cd6d27dd49f2f17b10cc71814f50f755c5766d91/_assets/arxiv_mcp_server_test.png -------------------------------------------------------------------------------- /_assets/asAgentStrategiesNode.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/3dify-project/dify-mcp-client/cd6d27dd49f2f17b10cc71814f50f755c5766d91/_assets/asAgentStrategiesNode.png -------------------------------------------------------------------------------- /_assets/everything_mcp_server_test_resource.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/3dify-project/dify-mcp-client/cd6d27dd49f2f17b10cc71814f50f755c5766d91/_assets/everything_mcp_server_test_resource.png -------------------------------------------------------------------------------- /_assets/icon.svg: -------------------------------------------------------------------------------- 1 | 2 | 6 | 7 | -------------------------------------------------------------------------------- /_assets/initial_mcp_plugin_settings.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/3dify-project/dify-mcp-client/cd6d27dd49f2f17b10cc71814f50f755c5766d91/_assets/initial_mcp_plugin_settings.png -------------------------------------------------------------------------------- /_assets/plugin_install_offline.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/3dify-project/dify-mcp-client/cd6d27dd49f2f17b10cc71814f50f755c5766d91/_assets/plugin_install_offline.png -------------------------------------------------------------------------------- /_assets/plugin_install_online.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/3dify-project/dify-mcp-client/cd6d27dd49f2f17b10cc71814f50f755c5766d91/_assets/plugin_install_online.png -------------------------------------------------------------------------------- /main.py: -------------------------------------------------------------------------------- 1 | from dify_plugin import Plugin, DifyPluginEnv 2 | 3 | plugin = Plugin(DifyPluginEnv(MAX_REQUEST_TIMEOUT=120)) 4 | 5 | if __name__ == '__main__': 6 | plugin.run() 7 | -------------------------------------------------------------------------------- /manifest.yaml: -------------------------------------------------------------------------------- 1 | version: 0.0.4 2 | type: plugin 3 | author: 3dify-project 4 | name: mcp_client 5 | label: 6 | en_US: mcp_client 7 | ja_JP: mcp_client 8 | zh_Hans: mcp_client 9 | pt_BR: mcp_client 10 | description: 11 | en_US: Claude MCP Client as Dify plugin. This maintains 1:1 connections with MCP servers. By default, list of {tool, resource, prompt} on the server is passed to LLM. Dify is regarded as Host (not MCP server). 12 | ja_JP: Claude MCP Client as Dify plugin. This maintains 1:1 connections with MCP servers. By default, list of {tool, resource, prompt} on the server is passed to LLM. Dify is regarded as Host (not MCP server). 13 | zh_Hans: Claude MCP Client as Dify plugin. This maintains 1:1 connections with MCP servers. By default, list of {tool, resource, prompt} on the server is passed to LLM. Dify is regarded as Host (not MCP server). 14 | pt_BR: Claude MCP Client as Dify plugin. This maintains 1:1 connections with MCP servers. By default, list of {tool, resource, prompt} on the server is passed to LLM. Dify is regarded as Host (not MCP server). 15 | icon: icon.svg 16 | resource: 17 | memory: 268435456 18 | permission: 19 | tool: 20 | enabled: true 21 | model: 22 | enabled: true 23 | llm: true 24 | text_embedding: false 25 | rerank: false 26 | tts: false 27 | speech2text: false 28 | moderation: false 29 | endpoint: 30 | enabled: true 31 | app: 32 | enabled: true 33 | storage: 34 | enabled: true 35 | size: 1048576 36 | plugins: 37 | agent_strategies: 38 | - provider/agent.yaml 39 | meta: 40 | version: 0.0.4 41 | arch: 42 | - amd64 43 | - arm64 44 | runner: 45 | language: python 46 | version: "3.12" 47 | entrypoint: main 48 | created_at: 2025-03-04T16:50:45.7131549+09:00 49 | privacy: PRIVACY.md 50 | verified: false 51 | -------------------------------------------------------------------------------- /output_parser/cot_output_parser.py: -------------------------------------------------------------------------------- 1 | import json 2 | import re 3 | from collections.abc import Generator 4 | from typing import Union 5 | 6 | from dify_plugin.entities.model.llm import LLMResultChunk 7 | from dify_plugin.interfaces.agent import AgentScratchpadUnit 8 | 9 | 10 | class CotAgentOutputParser: 11 | @classmethod 12 | def handle_react_stream_output( 13 | cls, llm_response: Generator[LLMResultChunk, None, None], usage_dict: dict 14 | ) -> Generator[Union[str, AgentScratchpadUnit.Action], None, None]: 15 | def parse_action(json_str): 16 | try: 17 | action = json.loads(json_str, strict=False) 18 | action_name = None 19 | action_input = None 20 | 21 | # cohere always returns a list 22 | if isinstance(action, list) and len(action) == 1: 23 | action = action[0] 24 | 25 | for key, value in action.items(): 26 | if "input" in key.lower(): 27 | action_input = value 28 | else: 29 | action_name = value 30 | 31 | if action_name is not None and action_input is not None: 32 | return AgentScratchpadUnit.Action( 33 | action_name=action_name, 34 | action_input=action_input, 35 | ) 36 | else: 37 | return json_str or "" 38 | except: 39 | return json_str or "" 40 | 41 | def extra_json_from_code_block( 42 | code_block, 43 | ) -> Generator[Union[str, AgentScratchpadUnit.Action], None, None]: 44 | code_blocks = re.findall(r"```(.*?)```", code_block, re.DOTALL) 45 | if not code_blocks: 46 | return 47 | for block in code_blocks: 48 | json_text = re.sub( 49 | r"^[a-zA-Z]+\n", "", block.strip(), flags=re.MULTILINE 50 | ) 51 | yield parse_action(json_text) 52 | 53 | code_block_cache = "" 54 | code_block_delimiter_count = 0 55 | in_code_block = False 56 | json_cache = "" 57 | json_quote_count = 0 58 | in_json = False 59 | got_json = False 60 | 61 | action_cache = "" 62 | action_str = "action:" 63 | action_idx = 0 64 | 65 | thought_cache = "" 66 | thought_str = "thought:" 67 | thought_idx = 0 68 | 69 | last_character = "" 70 | 71 | for response in llm_response: 72 | if response.delta.usage: 73 | usage_dict["usage"] = response.delta.usage 74 | response_content = response.delta.message.content 75 | if not isinstance(response_content, str): 76 | continue 77 | 78 | # stream 79 | index = 0 80 | while index < len(response_content): 81 | steps = 1 82 | delta = response_content[index : index + steps] 83 | yield_delta = False 84 | 85 | if delta == "`": 86 | last_character = delta 87 | code_block_cache += delta 88 | code_block_delimiter_count += 1 89 | else: 90 | if not in_code_block: 91 | if code_block_delimiter_count > 0: 92 | last_character = delta 93 | yield code_block_cache 94 | code_block_cache = "" 95 | else: 96 | last_character = delta 97 | code_block_cache += delta 98 | code_block_delimiter_count = 0 99 | 100 | if not in_code_block and not in_json: 101 | if delta.lower() == action_str[action_idx] and action_idx == 0: 102 | if last_character not in {"\n", " ", ""}: 103 | yield_delta = True 104 | else: 105 | last_character = delta 106 | action_cache += delta 107 | action_idx += 1 108 | if action_idx == len(action_str): 109 | action_cache = "" 110 | action_idx = 0 111 | index += steps 112 | continue 113 | elif delta.lower() == action_str[action_idx] and action_idx > 0: 114 | last_character = delta 115 | action_cache += delta 116 | action_idx += 1 117 | if action_idx == len(action_str): 118 | action_cache = "" 119 | action_idx = 0 120 | index += steps 121 | continue 122 | else: 123 | if action_cache: 124 | last_character = delta 125 | yield action_cache 126 | action_cache = "" 127 | action_idx = 0 128 | 129 | if delta.lower() == thought_str[thought_idx] and thought_idx == 0: 130 | if last_character not in {"\n", " ", ""}: 131 | yield_delta = True 132 | else: 133 | last_character = delta 134 | thought_cache += delta 135 | thought_idx += 1 136 | if thought_idx == len(thought_str): 137 | thought_cache = "" 138 | thought_idx = 0 139 | index += steps 140 | continue 141 | elif delta.lower() == thought_str[thought_idx] and thought_idx > 0: 142 | last_character = delta 143 | thought_cache += delta 144 | thought_idx += 1 145 | if thought_idx == len(thought_str): 146 | thought_cache = "" 147 | thought_idx = 0 148 | index += steps 149 | continue 150 | else: 151 | if thought_cache: 152 | last_character = delta 153 | yield thought_cache 154 | thought_cache = "" 155 | thought_idx = 0 156 | 157 | if yield_delta: 158 | index += steps 159 | last_character = delta 160 | yield delta 161 | continue 162 | 163 | if code_block_delimiter_count == 3: 164 | if in_code_block: 165 | last_character = delta 166 | yield from extra_json_from_code_block(code_block_cache) 167 | code_block_cache = "" 168 | 169 | in_code_block = not in_code_block 170 | code_block_delimiter_count = 0 171 | 172 | if not in_code_block: 173 | # handle single json 174 | if delta == "{": 175 | json_quote_count += 1 176 | in_json = True 177 | last_character = delta 178 | json_cache += delta 179 | elif delta == "}": 180 | last_character = delta 181 | json_cache += delta 182 | if json_quote_count > 0: 183 | json_quote_count -= 1 184 | if json_quote_count == 0: 185 | in_json = False 186 | got_json = True 187 | index += steps 188 | continue 189 | else: 190 | if in_json: 191 | last_character = delta 192 | json_cache += delta 193 | 194 | if got_json: 195 | got_json = False 196 | last_character = delta 197 | yield parse_action(json_cache) 198 | json_cache = "" 199 | json_quote_count = 0 200 | in_json = False 201 | 202 | if not in_code_block and not in_json: 203 | last_character = delta 204 | yield delta.replace("`", "") 205 | 206 | index += steps 207 | 208 | if code_block_cache: 209 | yield code_block_cache 210 | 211 | if json_cache: 212 | yield parse_action(json_cache) 213 | -------------------------------------------------------------------------------- /prompt/template.py: -------------------------------------------------------------------------------- 1 | ENGLISH_REACT_COMPLETION_PROMPT_TEMPLATES = """Respond to the human as helpfully and accurately as possible. 2 | 3 | {{instruction}} 4 | 5 | You have access to the following tools: 6 | 7 | {{tools}} 8 | 9 | Use a json blob to specify a tool by providing an action key (tool name) and an action_input key (tool input). 10 | Valid "action" values: "Final Answer" or {{tool_names}} 11 | 12 | Provide only ONE action per $JSON_BLOB, as shown: 13 | 14 | ``` 15 | { 16 | "action": $TOOL_NAME, 17 | "action_input": $ACTION_INPUT 18 | } 19 | ``` 20 | 21 | Follow this format: 22 | 23 | Question: input question to answer 24 | Thought: consider previous and subsequent steps 25 | Action: 26 | ``` 27 | $JSON_BLOB 28 | ``` 29 | Observation: action result 30 | ... (repeat Thought/Action/Observation N times) 31 | Thought: I know what to respond 32 | Action: 33 | ``` 34 | { 35 | "action": "Final Answer", 36 | "action_input": "Final response to human" 37 | } 38 | ``` 39 | 40 | Begin! Reminder to ALWAYS respond with a valid json blob of a single action. Use tools if necessary. Respond directly if appropriate. Format is Action:```$JSON_BLOB```then Observation:. 41 | {{historic_messages}} 42 | Question: {{query}} 43 | {{agent_scratchpad}} 44 | Thought:""" # noqa: E501 45 | 46 | 47 | ENGLISH_REACT_COMPLETION_AGENT_SCRATCHPAD_TEMPLATES = """Observation: {{observation}} 48 | Thought:""" 49 | 50 | ENGLISH_REACT_CHAT_PROMPT_TEMPLATES = """Respond to the human as helpfully and accurately as possible. 51 | 52 | {{instruction}} 53 | 54 | You have access to the following tools: 55 | 56 | {{tools}} 57 | 58 | Use a json blob to specify a tool by providing an action key (tool name) and an action_input key (tool input). 59 | Valid "action" values: "Final Answer" or {{tool_names}} 60 | 61 | Provide only ONE action per $JSON_BLOB, as shown: 62 | 63 | ``` 64 | { 65 | "action": $TOOL_NAME, 66 | "action_input": $ACTION_INPUT 67 | } 68 | ``` 69 | 70 | Follow this format: 71 | 72 | Question: input question to answer 73 | Thought: consider previous and subsequent steps 74 | Action: 75 | ``` 76 | $JSON_BLOB 77 | ``` 78 | Observation: action result 79 | ... (repeat Thought/Action/Observation N times) 80 | Thought: I know what to respond 81 | Action: 82 | ``` 83 | { 84 | "action": "Final Answer", 85 | "action_input": "Final response to human" 86 | } 87 | ``` 88 | 89 | Begin! Reminder to ALWAYS respond with a valid json blob of a single action. Use tools if necessary. Respond directly if appropriate. Format is Action:```$JSON_BLOB```then Observation:. 90 | """ # noqa: E501 91 | 92 | 93 | ENGLISH_REACT_CHAT_AGENT_SCRATCHPAD_TEMPLATES = "" 94 | 95 | REACT_PROMPT_TEMPLATES = { 96 | "english": { 97 | "chat": { 98 | "prompt": ENGLISH_REACT_CHAT_PROMPT_TEMPLATES, 99 | "agent_scratchpad": ENGLISH_REACT_CHAT_AGENT_SCRATCHPAD_TEMPLATES, 100 | }, 101 | "completion": { 102 | "prompt": ENGLISH_REACT_COMPLETION_PROMPT_TEMPLATES, 103 | "agent_scratchpad": ENGLISH_REACT_COMPLETION_AGENT_SCRATCHPAD_TEMPLATES, 104 | }, 105 | } 106 | } 107 | -------------------------------------------------------------------------------- /provider/agent.py: -------------------------------------------------------------------------------- 1 | from dify_plugin.interfaces.agent import AgentProvider 2 | 3 | 4 | class LanggeniusAgentProvider(AgentProvider): 5 | pass 6 | -------------------------------------------------------------------------------- /provider/agent.yaml: -------------------------------------------------------------------------------- 1 | identity: 2 | author: langgenius 3 | name: agent 4 | label: 5 | en_US: Agent 6 | zh_Hans: Agent 7 | pt_BR: Agent 8 | description: 9 | en_US: Agent 10 | zh_Hans: Agent 11 | pt_BR: Agent 12 | icon: icon.svg 13 | strategies: 14 | - strategies/function_calling.yaml 15 | - strategies/ReAct.yaml 16 | - strategies/mcpReAct.yaml 17 | extra: 18 | python: 19 | source: provider/agent.py 20 | -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | dify_plugin==0.1.0 2 | mcp>=1.6.0 -------------------------------------------------------------------------------- /strategies/ReAct.py: -------------------------------------------------------------------------------- 1 | import json 2 | import time 3 | from collections.abc import Generator, Mapping 4 | from typing import Any, Optional, cast 5 | 6 | from dify_plugin.entities.agent import AgentInvokeMessage 7 | from dify_plugin.entities.model.llm import LLMModelConfig, LLMUsage 8 | from dify_plugin.entities.model.message import ( 9 | AssistantPromptMessage, 10 | PromptMessage, 11 | SystemPromptMessage, 12 | ToolPromptMessage, 13 | UserPromptMessage, 14 | ) 15 | from dify_plugin.entities.tool import ( 16 | LogMetadata, 17 | ToolInvokeMessage, 18 | ToolParameter, 19 | ToolProviderType, 20 | ) 21 | from dify_plugin.interfaces.agent import ( 22 | AgentModelConfig, 23 | AgentScratchpadUnit, 24 | AgentStrategy, 25 | ToolEntity, 26 | ) 27 | from output_parser.cot_output_parser import CotAgentOutputParser 28 | from prompt.template import REACT_PROMPT_TEMPLATES 29 | from pydantic import BaseModel, Field 30 | 31 | ignore_observation_providers = ["wenxin"] 32 | 33 | 34 | class ReActParams(BaseModel): 35 | query: str 36 | instruction: str | None 37 | model: AgentModelConfig 38 | tools: list[ToolEntity] | None 39 | inputs: dict[str, Any] = {} 40 | maximum_iterations: int = 3 41 | 42 | 43 | class AgentPromptEntity(BaseModel): 44 | """ 45 | Agent Prompt Entity. 46 | """ 47 | 48 | first_prompt: str 49 | next_iteration: str 50 | 51 | 52 | class ToolInvokeMeta(BaseModel): 53 | """ 54 | Tool invoke meta 55 | """ 56 | 57 | time_cost: float = Field(..., description="The time cost of the tool invoke") 58 | error: Optional[str] = None 59 | tool_config: Optional[dict] = None 60 | 61 | @classmethod 62 | def empty(cls) -> "ToolInvokeMeta": 63 | """ 64 | Get an empty instance of ToolInvokeMeta 65 | """ 66 | return cls(time_cost=0.0, error=None, tool_config={}) 67 | 68 | @classmethod 69 | def error_instance(cls, error: str) -> "ToolInvokeMeta": 70 | """ 71 | Get an instance of ToolInvokeMeta with error 72 | """ 73 | return cls(time_cost=0.0, error=error, tool_config={}) 74 | 75 | def to_dict(self) -> dict: 76 | return { 77 | "time_cost": self.time_cost, 78 | "error": self.error, 79 | "tool_config": self.tool_config, 80 | } 81 | 82 | 83 | class ReActAgentStrategy(AgentStrategy): 84 | def _invoke(self, parameters: dict[str, Any]) -> Generator[AgentInvokeMessage]: 85 | react_params = ReActParams(**parameters) 86 | query = react_params.query 87 | model = react_params.model 88 | agent_scratchpad = [] 89 | history_prompt_messages: list[PromptMessage] = [] 90 | current_session_messages = [] 91 | self._organize_historic_prompt_messages( 92 | history_prompt_messages, current_session_messages=current_session_messages 93 | ) 94 | tools = react_params.tools 95 | tool_instances = {tool.identity.name: tool for tool in tools} if tools else {} 96 | react_params.model.completion_params = ( 97 | react_params.model.completion_params or {} 98 | ) 99 | # check model mode 100 | stop = ( 101 | react_params.model.completion_params.get("stop", []) 102 | if react_params.model.completion_params 103 | else [] 104 | ) 105 | 106 | if ( 107 | "Observation" not in stop 108 | and model.provider not in ignore_observation_providers 109 | ): 110 | stop.append("Observation") 111 | # init instruction 112 | inputs = react_params.inputs 113 | instruction = react_params.instruction or "" 114 | self._instruction = self._fill_in_inputs_from_external_data_tools( 115 | instruction, inputs 116 | ) 117 | 118 | iteration_step = 1 119 | max_iteration_steps = react_params.maximum_iterations 120 | 121 | # convert tools into ModelRuntime Tool format 122 | prompt_messages_tools = self._init_prompt_tools(tools) 123 | self._prompt_messages_tools = prompt_messages_tools 124 | 125 | run_agent_state = True 126 | llm_usage: dict[str, Optional[LLMUsage]] = {"usage": None} 127 | final_answer = "" 128 | prompt_messages = [] 129 | while run_agent_state and iteration_step <= max_iteration_steps: 130 | # continue to run until there is not any tool call 131 | run_agent_state = False 132 | round_started_at = time.perf_counter() 133 | round_log = self.create_log_message( 134 | label=f"ROUND {iteration_step}", 135 | data={}, 136 | metadata={ 137 | LogMetadata.STARTED_AT: round_started_at, 138 | }, 139 | status=ToolInvokeMessage.LogMessage.LogStatus.START, 140 | ) 141 | yield round_log 142 | if iteration_step == max_iteration_steps: 143 | # the last iteration, remove all tools 144 | self._prompt_messages_tools = [] 145 | 146 | message_file_ids: list[str] = [] 147 | 148 | # recalc llm max tokens 149 | prompt_messages = self._organize_prompt_messages(agent_scratchpad, query) 150 | if model.completion_params: 151 | self.recalc_llm_max_tokens( 152 | model.entity, prompt_messages, model.completion_params 153 | ) 154 | # invoke model 155 | chunks = self.session.model.llm.invoke( 156 | model_config=LLMModelConfig(**model.model_dump(mode="json")), 157 | prompt_messages=prompt_messages, 158 | stream=True, 159 | stop=stop, 160 | ) 161 | 162 | usage_dict = {} 163 | react_chunks = CotAgentOutputParser.handle_react_stream_output( 164 | chunks, usage_dict 165 | ) 166 | scratchpad = AgentScratchpadUnit( 167 | agent_response="", 168 | thought="", 169 | action_str="", 170 | observation="", 171 | action=None, 172 | ) 173 | 174 | model_started_at = time.perf_counter() 175 | model_log = self.create_log_message( 176 | label=f"{model.model} Thought", 177 | data={}, 178 | metadata={ 179 | LogMetadata.STARTED_AT: model_started_at, 180 | LogMetadata.PROVIDER: model.provider, 181 | }, 182 | parent=round_log, 183 | status=ToolInvokeMessage.LogMessage.LogStatus.START, 184 | ) 185 | yield model_log 186 | 187 | for chunk in react_chunks: 188 | if isinstance(chunk, AgentScratchpadUnit.Action): 189 | action = chunk 190 | # detect action 191 | assert scratchpad.agent_response is not None 192 | scratchpad.agent_response += json.dumps(chunk.model_dump()) 193 | 194 | scratchpad.action_str = json.dumps(chunk.model_dump()) 195 | scratchpad.action = action 196 | else: 197 | scratchpad.agent_response = scratchpad.agent_response or "" 198 | scratchpad.thought = scratchpad.thought or "" 199 | scratchpad.agent_response += chunk 200 | scratchpad.thought += chunk 201 | scratchpad.thought = ( 202 | scratchpad.thought.strip() 203 | if scratchpad.thought 204 | else "I am thinking about how to help you" 205 | ) 206 | agent_scratchpad.append(scratchpad) 207 | 208 | # get llm usage 209 | if "usage" in usage_dict: 210 | if usage_dict["usage"] is not None: 211 | self.increase_usage(llm_usage, usage_dict["usage"]) 212 | else: 213 | usage_dict["usage"] = LLMUsage.empty_usage() 214 | 215 | action = ( 216 | scratchpad.action.to_dict() 217 | if scratchpad.action 218 | else {"action": scratchpad.agent_response} 219 | ) 220 | 221 | yield self.finish_log_message( 222 | log=model_log, 223 | data={"thought": scratchpad.thought, **action}, 224 | metadata={ 225 | LogMetadata.STARTED_AT: model_started_at, 226 | LogMetadata.FINISHED_AT: time.perf_counter(), 227 | LogMetadata.ELAPSED_TIME: time.perf_counter() - model_started_at, 228 | LogMetadata.PROVIDER: model.provider, 229 | LogMetadata.TOTAL_PRICE: usage_dict["usage"].total_price 230 | if usage_dict["usage"] 231 | else 0, 232 | LogMetadata.CURRENCY: usage_dict["usage"].currency 233 | if usage_dict["usage"] 234 | else "", 235 | LogMetadata.TOTAL_TOKENS: usage_dict["usage"].total_tokens 236 | if usage_dict["usage"] 237 | else 0, 238 | }, 239 | ) 240 | if not scratchpad.action: 241 | final_answer = scratchpad.thought 242 | else: 243 | if scratchpad.action.action_name.lower() == "final answer": 244 | # action is final answer, return final answer directly 245 | try: 246 | if isinstance(scratchpad.action.action_input, dict): 247 | final_answer = json.dumps(scratchpad.action.action_input) 248 | elif isinstance(scratchpad.action.action_input, str): 249 | final_answer = scratchpad.action.action_input 250 | else: 251 | final_answer = f"{scratchpad.action.action_input}" 252 | except json.JSONDecodeError: 253 | final_answer = f"{scratchpad.action.action_input}" 254 | else: 255 | run_agent_state = True 256 | # action is tool call, invoke tool 257 | tool_call_started_at = time.perf_counter() 258 | tool_name = scratchpad.action.action_name 259 | tool_call_log = self.create_log_message( 260 | label=f"CALL {tool_name}", 261 | data={}, 262 | metadata={ 263 | LogMetadata.STARTED_AT: time.perf_counter(), 264 | LogMetadata.PROVIDER: tool_instances[ 265 | tool_name 266 | ].identity.provider 267 | if tool_instances.get(tool_name) 268 | else "", 269 | }, 270 | parent=round_log, 271 | status=ToolInvokeMessage.LogMessage.LogStatus.START, 272 | ) 273 | yield tool_call_log 274 | tool_invoke_response, tool_invoke_parameters = ( 275 | self._handle_invoke_action( 276 | action=scratchpad.action, 277 | tool_instances=tool_instances, 278 | message_file_ids=message_file_ids, 279 | ) 280 | ) 281 | scratchpad.observation = tool_invoke_response 282 | scratchpad.agent_response = tool_invoke_response 283 | yield self.finish_log_message( 284 | log=tool_call_log, 285 | data={ 286 | "tool_name": tool_name, 287 | "tool_call_args": tool_invoke_parameters, 288 | "output": tool_invoke_response, 289 | }, 290 | metadata={ 291 | LogMetadata.STARTED_AT: tool_call_started_at, 292 | LogMetadata.PROVIDER: tool_instances[ 293 | tool_name 294 | ].identity.provider 295 | if tool_instances.get(tool_name) 296 | else "", 297 | LogMetadata.FINISHED_AT: time.perf_counter(), 298 | LogMetadata.ELAPSED_TIME: time.perf_counter() 299 | - tool_call_started_at, 300 | }, 301 | ) 302 | 303 | # update prompt tool message 304 | for prompt_tool in self._prompt_messages_tools: 305 | self.update_prompt_message_tool( 306 | tool_instances[prompt_tool.name], prompt_tool 307 | ) 308 | yield self.finish_log_message( 309 | log=round_log, 310 | data={ 311 | "action_name": scratchpad.action.action_name 312 | if scratchpad.action 313 | else "", 314 | "action_input": scratchpad.action.action_input 315 | if scratchpad.action 316 | else "", 317 | "thought": scratchpad.thought, 318 | "observation": scratchpad.observation, 319 | }, 320 | metadata={ 321 | LogMetadata.STARTED_AT: round_started_at, 322 | LogMetadata.FINISHED_AT: time.perf_counter(), 323 | LogMetadata.ELAPSED_TIME: time.perf_counter() - round_started_at, 324 | LogMetadata.TOTAL_PRICE: usage_dict["usage"].total_price 325 | if usage_dict["usage"] 326 | else 0, 327 | LogMetadata.CURRENCY: usage_dict["usage"].currency 328 | if usage_dict["usage"] 329 | else "", 330 | LogMetadata.TOTAL_TOKENS: usage_dict["usage"].total_tokens 331 | if usage_dict["usage"] 332 | else 0, 333 | }, 334 | ) 335 | iteration_step += 1 336 | 337 | yield self.create_text_message(final_answer) 338 | yield self.create_json_message( 339 | { 340 | "execution_metadata": { 341 | LogMetadata.TOTAL_PRICE: llm_usage["usage"].total_price 342 | if llm_usage["usage"] is not None 343 | else 0, 344 | LogMetadata.CURRENCY: llm_usage["usage"].currency 345 | if llm_usage["usage"] is not None 346 | else "", 347 | LogMetadata.TOTAL_TOKENS: llm_usage["usage"].total_tokens 348 | if llm_usage["usage"] is not None 349 | else 0, 350 | } 351 | } 352 | ) 353 | 354 | def _organize_system_prompt(self) -> SystemPromptMessage: 355 | """ 356 | Organize system prompt 357 | """ 358 | 359 | prompt_entity = AgentPromptEntity( 360 | first_prompt=REACT_PROMPT_TEMPLATES["english"]["chat"]["prompt"], 361 | next_iteration=REACT_PROMPT_TEMPLATES["english"]["chat"][ 362 | "agent_scratchpad" 363 | ], 364 | ) 365 | if not prompt_entity: 366 | raise ValueError("Agent prompt configuration is not set") 367 | first_prompt = prompt_entity.first_prompt 368 | 369 | system_prompt = ( 370 | first_prompt.replace("{{instruction}}", self._instruction) 371 | .replace( 372 | "{{tools}}", 373 | json.dumps( 374 | [ 375 | tool.model_dump(mode="json") 376 | for tool in self._prompt_messages_tools 377 | ] 378 | ), 379 | ) 380 | .replace( 381 | "{{tool_names}}", 382 | ", ".join([tool.name for tool in self._prompt_messages_tools]), 383 | ) 384 | ) 385 | 386 | return SystemPromptMessage(content=system_prompt) 387 | 388 | def _organize_user_query( 389 | self, query, prompt_messages: list[PromptMessage] 390 | ) -> list[PromptMessage]: 391 | """ 392 | Organize user query 393 | """ 394 | prompt_messages.append(UserPromptMessage(content=query)) 395 | 396 | return prompt_messages 397 | 398 | def _organize_prompt_messages( 399 | self, agent_scratchpad: list, query: str 400 | ) -> list[PromptMessage]: 401 | """ 402 | Organize 403 | """ 404 | # organize system prompt 405 | system_message = self._organize_system_prompt() 406 | 407 | # organize current assistant messages 408 | agent_scratchpad = agent_scratchpad 409 | if not agent_scratchpad: 410 | assistant_messages = [] 411 | else: 412 | assistant_message = AssistantPromptMessage(content="") 413 | assistant_message.content = ( 414 | "" # FIXME: type check tell mypy that assistant_message.content is str 415 | ) 416 | for unit in agent_scratchpad: 417 | if unit.is_final(): 418 | assert isinstance(assistant_message.content, str) 419 | assistant_message.content += f"Final Answer: {unit.agent_response}" 420 | else: 421 | assert isinstance(assistant_message.content, str) 422 | assistant_message.content += f"Thought: {unit.thought}\n\n" 423 | if unit.action_str: 424 | assistant_message.content += f"Action: {unit.action_str}\n\n" 425 | if unit.observation: 426 | assistant_message.content += ( 427 | f"Observation: {unit.observation}\n\n" 428 | ) 429 | 430 | assistant_messages = [assistant_message] 431 | 432 | # query messages 433 | query_messages = self._organize_user_query(query, []) 434 | 435 | if assistant_messages: 436 | # organize historic prompt messages 437 | historic_messages = self._organize_historic_prompt_messages( 438 | [ 439 | system_message, 440 | *query_messages, 441 | *assistant_messages, 442 | UserPromptMessage(content="continue"), 443 | ] 444 | ) 445 | messages = [ 446 | system_message, 447 | *historic_messages, 448 | *query_messages, 449 | *assistant_messages, 450 | UserPromptMessage(content="continue"), 451 | ] 452 | else: 453 | # organize historic prompt messages 454 | historic_messages = self._organize_historic_prompt_messages( 455 | [system_message, *query_messages] 456 | ) 457 | messages = [system_message, *historic_messages, *query_messages] 458 | 459 | # join all messages 460 | return messages 461 | 462 | def _handle_invoke_action( 463 | self, 464 | action: AgentScratchpadUnit.Action, 465 | tool_instances: Mapping[str, ToolEntity], 466 | message_file_ids: list[str], 467 | ) -> tuple[str, dict[str, Any] | str]: 468 | """ 469 | handle invoke action 470 | :param action: action 471 | :param tool_instances: tool instances 472 | :param message_file_ids: message file ids 473 | :param trace_manager: trace manager 474 | :return: observation, meta 475 | """ 476 | # action is tool call, invoke tool 477 | tool_call_name = action.action_name 478 | tool_call_args = action.action_input 479 | tool_instance = tool_instances.get(tool_call_name) 480 | 481 | if not tool_instance: 482 | answer = f"there is not a tool named {tool_call_name}" 483 | return answer, tool_call_args 484 | 485 | if isinstance(tool_call_args, str): 486 | try: 487 | tool_call_args = json.loads(tool_call_args) 488 | except json.JSONDecodeError as e: 489 | params = [ 490 | param.name 491 | for param in tool_instance.parameters 492 | if param.form == ToolParameter.ToolParameterForm.LLM 493 | ] 494 | if len(params) > 1: 495 | raise ValueError("tool call args is not a valid json string") from e 496 | tool_call_args = {params[0]: tool_call_args} if len(params) == 1 else {} 497 | 498 | tool_invoke_parameters = {**tool_instance.runtime_parameters, **tool_call_args} 499 | try: 500 | tool_invoke_responses = self.session.tool.invoke( 501 | provider_type=ToolProviderType(tool_instance.provider_type), 502 | provider=tool_instance.identity.provider, 503 | tool_name=tool_instance.identity.name, 504 | parameters=tool_invoke_parameters, 505 | ) 506 | result = "" 507 | for response in tool_invoke_responses: 508 | if response.type == ToolInvokeMessage.MessageType.TEXT: 509 | result += cast(ToolInvokeMessage.TextMessage, response.message).text 510 | elif response.type == ToolInvokeMessage.MessageType.LINK: 511 | result += ( 512 | f"result link: {cast(ToolInvokeMessage.TextMessage, response.message).text}." 513 | + " please tell user to check it." 514 | ) 515 | elif response.type in { 516 | ToolInvokeMessage.MessageType.IMAGE_LINK, 517 | ToolInvokeMessage.MessageType.IMAGE, 518 | }: 519 | result += ( 520 | "image has been created and sent to user already, " 521 | + "you do not need to create it, just tell the user to check it now." 522 | ) 523 | elif response.type == ToolInvokeMessage.MessageType.JSON: 524 | text = json.dumps( 525 | cast( 526 | ToolInvokeMessage.JsonMessage, response.message 527 | ).json_object, 528 | ensure_ascii=False, 529 | ) 530 | result += f"tool response: {text}." 531 | else: 532 | result += f"tool response: {response.message!r}." 533 | except Exception as e: 534 | result = f"tool invoke error: {str(e)}" 535 | 536 | return result, tool_invoke_parameters 537 | 538 | def _convert_dict_to_action(self, action: dict) -> AgentScratchpadUnit.Action: 539 | """ 540 | convert dict to action 541 | """ 542 | return AgentScratchpadUnit.Action( 543 | action_name=action["action"], action_input=action["action_input"] 544 | ) 545 | 546 | def _fill_in_inputs_from_external_data_tools( 547 | self, instruction: str, inputs: Mapping[str, Any] 548 | ) -> str: 549 | """ 550 | fill in inputs from external data tools 551 | """ 552 | for key, value in inputs.items(): 553 | try: 554 | instruction = instruction.replace(f"{{{{{key}}}}}", str(value)) 555 | except Exception: 556 | continue 557 | 558 | return instruction 559 | 560 | def _format_assistant_message( 561 | self, agent_scratchpad: list[AgentScratchpadUnit] 562 | ) -> str: 563 | """ 564 | format assistant message 565 | """ 566 | message = "" 567 | for scratchpad in agent_scratchpad: 568 | if scratchpad.is_final(): 569 | message += f"Final Answer: {scratchpad.agent_response}" 570 | else: 571 | message += f"Thought: {scratchpad.thought}\n\n" 572 | if scratchpad.action_str: 573 | message += f"Action: {scratchpad.action_str}\n\n" 574 | if scratchpad.observation: 575 | message += f"Observation: {scratchpad.observation}\n\n" 576 | 577 | return message 578 | 579 | def _organize_historic_prompt_messages( 580 | self, 581 | history_prompt_messages: list[PromptMessage], 582 | current_session_messages: list[PromptMessage] | None = None, 583 | ) -> list[PromptMessage]: 584 | """ 585 | organize historic prompt messages 586 | """ 587 | result: list[PromptMessage] = [] 588 | scratchpads: list[AgentScratchpadUnit] = [] 589 | current_scratchpad: AgentScratchpadUnit | None = None 590 | 591 | for message in history_prompt_messages: 592 | if isinstance(message, AssistantPromptMessage): 593 | if not current_scratchpad: 594 | assert isinstance(message.content, str) 595 | current_scratchpad = AgentScratchpadUnit( 596 | agent_response=message.content, 597 | thought=message.content 598 | or "I am thinking about how to help you", 599 | action_str="", 600 | action=None, 601 | observation=None, 602 | ) 603 | scratchpads.append(current_scratchpad) 604 | if message.tool_calls: 605 | try: 606 | current_scratchpad.action = AgentScratchpadUnit.Action( 607 | action_name=message.tool_calls[0].function.name, 608 | action_input=json.loads( 609 | message.tool_calls[0].function.arguments 610 | ), 611 | ) 612 | current_scratchpad.action_str = json.dumps( 613 | current_scratchpad.action.to_dict() 614 | ) 615 | except Exception: 616 | pass 617 | elif isinstance(message, ToolPromptMessage): 618 | if current_scratchpad: 619 | assert isinstance(message.content, str) 620 | current_scratchpad.observation = message.content 621 | else: 622 | raise NotImplementedError("expected str type") 623 | elif isinstance(message, UserPromptMessage): 624 | if scratchpads: 625 | result.append( 626 | AssistantPromptMessage( 627 | content=self._format_assistant_message(scratchpads) 628 | ) 629 | ) 630 | scratchpads = [] 631 | current_scratchpad = None 632 | 633 | result.append(message) 634 | 635 | if scratchpads: 636 | result.append( 637 | AssistantPromptMessage( 638 | content=self._format_assistant_message(scratchpads) 639 | ) 640 | ) 641 | 642 | return current_session_messages or [] 643 | -------------------------------------------------------------------------------- /strategies/ReAct.yaml: -------------------------------------------------------------------------------- 1 | identity: 2 | name: ReAct 3 | author: Dify 4 | label: 5 | en_US: ReAct 6 | zh_Hans: ReAct 7 | pt_BR: ReAct 8 | description: 9 | en_US: ReAct is a basic strategy for agent, model will use the tools provided to perform the task. 10 | zh_Hans: ReAct 是一个基本的 Agent 策略，模型将使用提供的工具来执行任务。 11 | pt_BR: ReAct is a basic strategy for agent, model will use the tools provided to perform the task. 12 | parameters: 13 | - name: model 14 | type: model-selector 15 | scope: tool-call&llm 16 | required: true 17 | label: 18 | en_US: Model 19 | zh_Hans: 模型 20 | pt_BR: Model 21 | - name: tools 22 | type: array[tools] 23 | required: true 24 | label: 25 | en_US: Tools list 26 | zh_Hans: 工具列表 27 | pt_BR: Tools list 28 | - name: instruction 29 | type: string 30 | required: true 31 | label: 32 | en_US: Instruction 33 | zh_Hans: 指令 34 | pt_BR: Instruction 35 | auto_generate: 36 | type: prompt_instruction 37 | template: 38 | enabled: true 39 | - name: query 40 | type: string 41 | required: true 42 | label: 43 | en_US: Query 44 | zh_Hans: 查询 45 | pt_BR: Query 46 | - name: maximum_iterations 47 | type: number 48 | required: true 49 | label: 50 | en_US: Maxium Iterations 51 | zh_Hans: 最大迭代次数 52 | pt_BR: Maxium Iterations 53 | default: 3 54 | min: 1 55 | max: 30 56 | extra: 57 | python: 58 | source: strategies/ReAct.py 59 | -------------------------------------------------------------------------------- /strategies/function_calling.py: -------------------------------------------------------------------------------- 1 | import json 2 | import time 3 | from collections.abc import Generator 4 | from copy import deepcopy 5 | from typing import Any, Optional, cast 6 | 7 | from dify_plugin.entities.agent import AgentInvokeMessage 8 | from dify_plugin.entities.model.llm import ( 9 | LLMModelConfig, 10 | LLMResult, 11 | LLMResultChunk, 12 | LLMUsage, 13 | ) 14 | from dify_plugin.entities.model.message import ( 15 | AssistantPromptMessage, 16 | PromptMessage, 17 | PromptMessageContentType, 18 | PromptMessageRole, 19 | SystemPromptMessage, 20 | ToolPromptMessage, 21 | UserPromptMessage, 22 | ) 23 | from dify_plugin.entities.tool import LogMetadata, ToolInvokeMessage, ToolProviderType 24 | from dify_plugin.interfaces.agent import AgentModelConfig, AgentStrategy, ToolEntity 25 | from pydantic import BaseModel, Field 26 | 27 | 28 | class FunctionCallingParams(BaseModel): 29 | query: str 30 | instruction: str | None 31 | model: AgentModelConfig 32 | tools: list[ToolEntity] | None 33 | maximum_iterations: int = 3 34 | 35 | 36 | class ToolInvokeMeta(BaseModel): 37 | """ 38 | Tool invoke meta 39 | """ 40 | 41 | time_cost: float = Field(..., description="The time cost of the tool invoke") 42 | error: Optional[str] = None 43 | tool_config: Optional[dict] = None 44 | 45 | @classmethod 46 | def empty(cls) -> "ToolInvokeMeta": 47 | """ 48 | Get an empty instance of ToolInvokeMeta 49 | """ 50 | return cls(time_cost=0.0, error=None, tool_config={}) 51 | 52 | @classmethod 53 | def error_instance(cls, error: str) -> "ToolInvokeMeta": 54 | """ 55 | Get an instance of ToolInvokeMeta with error 56 | """ 57 | return cls(time_cost=0.0, error=error, tool_config={}) 58 | 59 | def to_dict(self) -> dict: 60 | return { 61 | "time_cost": self.time_cost, 62 | "error": self.error, 63 | "tool_config": self.tool_config, 64 | } 65 | 66 | 67 | class FunctionCallingAgentStrategy(AgentStrategy): 68 | def __init__(self, session): 69 | super().__init__(session) 70 | self.query = "" 71 | 72 | def _invoke(self, parameters: dict[str, Any]) -> Generator[AgentInvokeMessage]: 73 | """ 74 | Run FunctionCall agent application 75 | """ 76 | fc_params = FunctionCallingParams(**parameters) 77 | query = fc_params.query 78 | self.query = query 79 | instruction = fc_params.instruction 80 | init_prompt_messages = [ 81 | PromptMessage(role=PromptMessageRole.SYSTEM, content=instruction) 82 | ] 83 | tools = fc_params.tools 84 | tool_instances = {tool.identity.name: tool for tool in tools} if tools else {} 85 | model = fc_params.model 86 | stop = ( 87 | fc_params.model.completion_params.get("stop", []) 88 | if fc_params.model.completion_params 89 | else [] 90 | ) 91 | # convert tools into ModelRuntime Tool format 92 | prompt_messages_tools = self._init_prompt_tools(tools) 93 | 94 | iteration_step = 1 95 | max_iteration_steps = fc_params.maximum_iterations 96 | current_thoughts: list[PromptMessage] = [] 97 | # continue to run until there is not any tool call 98 | function_call_state = True 99 | llm_usage: dict[str, Optional[LLMUsage]] = {"usage": None} 100 | final_answer = "" 101 | 102 | while function_call_state and iteration_step <= max_iteration_steps: 103 | function_call_state = False 104 | round_started_at = time.perf_counter() 105 | round_log = self.create_log_message( 106 | label=f"ROUND {iteration_step}", 107 | data={}, 108 | metadata={ 109 | LogMetadata.STARTED_AT: round_started_at, 110 | }, 111 | status=ToolInvokeMessage.LogMessage.LogStatus.START, 112 | ) 113 | yield round_log 114 | if iteration_step == max_iteration_steps: 115 | # the last iteration, remove all tools 116 | prompt_messages_tools = [] 117 | 118 | # recalc llm max tokens 119 | prompt_messages = self._organize_prompt_messages( 120 | history_prompt_messages=init_prompt_messages, 121 | current_thoughts=current_thoughts, 122 | ) 123 | if model.completion_params: 124 | self.recalc_llm_max_tokens( 125 | model.entity, prompt_messages, model.completion_params 126 | ) 127 | # invoke model 128 | model_started_at = time.perf_counter() 129 | model_log = self.create_log_message( 130 | label=f"{model.model} Thought", 131 | data={}, 132 | metadata={ 133 | LogMetadata.STARTED_AT: model_started_at, 134 | LogMetadata.PROVIDER: model.provider, 135 | }, 136 | parent=round_log, 137 | status=ToolInvokeMessage.LogMessage.LogStatus.START, 138 | ) 139 | yield model_log 140 | chunks: Generator[LLMResultChunk, None, None] | LLMResult = ( 141 | self.session.model.llm.invoke( 142 | model_config=LLMModelConfig(**model.model_dump(mode="json")), 143 | prompt_messages=prompt_messages, 144 | stream=True, 145 | stop=stop, 146 | tools=prompt_messages_tools, 147 | ) 148 | ) 149 | 150 | tool_calls: list[tuple[str, str, dict[str, Any]]] = [] 151 | 152 | # save full response 153 | response = "" 154 | 155 | # save tool call names and inputs 156 | tool_call_names = "" 157 | 158 | current_llm_usage = None 159 | 160 | if isinstance(chunks, Generator): 161 | for chunk in chunks: 162 | # check if there is any tool call 163 | if self.check_tool_calls(chunk): 164 | function_call_state = True 165 | tool_calls.extend(self.extract_tool_calls(chunk) or []) 166 | tool_call_names = ";".join( 167 | [tool_call[1] for tool_call in tool_calls] 168 | ) 169 | 170 | if chunk.delta.message and chunk.delta.message.content: 171 | if isinstance(chunk.delta.message.content, list): 172 | for content in chunk.delta.message.content: 173 | response += content.data 174 | if ( 175 | not function_call_state 176 | or iteration_step == max_iteration_steps 177 | ): 178 | yield self.create_text_message(content.data) 179 | else: 180 | response += str(chunk.delta.message.content) 181 | if ( 182 | not function_call_state 183 | or iteration_step == max_iteration_steps 184 | ): 185 | yield self.create_text_message( 186 | str(chunk.delta.message.content) 187 | ) 188 | 189 | if chunk.delta.usage: 190 | self.increase_usage(llm_usage, chunk.delta.usage) 191 | current_llm_usage = chunk.delta.usage 192 | 193 | else: 194 | result = chunks 195 | # check if there is any tool call 196 | if self.check_blocking_tool_calls(result): 197 | function_call_state = True 198 | tool_calls.extend(self.extract_blocking_tool_calls(result) or []) 199 | tool_call_names = ";".join( 200 | [tool_call[1] for tool_call in tool_calls] 201 | ) 202 | 203 | if result.usage: 204 | self.increase_usage(llm_usage, result.usage) 205 | current_llm_usage = result.usage 206 | 207 | if result.message and result.message.content: 208 | if isinstance(result.message.content, list): 209 | for content in result.message.content: 210 | response += content.data 211 | else: 212 | response += str(result.message.content) 213 | 214 | if not result.message.content: 215 | result.message.content = "" 216 | yield self.finish_log_message( 217 | log=model_log, 218 | data={ 219 | "output": response, 220 | "tool_name": tool_call_names, 221 | "tool_input": { 222 | tool_call[1]: tool_call[2] for tool_call in tool_calls 223 | }, 224 | }, 225 | metadata={ 226 | LogMetadata.STARTED_AT: model_started_at, 227 | LogMetadata.FINISHED_AT: time.perf_counter(), 228 | LogMetadata.ELAPSED_TIME: time.perf_counter() - model_started_at, 229 | LogMetadata.PROVIDER: model.provider, 230 | LogMetadata.TOTAL_PRICE: current_llm_usage.total_price 231 | if current_llm_usage 232 | else 0, 233 | LogMetadata.CURRENCY: current_llm_usage.currency 234 | if current_llm_usage 235 | else "", 236 | LogMetadata.TOTAL_TOKENS: current_llm_usage.total_tokens 237 | if current_llm_usage 238 | else 0, 239 | }, 240 | ) 241 | assistant_message = AssistantPromptMessage(content="", tool_calls=[]) 242 | if tool_calls: 243 | assistant_message.tool_calls = [ 244 | AssistantPromptMessage.ToolCall( 245 | id=tool_call[0], 246 | type="function", 247 | function=AssistantPromptMessage.ToolCall.ToolCallFunction( 248 | name=tool_call[1], 249 | arguments=json.dumps(tool_call[2], ensure_ascii=False), 250 | ), 251 | ) 252 | for tool_call in tool_calls 253 | ] 254 | else: 255 | assistant_message.content = response 256 | 257 | current_thoughts.append(assistant_message) 258 | 259 | final_answer += response + "\n" 260 | 261 | # call tools 262 | tool_responses = [] 263 | for tool_call_id, tool_call_name, tool_call_args in tool_calls: 264 | tool_instance = tool_instances[tool_call_name] 265 | tool_call_started_at = time.perf_counter() 266 | tool_call_log = self.create_log_message( 267 | label=f"CALL {tool_call_name}", 268 | data={}, 269 | metadata={ 270 | LogMetadata.STARTED_AT: time.perf_counter(), 271 | LogMetadata.PROVIDER: tool_instance.identity.provider, 272 | }, 273 | parent=round_log, 274 | status=ToolInvokeMessage.LogMessage.LogStatus.START, 275 | ) 276 | yield tool_call_log 277 | if not tool_instance: 278 | tool_response = { 279 | "tool_call_id": tool_call_id, 280 | "tool_call_name": tool_call_name, 281 | "tool_response": f"there is not a tool named {tool_call_name}", 282 | "meta": ToolInvokeMeta.error_instance( 283 | f"there is not a tool named {tool_call_name}" 284 | ).to_dict(), 285 | } 286 | else: 287 | # invoke tool 288 | try: 289 | tool_invoke_responses = self.session.tool.invoke( 290 | provider_type=ToolProviderType(tool_instance.provider_type), 291 | provider=tool_instance.identity.provider, 292 | tool_name=tool_instance.identity.name, 293 | parameters={ 294 | **tool_instance.runtime_parameters, 295 | **tool_call_args, 296 | }, 297 | ) 298 | result = "" 299 | for response in tool_invoke_responses: 300 | if response.type == ToolInvokeMessage.MessageType.TEXT: 301 | result += cast( 302 | ToolInvokeMessage.TextMessage, response.message 303 | ).text 304 | elif response.type == ToolInvokeMessage.MessageType.LINK: 305 | result += ( 306 | f"result link: {cast(ToolInvokeMessage.TextMessage, response.message).text}." 307 | + " please tell user to check it." 308 | ) 309 | elif response.type in { 310 | ToolInvokeMessage.MessageType.IMAGE_LINK, 311 | ToolInvokeMessage.MessageType.IMAGE, 312 | }: 313 | result += ( 314 | "image has been created and sent to user already, " 315 | + "you do not need to create it, just tell the user to check it now." 316 | ) 317 | elif response.type == ToolInvokeMessage.MessageType.JSON: 318 | text = json.dumps( 319 | cast( 320 | ToolInvokeMessage.JsonMessage, response.message 321 | ).json_object, 322 | ensure_ascii=False, 323 | ) 324 | result += f"tool response: {text}." 325 | else: 326 | result += f"tool response: {response.message!r}." 327 | except Exception as e: 328 | result = f"tool invoke error: {str(e)}" 329 | tool_response = { 330 | "tool_call_id": tool_call_id, 331 | "tool_call_name": tool_call_name, 332 | "tool_call_input": { 333 | **tool_instance.runtime_parameters, 334 | **tool_call_args, 335 | }, 336 | "tool_response": result, 337 | } 338 | 339 | yield self.finish_log_message( 340 | log=tool_call_log, 341 | data={ 342 | "output": tool_response, 343 | }, 344 | metadata={ 345 | LogMetadata.STARTED_AT: tool_call_started_at, 346 | LogMetadata.PROVIDER: tool_instance.identity.provider, 347 | LogMetadata.FINISHED_AT: time.perf_counter(), 348 | LogMetadata.ELAPSED_TIME: time.perf_counter() 349 | - tool_call_started_at, 350 | }, 351 | ) 352 | tool_responses.append(tool_response) 353 | if tool_response["tool_response"] is not None: 354 | current_thoughts.append( 355 | ToolPromptMessage( 356 | content=str(tool_response["tool_response"]), 357 | tool_call_id=tool_call_id, 358 | name=tool_call_name, 359 | ) 360 | ) 361 | 362 | # update prompt tool 363 | for prompt_tool in prompt_messages_tools: 364 | self.update_prompt_message_tool( 365 | tool_instances[prompt_tool.name], prompt_tool 366 | ) 367 | yield self.finish_log_message( 368 | log=round_log, 369 | data={ 370 | "output": { 371 | "llm_response": response, 372 | "tool_responses": tool_responses, 373 | }, 374 | }, 375 | metadata={ 376 | LogMetadata.STARTED_AT: round_started_at, 377 | LogMetadata.FINISHED_AT: time.perf_counter(), 378 | LogMetadata.ELAPSED_TIME: time.perf_counter() - round_started_at, 379 | LogMetadata.TOTAL_PRICE: current_llm_usage.total_price 380 | if current_llm_usage 381 | else 0, 382 | LogMetadata.CURRENCY: current_llm_usage.currency 383 | if current_llm_usage 384 | else "", 385 | LogMetadata.TOTAL_TOKENS: current_llm_usage.total_tokens 386 | if current_llm_usage 387 | else 0, 388 | }, 389 | ) 390 | iteration_step += 1 391 | 392 | yield self.create_json_message( 393 | { 394 | "execution_metadata": { 395 | LogMetadata.TOTAL_PRICE: llm_usage["usage"].total_price 396 | if llm_usage["usage"] is not None 397 | else 0, 398 | LogMetadata.CURRENCY: llm_usage["usage"].currency 399 | if llm_usage["usage"] is not None 400 | else "", 401 | LogMetadata.TOTAL_TOKENS: llm_usage["usage"].total_tokens 402 | if llm_usage["usage"] is not None 403 | else 0, 404 | } 405 | } 406 | ) 407 | 408 | def check_tool_calls(self, llm_result_chunk: LLMResultChunk) -> bool: 409 | """ 410 | Check if there is any tool call in llm result chunk 411 | """ 412 | return bool(llm_result_chunk.delta.message.tool_calls) 413 | 414 | def check_blocking_tool_calls(self, llm_result: LLMResult) -> bool: 415 | """ 416 | Check if there is any blocking tool call in llm result 417 | """ 418 | return bool(llm_result.message.tool_calls) 419 | 420 | def extract_tool_calls( 421 | self, llm_result_chunk: LLMResultChunk 422 | ) -> list[tuple[str, str, dict[str, Any]]]: 423 | """ 424 | Extract tool calls from llm result chunk 425 | 426 | Returns: 427 | List[Tuple[str, str, Dict[str, Any]]]: [(tool_call_id, tool_call_name, tool_call_args)] 428 | """ 429 | tool_calls = [] 430 | for prompt_message in llm_result_chunk.delta.message.tool_calls: 431 | args = {} 432 | if prompt_message.function.arguments != "": 433 | args = json.loads(prompt_message.function.arguments) 434 | 435 | tool_calls.append( 436 | ( 437 | prompt_message.id, 438 | prompt_message.function.name, 439 | args, 440 | ) 441 | ) 442 | 443 | return tool_calls 444 | 445 | def extract_blocking_tool_calls( 446 | self, llm_result: LLMResult 447 | ) -> list[tuple[str, str, dict[str, Any]]]: 448 | """ 449 | Extract blocking tool calls from llm result 450 | 451 | Returns: 452 | List[Tuple[str, str, Dict[str, Any]]]: [(tool_call_id, tool_call_name, tool_call_args)] 453 | """ 454 | tool_calls = [] 455 | for prompt_message in llm_result.message.tool_calls: 456 | args = {} 457 | if prompt_message.function.arguments != "": 458 | args = json.loads(prompt_message.function.arguments) 459 | 460 | tool_calls.append( 461 | ( 462 | prompt_message.id, 463 | prompt_message.function.name, 464 | args, 465 | ) 466 | ) 467 | 468 | return tool_calls 469 | 470 | def _init_system_message( 471 | self, prompt_template: str, prompt_messages: list[PromptMessage] 472 | ) -> list[PromptMessage]: 473 | """ 474 | Initialize system message 475 | """ 476 | if not prompt_messages and prompt_template: 477 | return [ 478 | SystemPromptMessage(content=prompt_template), 479 | ] 480 | 481 | if ( 482 | prompt_messages 483 | and not isinstance(prompt_messages[0], SystemPromptMessage) 484 | and prompt_template 485 | ): 486 | prompt_messages.insert(0, SystemPromptMessage(content=prompt_template)) 487 | 488 | return prompt_messages or [] 489 | 490 | def _organize_user_query( 491 | self, query: str, prompt_messages: list[PromptMessage] 492 | ) -> list[PromptMessage]: 493 | """ 494 | Organize user query 495 | """ 496 | 497 | prompt_messages.append(UserPromptMessage(content=query)) 498 | 499 | return prompt_messages 500 | 501 | def _clear_user_prompt_image_messages( 502 | self, prompt_messages: list[PromptMessage] 503 | ) -> list[PromptMessage]: 504 | """ 505 | As for now, gpt supports both fc and vision at the first iteration. 506 | We need to remove the image messages from the prompt messages at the first iteration. 507 | """ 508 | prompt_messages = deepcopy(prompt_messages) 509 | 510 | for prompt_message in prompt_messages: 511 | if isinstance(prompt_message, UserPromptMessage) and isinstance( 512 | prompt_message.content, list 513 | ): 514 | prompt_message.content = "\n".join( 515 | [ 516 | content.data 517 | if content.type == PromptMessageContentType.TEXT 518 | else "[image]" 519 | if content.type == PromptMessageContentType.IMAGE 520 | else "[file]" 521 | for content in prompt_message.content 522 | ] 523 | ) 524 | 525 | return prompt_messages 526 | 527 | def _organize_prompt_messages( 528 | self, 529 | current_thoughts: list[PromptMessage], 530 | history_prompt_messages: list[PromptMessage], 531 | ) -> list[PromptMessage]: 532 | prompt_template = "" 533 | history_prompt_messages = self._init_system_message( 534 | prompt_template, history_prompt_messages 535 | ) 536 | query_prompt_messages = self._organize_user_query(self.query or "", []) 537 | 538 | prompt_messages = [ 539 | *history_prompt_messages, 540 | *query_prompt_messages, 541 | *current_thoughts, 542 | ] 543 | if len(current_thoughts) != 0: 544 | # clear messages after the first iteration 545 | prompt_messages = self._clear_user_prompt_image_messages(prompt_messages) 546 | return prompt_messages 547 | -------------------------------------------------------------------------------- /strategies/function_calling.yaml: -------------------------------------------------------------------------------- 1 | identity: 2 | name: function_calling 3 | author: Dify 4 | label: 5 | en_US: FunctionCalling 6 | zh_Hans: FunctionCalling 7 | pt_BR: FunctionCalling 8 | description: 9 | en_US: Function Calling is a basic strategy for agent, model will use the tools provided to perform the task. 10 | zh_Hans: Function Calling 是一个基本的 Agent 策略，模型将使用提供的工具来执行任务。 11 | pt_BR: Function Calling is a basic strategy for agent, model will use the tools provided to perform the task. 12 | parameters: 13 | - name: model 14 | type: model-selector 15 | scope: tool-call&llm 16 | required: true 17 | label: 18 | en_US: Model 19 | zh_Hans: 模型 20 | pt_BR: Model 21 | - name: tools 22 | type: array[tools] 23 | required: true 24 | label: 25 | en_US: Tools list 26 | zh_Hans: 工具列表 27 | pt_BR: Tools list 28 | - name: instruction 29 | type: string 30 | required: true 31 | label: 32 | en_US: Instruction 33 | zh_Hans: 指令 34 | pt_BR: Instruction 35 | auto_generate: 36 | type: prompt_instruction 37 | template: 38 | enabled: true 39 | - name: query 40 | type: string 41 | required: true 42 | label: 43 | en_US: Query 44 | zh_Hans: 查询 45 | pt_BR: Query 46 | - name: maximum_iterations 47 | type: number 48 | required: true 49 | label: 50 | en_US: Maxium Iterations 51 | zh_Hans: 最大迭代次数 52 | pt_BR: Maxium Iterations 53 | default: 3 54 | max: 30 55 | min: 1 56 | extra: 57 | python: 58 | source: strategies/function_calling.py 59 | -------------------------------------------------------------------------------- /strategies/mcpReAct.yaml: -------------------------------------------------------------------------------- 1 | identity: 2 | name: mcpReAct 3 | author: 3dify 4 | label: 5 | en_US: mcpReAct 6 | zh_Hans: mcpReAct 7 | pt_BR: mcpReAct 8 | description: 9 | en_US: mcpReAct is a basic strategy for agent, model will use the tools provided to perform the task. This node is MCP client that can connect to a MCP server. Dify is regarded as Host (not MCP server). 10 | zh_Hans: mcpReAct 是一个基本的 Agent 策略，模型将使用提供的工具来执行任务。 11 | pt_BR: mcpReAct is a basic strategy for agent, model will use the tools provided to perform the task. This node is MCP client that can connect to a MCP server. Dify is regarded as Host (not MCP server). 12 | parameters: 13 | - name: model 14 | type: model-selector 15 | scope: tool-call&llm 16 | required: true 17 | label: 18 | en_US: Model 19 | zh_Hans: 模型 20 | pt_BR: Model 21 | - name: config_json 22 | type: string 23 | required: true 24 | label: 25 | en_US: Commands to awake each MCP servers (claude_desktop_config.json) 26 | zh_Hans: 唤醒每个 MCP 服务器的命令（claude_desktop_config.json） 27 | pt_BR: Commands to awake each MCP servers (claude_desktop_config.json) 28 | - name: tools 29 | type: array[tools] 30 | required: false 31 | label: 32 | en_US: Tools list 33 | zh_Hans: 工具列表 34 | pt_BR: Tools list 35 | - name: instruction 36 | type: string 37 | required: true 38 | label: 39 | en_US: Instruction 40 | zh_Hans: 指令 41 | pt_BR: Instruction 42 | auto_generate: 43 | type: prompt_instruction 44 | template: 45 | enabled: true 46 | - name: query 47 | type: string 48 | required: true 49 | label: 50 | en_US: Query 51 | zh_Hans: 查询 52 | pt_BR: Query 53 | - name: maximum_iterations 54 | type: number 55 | required: true 56 | label: 57 | en_US: Maxium Iterations 58 | zh_Hans: 最大迭代次数 59 | pt_BR: Maxium Iterations 60 | default: 3 61 | min: 1 62 | max: 30 63 | extra: 64 | python: 65 | source: strategies/mcpReAct.py 66 | -------------------------------------------------------------------------------- /test/chatflow/README.md: -------------------------------------------------------------------------------- 1 | 2 | -------------------------------------------------------------------------------- /test/chatflow/mcp_multi_sse_chatflow_v0.0.3.yml: -------------------------------------------------------------------------------- 1 | app: 2 | description: 'Test for MCP client as Dify Agent Strategy plugin. 3 | 4 | used "Everything MCP server" 5 | 6 | Resource and Prompt are working correctly. 7 | 8 | Tool also Succussed, except for "samplingLLM", "annotatedMessage"' 9 | icon: 🤖 10 | icon_background: '#FFEAD5' 11 | mode: advanced-chat 12 | name: multi MCP server test 13 | use_icon_as_answer_icon: true 14 | dependencies: [] 15 | kind: app 16 | version: 0.1.5 17 | workflow: 18 | conversation_variables: [] 19 | environment_variables: [] 20 | features: 21 | file_upload: 22 | allowed_file_extensions: [] 23 | allowed_file_types: 24 | - image 25 | allowed_file_upload_methods: 26 | - remote_url 27 | - local_file 28 | enabled: true 29 | fileUploadConfig: 30 | audio_file_size_limit: 50 31 | batch_count_limit: 5 32 | file_size_limit: 15 33 | image_file_size_limit: 10 34 | video_file_size_limit: 100 35 | workflow_file_upload_limit: 10 36 | image: 37 | enabled: false 38 | number_limits: 3 39 | transfer_methods: 40 | - local_file 41 | - remote_url 42 | number_limits: 1 43 | opening_statement: '' 44 | retriever_resource: 45 | enabled: true 46 | sensitive_word_avoidance: 47 | enabled: false 48 | speech_to_text: 49 | enabled: false 50 | suggested_questions: [] 51 | suggested_questions_after_answer: 52 | enabled: false 53 | text_to_speech: 54 | enabled: false 55 | language: '' 56 | voice: '' 57 | graph: 58 | edges: 59 | - data: 60 | isInIteration: false 61 | sourceType: start 62 | targetType: agent 63 | id: 1741367649354-source-1742356283732-target 64 | source: '1741367649354' 65 | sourceHandle: source 66 | target: '1742356283732' 67 | targetHandle: target 68 | type: custom 69 | zIndex: 0 70 | - data: 71 | isInIteration: false 72 | sourceType: agent 73 | targetType: answer 74 | id: 1742356283732-source-answer-target 75 | source: '1742356283732' 76 | sourceHandle: source 77 | target: answer 78 | targetHandle: target 79 | type: custom 80 | zIndex: 0 81 | nodes: 82 | - data: 83 | desc: '' 84 | selected: false 85 | title: Start 86 | type: start 87 | variables: [] 88 | height: 53 89 | id: '1741367649354' 90 | position: 91 | x: 87.25729350925093 92 | y: 282 93 | positionAbsolute: 94 | x: 87.25729350925093 95 | y: 282 96 | selected: false 97 | sourcePosition: right 98 | targetPosition: left 99 | type: custom 100 | width: 243 101 | - data: 102 | answer: '{{#1742356283732.text#}}' 103 | desc: '' 104 | selected: false 105 | title: Answer 106 | type: answer 107 | variables: [] 108 | height: 101 109 | id: answer 110 | position: 111 | x: 629.2361265621643 112 | y: 282 113 | positionAbsolute: 114 | x: 629.2361265621643 115 | y: 282 116 | selected: false 117 | sourcePosition: right 118 | targetPosition: left 119 | type: custom 120 | width: 243 121 | - data: 122 | author: 3Dify-developer 123 | desc: '' 124 | height: 194 125 | selected: false 126 | showAuthor: true 127 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 128 | 16px;","text":"User prompt base","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 129 | 16px;","text":"I''m developing MCP client as Dify Agent Strategy Plugin.","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 130 | 16px;","text":"- Some features haven''t implemented and might not work as 131 | I expected.","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 132 | 16px;","text":"- I found good MCP server for testing.","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 133 | 16px;","text":"- You should ignore \"Current Time Tool\". ","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 134 | theme: blue 135 | title: '' 136 | type: '' 137 | width: 569 138 | height: 194 139 | id: '1741624678131' 140 | position: 141 | x: 87.25729350925093 142 | y: 508.8200615915433 143 | positionAbsolute: 144 | x: 87.25729350925093 145 | y: 508.8200615915433 146 | selected: false 147 | sourcePosition: right 148 | targetPosition: left 149 | type: custom-note 150 | width: 569 151 | - data: 152 | author: 3Dify-developer 153 | desc: '' 154 | height: 88 155 | selected: false 156 | showAuthor: false 157 | text: '{"root":{"children":[{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 158 | 16px;","text":"Sample MCP server for testing","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"link","version":1,"rel":"noreferrer","target":null,"title":null,"url":""}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":"font-size: 159 | 16px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":"https://github.com/modelcontextprotocol/servers/tree/main/src/everything","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 160 | theme: cyan 161 | title: '' 162 | type: '' 163 | width: 449 164 | height: 88 165 | id: '1741626728667' 166 | position: 167 | x: 690.5542795067578 168 | y: 763.3257650174893 169 | positionAbsolute: 170 | x: 690.5542795067578 171 | y: 763.3257650174893 172 | selected: false 173 | sourcePosition: right 174 | targetPosition: left 175 | type: custom-note 176 | width: 449 177 | - data: 178 | author: 3Dify-developer 179 | desc: '' 180 | height: 151 181 | selected: false 182 | showAuthor: true 183 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 184 | 16px;","text":"MCP Tool Test prompt","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 185 | 16px;","text":"Try mcp tools one-by-one","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 186 | 16px;","text":"- judge result in \"thought\" phase (I want to know at least 187 | Fail or Success)","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 188 | 16px;","text":"- avoid \"sampleLLM\" which is not compatible feature yet. 189 | ","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 190 | theme: green 191 | title: '' 192 | type: '' 193 | width: 570 194 | height: 151 195 | id: '1741669149712' 196 | position: 197 | x: 87.25729350925093 198 | y: 718.2693074279789 199 | positionAbsolute: 200 | x: 87.25729350925093 201 | y: 718.2693074279789 202 | selected: false 203 | sourcePosition: right 204 | targetPosition: left 205 | type: custom-note 206 | width: 570 207 | - data: 208 | author: 3Dify-developer 209 | desc: '' 210 | height: 140 211 | selected: false 212 | showAuthor: true 213 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 214 | 16px;","text":"MCP Resource Test prompt","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 215 | 16px;","text":"Try mcp resources one-by-one","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":"font-size: 216 | 16px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 217 | 16px;","text":"- judge result in \"thought\" phase (I want to know at least 218 | Fail or Success)","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 219 | theme: green 220 | title: '' 221 | type: '' 222 | width: 568 223 | height: 140 224 | id: '1741669231167' 225 | position: 226 | x: 88.65225939320432 227 | y: 879.6113451588606 228 | positionAbsolute: 229 | x: 88.65225939320432 230 | y: 879.6113451588606 231 | selected: false 232 | sourcePosition: right 233 | targetPosition: left 234 | type: custom-note 235 | width: 568 236 | - data: 237 | author: 3Dify-developer 238 | desc: '' 239 | height: 128 240 | selected: false 241 | showAuthor: true 242 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 243 | 16px;","text":"MCP Prompt Test prompt","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 244 | 16px;","text":"Try mcp prompts one-by-one","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 245 | 16px;","text":"- judge result in \"thought\" phase (I want to know at least 246 | Fail or Success)","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 247 | theme: green 248 | title: '' 249 | type: '' 250 | width: 568 251 | height: 128 252 | id: '1741669352719' 253 | position: 254 | x: 87.25729350925093 255 | y: 1033.9889783833733 256 | positionAbsolute: 257 | x: 87.25729350925093 258 | y: 1033.9889783833733 259 | selected: false 260 | sourcePosition: right 261 | targetPosition: left 262 | type: custom-note 263 | width: 568 264 | - data: 265 | author: 3Dify-developer 266 | desc: '' 267 | height: 264 268 | selected: false 269 | showAuthor: true 270 | text: '{"root":{"children":[{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 271 | 16px;","text":"config.json example ","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 272 | 16px;","text":"(","type":"text","version":1},{"detail":0,"format":1,"mode":"normal","style":"font-size: 273 | 16px;","text":"Cloud?","type":"text","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 274 | 16px;","text":")","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"","text":"{","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"mcpServers\": 275 | {","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"everything\": 276 | {","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"url\": 277 | \"http://127.0.0.1:8080/sse\"","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" },","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"simple-arxiv\": 278 | {","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"url\": 279 | \"http://127.0.0.1:8008/sse\"","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" }","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" }","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":"}","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 280 | theme: pink 281 | title: '' 282 | type: '' 283 | width: 283 284 | height: 264 285 | id: '1741671860068' 286 | position: 287 | x: 1027.6182377664975 288 | y: 988.1065026922056 289 | positionAbsolute: 290 | x: 1027.6182377664975 291 | y: 988.1065026922056 292 | selected: false 293 | sourcePosition: right 294 | targetPosition: left 295 | type: custom-note 296 | width: 283 297 | - data: 298 | author: 3Dify-developer 299 | desc: '' 300 | height: 93 301 | selected: false 302 | showAuthor: true 303 | text: '{"root":{"children":[{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 304 | 14px;","text":"TOOL LIST should not blank. Choose some plugin.","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":"font-size: 305 | 14px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 306 | 14px;","text":"I recommend official Dify plugin like \"Current Time\".","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":"font-size: 307 | 14px;"}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 308 | theme: yellow 309 | title: '' 310 | type: '' 311 | width: 363 312 | height: 93 313 | id: '1741766414431' 314 | position: 315 | x: 289.33163665746144 316 | y: 178.12817056680163 317 | positionAbsolute: 318 | x: 289.33163665746144 319 | y: 178.12817056680163 320 | selected: false 321 | sourcePosition: right 322 | targetPosition: left 323 | type: custom-note 324 | width: 363 325 | - data: 326 | author: 3Dify-developer 327 | desc: '' 328 | height: 362 329 | selected: false 330 | showAuthor: true 331 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 332 | 16px;","text":"Example query","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 333 | 16px;","text":"I''m developing MCP client as Dify Agent Strategy Plugin.","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 334 | 16px;","text":"- Some features haven''t implemented and might not work as 335 | I expected.","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 336 | 16px;","text":"- I found good MCP server for testing.","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 337 | 16px;","text":"- You should ignore \"Current Time Tool\". ","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 338 | 16px;","text":"At first, tell me all tools you are available in ","type":"text","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 339 | 14px;","text":"\"thought\".","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 340 | 16px;","text":"Try mcp tools one-by-one","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 341 | 16px;","text":"- judge result in \"thought\" phase (I want to know at least 342 | Fail or Success)","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 343 | 16px;","text":"- avoid \"sampleLLM\" which is not compatible feature yet. 344 | ","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 345 | theme: blue 346 | title: '' 347 | type: '' 348 | width: 502 349 | height: 362 350 | id: '1741772186537' 351 | position: 352 | x: 673.1166803440168 353 | y: 375.05515998236154 354 | positionAbsolute: 355 | x: 673.1166803440168 356 | y: 375.05515998236154 357 | selected: false 358 | sourcePosition: right 359 | targetPosition: left 360 | type: custom-note 361 | width: 502 362 | - data: 363 | author: 3Dify-developer 364 | desc: '' 365 | height: 97 366 | selected: false 367 | showAuthor: true 368 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 369 | 16px;","text":"Instruction","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"","text":"You 370 | are helpful LLM Agent","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 371 | theme: violet 372 | title: '' 373 | type: '' 374 | width: 240 375 | height: 97 376 | id: '1741802957289' 377 | position: 378 | x: 87.25729350925093 379 | y: 375.05515998236154 380 | positionAbsolute: 381 | x: 87.25729350925093 382 | y: 375.05515998236154 383 | selected: false 384 | sourcePosition: right 385 | targetPosition: left 386 | type: custom-note 387 | width: 240 388 | - data: 389 | agent_parameters: 390 | config_json: 391 | type: constant 392 | value: "{\n \"mcpServers\": {\n \"everything\": {\n \"url\"\ 393 | : \"http://host.docker.internal:8080/sse\"\n },\n \"simple-arxiv\"\ 394 | : {\n \"url\": \"http://host.docker.internal:8008/sse\"\n \ 395 | \ }\n }\n}" 396 | instruction: 397 | type: constant 398 | value: You are helpful LLM Agent 399 | maximum_iterations: 400 | type: constant 401 | value: 3 402 | model: 403 | type: constant 404 | value: 405 | completion_params: {} 406 | mode: chat 407 | model: gemini-2.0-flash-exp 408 | model_type: llm 409 | provider: langgenius/gemini/google 410 | type: model-selector 411 | query: 412 | type: constant 413 | value: '{{#sys.query#}}' 414 | tools: 415 | type: constant 416 | value: 417 | - enabled: true 418 | extra: 419 | description: '' 420 | parameters: {} 421 | provider_name: time 422 | schemas: 423 | - auto_generate: null 424 | default: '%Y-%m-%d %H:%M:%S' 425 | form: form 426 | human_description: 427 | en_US: Time format in strftime standard. 428 | ja_JP: Time format in strftime standard. 429 | pt_BR: Time format in strftime standard. 430 | zh_Hans: strftime 标准的时间格式。 431 | label: 432 | en_US: Format 433 | ja_JP: Format 434 | pt_BR: Format 435 | zh_Hans: 格式 436 | llm_description: null 437 | max: null 438 | min: null 439 | name: format 440 | options: [] 441 | placeholder: null 442 | precision: null 443 | required: false 444 | scope: null 445 | template: null 446 | type: string 447 | - auto_generate: null 448 | default: UTC 449 | form: form 450 | human_description: 451 | en_US: Timezone 452 | ja_JP: Timezone 453 | pt_BR: Timezone 454 | zh_Hans: 时区 455 | label: 456 | en_US: Timezone 457 | ja_JP: Timezone 458 | pt_BR: Timezone 459 | zh_Hans: 时区 460 | llm_description: null 461 | max: null 462 | min: null 463 | name: timezone 464 | options: 465 | - label: 466 | en_US: UTC 467 | ja_JP: UTC 468 | pt_BR: UTC 469 | zh_Hans: UTC 470 | value: UTC 471 | - label: 472 | en_US: America/New_York 473 | ja_JP: America/New_York 474 | pt_BR: America/New_York 475 | zh_Hans: 美洲/纽约 476 | value: America/New_York 477 | - label: 478 | en_US: America/Los_Angeles 479 | ja_JP: America/Los_Angeles 480 | pt_BR: America/Los_Angeles 481 | zh_Hans: 美洲/洛杉矶 482 | value: America/Los_Angeles 483 | - label: 484 | en_US: America/Chicago 485 | ja_JP: America/Chicago 486 | pt_BR: America/Chicago 487 | zh_Hans: 美洲/芝加哥 488 | value: America/Chicago 489 | - label: 490 | en_US: America/Sao_Paulo 491 | ja_JP: America/Sao_Paulo 492 | pt_BR: América/São Paulo 493 | zh_Hans: 美洲/圣保罗 494 | value: America/Sao_Paulo 495 | - label: 496 | en_US: Asia/Shanghai 497 | ja_JP: Asia/Shanghai 498 | pt_BR: Asia/Shanghai 499 | zh_Hans: 亚洲/上海 500 | value: Asia/Shanghai 501 | - label: 502 | en_US: Asia/Ho_Chi_Minh 503 | ja_JP: Asia/Ho_Chi_Minh 504 | pt_BR: Ásia/Ho Chi Minh 505 | zh_Hans: 亚洲/胡志明市 506 | value: Asia/Ho_Chi_Minh 507 | - label: 508 | en_US: Asia/Tokyo 509 | ja_JP: Asia/Tokyo 510 | pt_BR: Asia/Tokyo 511 | zh_Hans: 亚洲/东京 512 | value: Asia/Tokyo 513 | - label: 514 | en_US: Asia/Dubai 515 | ja_JP: Asia/Dubai 516 | pt_BR: Asia/Dubai 517 | zh_Hans: 亚洲/迪拜 518 | value: Asia/Dubai 519 | - label: 520 | en_US: Asia/Kolkata 521 | ja_JP: Asia/Kolkata 522 | pt_BR: Asia/Kolkata 523 | zh_Hans: 亚洲/加尔各答 524 | value: Asia/Kolkata 525 | - label: 526 | en_US: Asia/Seoul 527 | ja_JP: Asia/Seoul 528 | pt_BR: Asia/Seoul 529 | zh_Hans: 亚洲/首尔 530 | value: Asia/Seoul 531 | - label: 532 | en_US: Asia/Singapore 533 | ja_JP: Asia/Singapore 534 | pt_BR: Asia/Singapore 535 | zh_Hans: 亚洲/新加坡 536 | value: Asia/Singapore 537 | - label: 538 | en_US: Europe/London 539 | ja_JP: Europe/London 540 | pt_BR: Europe/London 541 | zh_Hans: 欧洲/伦敦 542 | value: Europe/London 543 | - label: 544 | en_US: Europe/Berlin 545 | ja_JP: Europe/Berlin 546 | pt_BR: Europe/Berlin 547 | zh_Hans: 欧洲/柏林 548 | value: Europe/Berlin 549 | - label: 550 | en_US: Europe/Moscow 551 | ja_JP: Europe/Moscow 552 | pt_BR: Europe/Moscow 553 | zh_Hans: 欧洲/莫斯科 554 | value: Europe/Moscow 555 | - label: 556 | en_US: Australia/Sydney 557 | ja_JP: Australia/Sydney 558 | pt_BR: Australia/Sydney 559 | zh_Hans: 澳大利亚/悉尼 560 | value: Australia/Sydney 561 | - label: 562 | en_US: Pacific/Auckland 563 | ja_JP: Pacific/Auckland 564 | pt_BR: Pacific/Auckland 565 | zh_Hans: 太平洋/奥克兰 566 | value: Pacific/Auckland 567 | - label: 568 | en_US: Africa/Cairo 569 | ja_JP: Africa/Cairo 570 | pt_BR: Africa/Cairo 571 | zh_Hans: 非洲/开罗 572 | value: Africa/Cairo 573 | placeholder: null 574 | precision: null 575 | required: false 576 | scope: null 577 | template: null 578 | type: select 579 | settings: 580 | format: 581 | value: '%Y-%m-%d %H:%M:%S' 582 | timezone: 583 | value: UTC 584 | tool_label: Current Time 585 | tool_name: current_time 586 | type: builtin 587 | agent_strategy_label: mcpReAct 588 | agent_strategy_name: mcpReAct 589 | agent_strategy_provider_name: 3dify-project/mcp_client/agent 590 | desc: '' 591 | output_schema: null 592 | plugin_unique_identifier: 3dify-project/mcp_client:0.0.2@0a8f8d04a49549adc0fbe6118da99f846239aad74b5ede932e8bb40cec531491 593 | selected: false 594 | title: Agent 595 | type: agent 596 | height: 197 597 | id: '1742356283732' 598 | position: 599 | x: 361.24246556756697 600 | y: 282 601 | positionAbsolute: 602 | x: 361.24246556756697 603 | y: 282 604 | selected: true 605 | sourcePosition: right 606 | targetPosition: left 607 | type: custom 608 | width: 243 609 | - data: 610 | author: 3Dify-developer 611 | desc: '' 612 | height: 264 613 | selected: false 614 | showAuthor: false 615 | text: '{"root":{"children":[{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 616 | 16px;","text":"config.json example ","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 617 | 16px;","text":"(","type":"text","version":1},{"detail":0,"format":1,"mode":"normal","style":"font-size: 618 | 16px;","text":"Docker Compose self deploy","type":"text","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 619 | 16px;","text":")","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":"{","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"mcpServers\": 620 | {","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"everything\": 621 | {","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"url\": 622 | \"http://host.docker.internal:8080/sse\"","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" },","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"simple-arxiv\": 623 | {","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"url\": 624 | \"http://host.docker.internal:8008/sse\"","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" }","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" }","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":"}","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 625 | theme: pink 626 | title: '' 627 | type: '' 628 | width: 315 629 | height: 264 630 | id: '1742367595361' 631 | position: 632 | x: 690.5542795067578 633 | y: 988.1065026922056 634 | positionAbsolute: 635 | x: 690.5542795067578 636 | y: 988.1065026922056 637 | selected: false 638 | sourcePosition: right 639 | targetPosition: left 640 | type: custom-note 641 | width: 315 642 | - data: 643 | author: 3Dify-developer 644 | desc: '' 645 | height: 97 646 | selected: false 647 | showAuthor: false 648 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 649 | 16px;","text":"How to convert \"stdio\" to \"SSE\" MCP server","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":"README 650 | of my GitHub repository","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":"https://github.com/3dify-project/dify-mcp-client","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"link","version":1,"rel":"noreferrer","target":null,"title":null,"url":"https://github.com/3dify-project/dify-mcp-client"}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 651 | theme: cyan 652 | title: '' 653 | type: '' 654 | width: 447 655 | height: 97 656 | id: '1742452980902' 657 | position: 658 | x: 690.5542795067578 659 | y: 863.8984826162116 660 | positionAbsolute: 661 | x: 690.5542795067578 662 | y: 863.8984826162116 663 | selected: false 664 | sourcePosition: right 665 | targetPosition: left 666 | type: custom-note 667 | width: 447 668 | viewport: 669 | x: 73.56256265549337 670 | y: -104.78783867265224 671 | zoom: 0.8487039665217165 672 | -------------------------------------------------------------------------------- /test/chatflow/mcp_sse_chatflow_v0.0.2.yml: -------------------------------------------------------------------------------- 1 | app: 2 | description: 'Test for MCP client as Dify Agent Strategy plugin. 3 | 4 | used "Everything MCP server" 5 | 6 | Resource and Prompt are working correctly. 7 | 8 | Tool also Succussed, except for "samplingLLM", "annotatedMessage"' 9 | icon: 🤖 10 | icon_background: '#FFEAD5' 11 | mode: advanced-chat 12 | name: SSE MCP test 13 | use_icon_as_answer_icon: true 14 | dependencies: [] 15 | kind: app 16 | version: 0.1.5 17 | workflow: 18 | conversation_variables: [] 19 | environment_variables: [] 20 | features: 21 | file_upload: 22 | allowed_file_extensions: 23 | - .JPG 24 | - .JPEG 25 | - .PNG 26 | - .GIF 27 | - .WEBP 28 | - .SVG 29 | allowed_file_types: 30 | - image 31 | allowed_file_upload_methods: 32 | - local_file 33 | - remote_url 34 | enabled: false 35 | fileUploadConfig: 36 | audio_file_size_limit: 50 37 | batch_count_limit: 5 38 | file_size_limit: 15 39 | image_file_size_limit: 10 40 | video_file_size_limit: 100 41 | workflow_file_upload_limit: 10 42 | image: 43 | enabled: false 44 | number_limits: 3 45 | transfer_methods: 46 | - local_file 47 | - remote_url 48 | number_limits: 3 49 | opening_statement: '' 50 | retriever_resource: 51 | enabled: true 52 | sensitive_word_avoidance: 53 | enabled: false 54 | speech_to_text: 55 | enabled: false 56 | suggested_questions: [] 57 | suggested_questions_after_answer: 58 | enabled: false 59 | text_to_speech: 60 | enabled: false 61 | language: '' 62 | voice: '' 63 | graph: 64 | edges: 65 | - data: 66 | isInIteration: false 67 | sourceType: start 68 | targetType: agent 69 | id: 1741367649354-source-1742356283732-target 70 | source: '1741367649354' 71 | sourceHandle: source 72 | target: '1742356283732' 73 | targetHandle: target 74 | type: custom 75 | zIndex: 0 76 | - data: 77 | isInIteration: false 78 | sourceType: agent 79 | targetType: answer 80 | id: 1742356283732-source-answer-target 81 | source: '1742356283732' 82 | sourceHandle: source 83 | target: answer 84 | targetHandle: target 85 | type: custom 86 | zIndex: 0 87 | nodes: 88 | - data: 89 | desc: '' 90 | selected: false 91 | title: Start 92 | type: start 93 | variables: [] 94 | height: 53 95 | id: '1741367649354' 96 | position: 97 | x: 87.25729350925093 98 | y: 282 99 | positionAbsolute: 100 | x: 87.25729350925093 101 | y: 282 102 | selected: false 103 | sourcePosition: right 104 | targetPosition: left 105 | type: custom 106 | width: 243 107 | - data: 108 | answer: '{{#1742356283732.text#}}' 109 | desc: '' 110 | selected: false 111 | title: Answer 112 | type: answer 113 | variables: [] 114 | height: 101 115 | id: answer 116 | position: 117 | x: 629.2361265621643 118 | y: 282 119 | positionAbsolute: 120 | x: 629.2361265621643 121 | y: 282 122 | selected: false 123 | sourcePosition: right 124 | targetPosition: left 125 | type: custom 126 | width: 243 127 | - data: 128 | author: 3Dify-developer 129 | desc: '' 130 | height: 194 131 | selected: false 132 | showAuthor: true 133 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 134 | 16px;","text":"User prompt base","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 135 | 16px;","text":"I''m developing MCP client as Dify Agent Strategy Plugin.","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 136 | 16px;","text":"- Some features haven''t implemented and might not work as 137 | I expected.","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 138 | 16px;","text":"- I found good MCP server for testing.","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 139 | 16px;","text":"- You should ignore \"Current Time Tool\". ","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 140 | theme: blue 141 | title: '' 142 | type: '' 143 | width: 569 144 | height: 194 145 | id: '1741624678131' 146 | position: 147 | x: 87.25729350925093 148 | y: 508.8200615915433 149 | positionAbsolute: 150 | x: 87.25729350925093 151 | y: 508.8200615915433 152 | selected: false 153 | sourcePosition: right 154 | targetPosition: left 155 | type: custom-note 156 | width: 569 157 | - data: 158 | author: 3Dify-developer 159 | desc: '' 160 | height: 88 161 | selected: false 162 | showAuthor: false 163 | text: '{"root":{"children":[{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 164 | 16px;","text":"Sample MCP server for testing","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"link","version":1,"rel":"noreferrer","target":null,"title":null,"url":""}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":"font-size: 165 | 16px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":"https://github.com/modelcontextprotocol/servers/tree/main/src/everything","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 166 | theme: cyan 167 | title: '' 168 | type: '' 169 | width: 449 170 | height: 88 171 | id: '1741626728667' 172 | position: 173 | x: 690.5542795067578 174 | y: 763.3257650174893 175 | positionAbsolute: 176 | x: 690.5542795067578 177 | y: 763.3257650174893 178 | selected: false 179 | sourcePosition: right 180 | targetPosition: left 181 | type: custom-note 182 | width: 449 183 | - data: 184 | author: 3Dify-developer 185 | desc: '' 186 | height: 151 187 | selected: false 188 | showAuthor: true 189 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 190 | 16px;","text":"MCP Tool Test prompt","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 191 | 16px;","text":"Try mcp tools one-by-one","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 192 | 16px;","text":"- judge result in \"thought\" phase (I want to know at least 193 | Fail or Success)","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 194 | 16px;","text":"- avoid \"sampleLLM\" which is not compatible feature yet. 195 | ","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 196 | theme: green 197 | title: '' 198 | type: '' 199 | width: 570 200 | height: 151 201 | id: '1741669149712' 202 | position: 203 | x: 87.25729350925093 204 | y: 718.2693074279789 205 | positionAbsolute: 206 | x: 87.25729350925093 207 | y: 718.2693074279789 208 | selected: false 209 | sourcePosition: right 210 | targetPosition: left 211 | type: custom-note 212 | width: 570 213 | - data: 214 | author: 3Dify-developer 215 | desc: '' 216 | height: 140 217 | selected: false 218 | showAuthor: true 219 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 220 | 16px;","text":"MCP Resource Test prompt","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 221 | 16px;","text":"Try mcp resources one-by-one","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":"font-size: 222 | 16px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 223 | 16px;","text":"- judge result in \"thought\" phase (I want to know at least 224 | Fail or Success)","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 225 | theme: green 226 | title: '' 227 | type: '' 228 | width: 568 229 | height: 140 230 | id: '1741669231167' 231 | position: 232 | x: 88.65225939320432 233 | y: 879.6113451588606 234 | positionAbsolute: 235 | x: 88.65225939320432 236 | y: 879.6113451588606 237 | selected: false 238 | sourcePosition: right 239 | targetPosition: left 240 | type: custom-note 241 | width: 568 242 | - data: 243 | author: 3Dify-developer 244 | desc: '' 245 | height: 128 246 | selected: false 247 | showAuthor: true 248 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 249 | 16px;","text":"MCP Prompt Test prompt","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 250 | 16px;","text":"Try mcp prompts one-by-one","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 251 | 16px;","text":"- judge result in \"thought\" phase (I want to know at least 252 | Fail or Success)","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 253 | theme: green 254 | title: '' 255 | type: '' 256 | width: 568 257 | height: 128 258 | id: '1741669352719' 259 | position: 260 | x: 87.25729350925093 261 | y: 1033.9889783833733 262 | positionAbsolute: 263 | x: 87.25729350925093 264 | y: 1033.9889783833733 265 | selected: false 266 | sourcePosition: right 267 | targetPosition: left 268 | type: custom-note 269 | width: 568 270 | - data: 271 | author: 3Dify-developer 272 | desc: '' 273 | height: 228 274 | selected: false 275 | showAuthor: true 276 | text: '{"root":{"children":[{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 277 | 16px;","text":"config.json example ","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":"font-size: 278 | 16px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 279 | 16px;","text":"(","type":"text","version":1},{"detail":0,"format":1,"mode":"normal","style":"font-size: 280 | 16px;","text":"Cloud?","type":"text","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 281 | 16px;","text":")","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"","text":"{","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":"font-size: 282 | 16px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"mcpServers\": 283 | {","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"everything\": 284 | {","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"url\": 285 | \"http://127.0.0.1:8080/sse\"","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" }","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" }","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":"}","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 286 | theme: pink 287 | title: '' 288 | type: '' 289 | width: 243 290 | height: 228 291 | id: '1741671860068' 292 | position: 293 | x: 1060.1016025290248 294 | y: 969.1578732473977 295 | positionAbsolute: 296 | x: 1060.1016025290248 297 | y: 969.1578732473977 298 | selected: false 299 | sourcePosition: right 300 | targetPosition: left 301 | type: custom-note 302 | width: 243 303 | - data: 304 | author: 3Dify-developer 305 | desc: '' 306 | height: 93 307 | selected: false 308 | showAuthor: true 309 | text: '{"root":{"children":[{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 310 | 14px;","text":"TOOL LIST should not blank. Choose some plugin.","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":"font-size: 311 | 14px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 312 | 14px;","text":"I recommend official Dify plugin like \"Current Time\".","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":"font-size: 313 | 14px;"}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 314 | theme: yellow 315 | title: '' 316 | type: '' 317 | width: 363 318 | height: 93 319 | id: '1741766414431' 320 | position: 321 | x: 289.33163665746144 322 | y: 178.12817056680163 323 | positionAbsolute: 324 | x: 289.33163665746144 325 | y: 178.12817056680163 326 | selected: false 327 | sourcePosition: right 328 | targetPosition: left 329 | type: custom-note 330 | width: 363 331 | - data: 332 | author: 3Dify-developer 333 | desc: '' 334 | height: 362 335 | selected: false 336 | showAuthor: true 337 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 338 | 16px;","text":"Example query","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 339 | 16px;","text":"I''m developing MCP client as Dify Agent Strategy Plugin.","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":"font-size: 340 | 16px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 341 | 16px;","text":"- Some features haven''t implemented and might not work as 342 | I expected.","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 343 | 16px;","text":"- I found good MCP server for testing.","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 344 | 16px;","text":"- You should ignore \"Current Time Tool\". ","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 345 | 16px;","text":"At first tell me what tools are available","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 346 | 16px;","text":"Try mcp tools one-by-one","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 347 | 16px;","text":"- judge result in \"thought\" phase (I want to know at least 348 | Fail or Success)","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 349 | 16px;","text":"- avoid \"sampleLLM\" which is not compatible feature yet. 350 | ","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 351 | theme: blue 352 | title: '' 353 | type: '' 354 | width: 502 355 | height: 362 356 | id: '1741772186537' 357 | position: 358 | x: 690.5542795067578 359 | y: 397.44223574668086 360 | positionAbsolute: 361 | x: 690.5542795067578 362 | y: 397.44223574668086 363 | selected: false 364 | sourcePosition: right 365 | targetPosition: left 366 | type: custom-note 367 | width: 502 368 | - data: 369 | author: 3Dify-developer 370 | desc: '' 371 | height: 97 372 | selected: false 373 | showAuthor: true 374 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 375 | 16px;","text":"Instruction","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"","text":"You 376 | are helpful LLM Agent","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 377 | theme: violet 378 | title: '' 379 | type: '' 380 | width: 240 381 | height: 97 382 | id: '1741802957289' 383 | position: 384 | x: 87.25729350925093 385 | y: 375.05515998236154 386 | positionAbsolute: 387 | x: 87.25729350925093 388 | y: 375.05515998236154 389 | selected: true 390 | sourcePosition: right 391 | targetPosition: left 392 | type: custom-note 393 | width: 240 394 | - data: 395 | agent_parameters: 396 | config_json: 397 | type: constant 398 | value: "{\n \"mcpServers\": {\n \"everything\": {\n \"url\"\ 399 | : \"http://host.docker.internal:8080/sse\"\n }\n }\n}" 400 | instruction: 401 | type: constant 402 | value: You are helpful LLM Agent 403 | maximum_iterations: 404 | type: constant 405 | value: 3 406 | model: 407 | type: constant 408 | value: 409 | completion_params: {} 410 | mode: chat 411 | model: gemini-2.0-flash-exp 412 | model_type: llm 413 | provider: langgenius/gemini/google 414 | type: model-selector 415 | query: 416 | type: constant 417 | value: '{{#sys.query#}}' 418 | tools: 419 | type: constant 420 | value: 421 | - enabled: true 422 | extra: 423 | description: '' 424 | parameters: {} 425 | provider_name: time 426 | schemas: 427 | - auto_generate: null 428 | default: '%Y-%m-%d %H:%M:%S' 429 | form: form 430 | human_description: 431 | en_US: Time format in strftime standard. 432 | ja_JP: Time format in strftime standard. 433 | pt_BR: Time format in strftime standard. 434 | zh_Hans: strftime 标准的时间格式。 435 | label: 436 | en_US: Format 437 | ja_JP: Format 438 | pt_BR: Format 439 | zh_Hans: 格式 440 | llm_description: null 441 | max: null 442 | min: null 443 | name: format 444 | options: [] 445 | placeholder: null 446 | precision: null 447 | required: false 448 | scope: null 449 | template: null 450 | type: string 451 | - auto_generate: null 452 | default: UTC 453 | form: form 454 | human_description: 455 | en_US: Timezone 456 | ja_JP: Timezone 457 | pt_BR: Timezone 458 | zh_Hans: 时区 459 | label: 460 | en_US: Timezone 461 | ja_JP: Timezone 462 | pt_BR: Timezone 463 | zh_Hans: 时区 464 | llm_description: null 465 | max: null 466 | min: null 467 | name: timezone 468 | options: 469 | - label: 470 | en_US: UTC 471 | ja_JP: UTC 472 | pt_BR: UTC 473 | zh_Hans: UTC 474 | value: UTC 475 | - label: 476 | en_US: America/New_York 477 | ja_JP: America/New_York 478 | pt_BR: America/New_York 479 | zh_Hans: 美洲/纽约 480 | value: America/New_York 481 | - label: 482 | en_US: America/Los_Angeles 483 | ja_JP: America/Los_Angeles 484 | pt_BR: America/Los_Angeles 485 | zh_Hans: 美洲/洛杉矶 486 | value: America/Los_Angeles 487 | - label: 488 | en_US: America/Chicago 489 | ja_JP: America/Chicago 490 | pt_BR: America/Chicago 491 | zh_Hans: 美洲/芝加哥 492 | value: America/Chicago 493 | - label: 494 | en_US: America/Sao_Paulo 495 | ja_JP: America/Sao_Paulo 496 | pt_BR: América/São Paulo 497 | zh_Hans: 美洲/圣保罗 498 | value: America/Sao_Paulo 499 | - label: 500 | en_US: Asia/Shanghai 501 | ja_JP: Asia/Shanghai 502 | pt_BR: Asia/Shanghai 503 | zh_Hans: 亚洲/上海 504 | value: Asia/Shanghai 505 | - label: 506 | en_US: Asia/Ho_Chi_Minh 507 | ja_JP: Asia/Ho_Chi_Minh 508 | pt_BR: Ásia/Ho Chi Minh 509 | zh_Hans: 亚洲/胡志明市 510 | value: Asia/Ho_Chi_Minh 511 | - label: 512 | en_US: Asia/Tokyo 513 | ja_JP: Asia/Tokyo 514 | pt_BR: Asia/Tokyo 515 | zh_Hans: 亚洲/东京 516 | value: Asia/Tokyo 517 | - label: 518 | en_US: Asia/Dubai 519 | ja_JP: Asia/Dubai 520 | pt_BR: Asia/Dubai 521 | zh_Hans: 亚洲/迪拜 522 | value: Asia/Dubai 523 | - label: 524 | en_US: Asia/Kolkata 525 | ja_JP: Asia/Kolkata 526 | pt_BR: Asia/Kolkata 527 | zh_Hans: 亚洲/加尔各答 528 | value: Asia/Kolkata 529 | - label: 530 | en_US: Asia/Seoul 531 | ja_JP: Asia/Seoul 532 | pt_BR: Asia/Seoul 533 | zh_Hans: 亚洲/首尔 534 | value: Asia/Seoul 535 | - label: 536 | en_US: Asia/Singapore 537 | ja_JP: Asia/Singapore 538 | pt_BR: Asia/Singapore 539 | zh_Hans: 亚洲/新加坡 540 | value: Asia/Singapore 541 | - label: 542 | en_US: Europe/London 543 | ja_JP: Europe/London 544 | pt_BR: Europe/London 545 | zh_Hans: 欧洲/伦敦 546 | value: Europe/London 547 | - label: 548 | en_US: Europe/Berlin 549 | ja_JP: Europe/Berlin 550 | pt_BR: Europe/Berlin 551 | zh_Hans: 欧洲/柏林 552 | value: Europe/Berlin 553 | - label: 554 | en_US: Europe/Moscow 555 | ja_JP: Europe/Moscow 556 | pt_BR: Europe/Moscow 557 | zh_Hans: 欧洲/莫斯科 558 | value: Europe/Moscow 559 | - label: 560 | en_US: Australia/Sydney 561 | ja_JP: Australia/Sydney 562 | pt_BR: Australia/Sydney 563 | zh_Hans: 澳大利亚/悉尼 564 | value: Australia/Sydney 565 | - label: 566 | en_US: Pacific/Auckland 567 | ja_JP: Pacific/Auckland 568 | pt_BR: Pacific/Auckland 569 | zh_Hans: 太平洋/奥克兰 570 | value: Pacific/Auckland 571 | - label: 572 | en_US: Africa/Cairo 573 | ja_JP: Africa/Cairo 574 | pt_BR: Africa/Cairo 575 | zh_Hans: 非洲/开罗 576 | value: Africa/Cairo 577 | placeholder: null 578 | precision: null 579 | required: false 580 | scope: null 581 | template: null 582 | type: select 583 | settings: 584 | format: 585 | value: '%Y-%m-%d %H:%M:%S' 586 | timezone: 587 | value: UTC 588 | tool_label: Current Time 589 | tool_name: current_time 590 | type: builtin 591 | agent_strategy_label: mcpReAct 592 | agent_strategy_name: mcpReAct 593 | agent_strategy_provider_name: 3dify-project/mcp_client/agent 594 | desc: '' 595 | output_schema: null 596 | plugin_unique_identifier: 3dify-project/mcp_client:0.0.2@0a8f8d04a49549adc0fbe6118da99f846239aad74b5ede932e8bb40cec531491 597 | selected: false 598 | title: Agent 599 | type: agent 600 | height: 197 601 | id: '1742356283732' 602 | position: 603 | x: 361.24246556756697 604 | y: 282 605 | positionAbsolute: 606 | x: 361.24246556756697 607 | y: 282 608 | selected: false 609 | sourcePosition: right 610 | targetPosition: left 611 | type: custom 612 | width: 243 613 | - data: 614 | author: 3Dify-developer 615 | desc: '' 616 | height: 231 617 | selected: false 618 | showAuthor: true 619 | text: '{"root":{"children":[{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 620 | 16px;","text":"config.json example ","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":"font-size: 621 | 16px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 622 | 16px;","text":"(","type":"text","version":1},{"detail":0,"format":1,"mode":"normal","style":"font-size: 623 | 16px;","text":"Docker Compose self deploy","type":"text","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 624 | 16px;","text":")","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":"font-size: 625 | 16px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":"{","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"mcpServers\": 626 | {","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"everything\": 627 | {","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"url\": 628 | \"http://host.docker.internal:8080/sse\"","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" }","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" }","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":"}","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 629 | theme: pink 630 | title: '' 631 | type: '' 632 | width: 357 633 | height: 231 634 | id: '1742367595361' 635 | position: 636 | x: 690.5542795067578 637 | y: 969.1578732473977 638 | positionAbsolute: 639 | x: 690.5542795067578 640 | y: 969.1578732473977 641 | selected: false 642 | sourcePosition: right 643 | targetPosition: left 644 | type: custom-note 645 | width: 357 646 | - data: 647 | author: 3Dify-developer 648 | desc: '' 649 | height: 97 650 | selected: false 651 | showAuthor: false 652 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 653 | 16px;","text":"How to convert \"stdio\" to \"SSE\" MCP server","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":"README 654 | of my GitHub repository","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":"https://github.com/3dify-project/dify-mcp-client","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"link","version":1,"rel":"noreferrer","target":null,"title":null,"url":"https://github.com/3dify-project/dify-mcp-client"}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 655 | theme: cyan 656 | title: '' 657 | type: '' 658 | width: 447 659 | height: 97 660 | id: '1742452980902' 661 | position: 662 | x: 690.5542795067578 663 | y: 863.8984826162116 664 | positionAbsolute: 665 | x: 690.5542795067578 666 | y: 863.8984826162116 667 | selected: false 668 | sourcePosition: right 669 | targetPosition: left 670 | type: custom-note 671 | width: 447 672 | viewport: 673 | x: -33.95272842445752 674 | y: -104.19406886847042 675 | zoom: 0.8487039665217158 676 | -------------------------------------------------------------------------------- /test/chatflow/mcp_stdio_chatflow_v0.0.1.yml: -------------------------------------------------------------------------------- 1 | app: 2 | description: 'Test for MCP client as Dify Agent Strategy plugin. 3 | 4 | used "Everything MCP server" 5 | 6 | Resource and Prompt are working correctly. 7 | 8 | Tool also Succussed, except for "samplingLLM", "annotatedMessage"' 9 | icon: 🤖 10 | icon_background: '#FFEAD5' 11 | mode: advanced-chat 12 | name: MCP test 13 | use_icon_as_answer_icon: true 14 | dependencies: [] 15 | kind: app 16 | version: 0.1.5 17 | workflow: 18 | conversation_variables: [] 19 | environment_variables: [] 20 | features: 21 | file_upload: 22 | allowed_file_extensions: 23 | - .JPG 24 | - .JPEG 25 | - .PNG 26 | - .GIF 27 | - .WEBP 28 | - .SVG 29 | allowed_file_types: 30 | - image 31 | allowed_file_upload_methods: 32 | - local_file 33 | - remote_url 34 | enabled: false 35 | fileUploadConfig: 36 | audio_file_size_limit: 50 37 | batch_count_limit: 5 38 | file_size_limit: 15 39 | image_file_size_limit: 10 40 | video_file_size_limit: 100 41 | workflow_file_upload_limit: 10 42 | image: 43 | enabled: false 44 | number_limits: 3 45 | transfer_methods: 46 | - local_file 47 | - remote_url 48 | number_limits: 3 49 | opening_statement: '' 50 | retriever_resource: 51 | enabled: true 52 | sensitive_word_avoidance: 53 | enabled: false 54 | speech_to_text: 55 | enabled: false 56 | suggested_questions: [] 57 | suggested_questions_after_answer: 58 | enabled: false 59 | text_to_speech: 60 | enabled: false 61 | language: '' 62 | voice: '' 63 | graph: 64 | edges: 65 | - data: 66 | isInIteration: false 67 | sourceType: start 68 | targetType: agent 69 | id: 1741367649354-source-1741765826830-target 70 | source: '1741367649354' 71 | sourceHandle: source 72 | target: '1741765826830' 73 | targetHandle: target 74 | type: custom 75 | zIndex: 0 76 | - data: 77 | isInIteration: false 78 | sourceType: agent 79 | targetType: answer 80 | id: 1741765826830-source-answer-target 81 | source: '1741765826830' 82 | sourceHandle: source 83 | target: answer 84 | targetHandle: target 85 | type: custom 86 | zIndex: 0 87 | nodes: 88 | - data: 89 | desc: '' 90 | selected: false 91 | title: Start 92 | type: start 93 | variables: [] 94 | height: 53 95 | id: '1741367649354' 96 | position: 97 | x: 87.25729350925093 98 | y: 282 99 | positionAbsolute: 100 | x: 87.25729350925093 101 | y: 282 102 | selected: false 103 | sourcePosition: right 104 | targetPosition: left 105 | type: custom 106 | width: 243 107 | - data: 108 | answer: '{{#1741367691327.text#}}' 109 | desc: '' 110 | selected: false 111 | title: Answer 112 | type: answer 113 | variables: [] 114 | height: 101 115 | id: answer 116 | position: 117 | x: 627.4767826811337 118 | y: 282 119 | positionAbsolute: 120 | x: 627.4767826811337 121 | y: 282 122 | selected: false 123 | sourcePosition: right 124 | targetPosition: left 125 | type: custom 126 | width: 243 127 | - data: 128 | author: 3Dify-developer 129 | desc: '' 130 | height: 351 131 | selected: false 132 | showAuthor: true 133 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 134 | 16px;","text":"Mac ","type":"text","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 135 | 16px;","text":"config.json example","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":"font-size: 136 | 16px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":"{","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"mcpServers\": 137 | {","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"everything\": 138 | {","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"command\": 139 | \"npx\",","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"args\": 140 | [","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"-y\",","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"@modelcontextprotocol/server-everything\"","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" ]","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" }","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" }","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":"}","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 141 | theme: pink 142 | title: '' 143 | type: '' 144 | width: 289 145 | height: 351 146 | id: '1741579814015' 147 | position: 148 | x: 973.6650449303346 149 | y: 610.9646610182883 150 | positionAbsolute: 151 | x: 973.6650449303346 152 | y: 610.9646610182883 153 | selected: false 154 | sourcePosition: right 155 | targetPosition: left 156 | type: custom-note 157 | width: 289 158 | - data: 159 | author: 3Dify-developer 160 | desc: '' 161 | height: 194 162 | selected: false 163 | showAuthor: true 164 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 165 | 16px;","text":"User prompt base","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 166 | 16px;","text":"I''m developing MCP client as Dify Agent Strategy Plugin.","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 167 | 16px;","text":"- Some features haven''t implemented and might not work as 168 | I expected.","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 169 | 16px;","text":"- I found good MCP server for testing.","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 170 | 16px;","text":"- You should ignore \"Current Time Tool\". ","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":"font-size: 171 | 16px;"},{"children":[],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 172 | theme: blue 173 | title: '' 174 | type: '' 175 | width: 569 176 | height: 194 177 | id: '1741624678131' 178 | position: 179 | x: 87.25729350925093 180 | y: 508.8200615915433 181 | positionAbsolute: 182 | x: 87.25729350925093 183 | y: 508.8200615915433 184 | selected: false 185 | sourcePosition: right 186 | targetPosition: left 187 | type: custom-note 188 | width: 569 189 | - data: 190 | author: 3Dify-developer 191 | desc: '' 192 | height: 88 193 | selected: false 194 | showAuthor: true 195 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 196 | 16px;","text":"Sample MCP server for testing","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":"font-size: 197 | 16px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":"https://github.com/modelcontextprotocol/servers/tree/main/src/everything","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 198 | theme: cyan 199 | title: '' 200 | type: '' 201 | width: 449 202 | height: 88 203 | id: '1741626728667' 204 | position: 205 | x: 737.2683807703559 206 | y: 508.8200615915433 207 | positionAbsolute: 208 | x: 737.2683807703559 209 | y: 508.8200615915433 210 | selected: false 211 | sourcePosition: right 212 | targetPosition: left 213 | type: custom-note 214 | width: 449 215 | - data: 216 | author: 3Dify-developer 217 | desc: '' 218 | height: 151 219 | selected: false 220 | showAuthor: true 221 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 222 | 16px;","text":"MCP Tool Test prompt","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 223 | 16px;","text":"Try mcp tools one-by-one","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":"font-size: 224 | 16px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 225 | 16px;","text":"- judge result in \"thought\" phase (I want to know at least 226 | Fail or Success)","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 227 | 16px;","text":"- avoid \"sampleLLM\" which is not compatible feature yet. 228 | ","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 229 | theme: violet 230 | title: '' 231 | type: '' 232 | width: 570 233 | height: 151 234 | id: '1741669149712' 235 | position: 236 | x: 87.25729350925093 237 | y: 718.2693074279789 238 | positionAbsolute: 239 | x: 87.25729350925093 240 | y: 718.2693074279789 241 | selected: false 242 | sourcePosition: right 243 | targetPosition: left 244 | type: custom-note 245 | width: 570 246 | - data: 247 | author: 3Dify-developer 248 | desc: '' 249 | height: 140 250 | selected: false 251 | showAuthor: true 252 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 253 | 16px;","text":"MCP Resource Test prompt","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 254 | 16px;","text":"Try mcp resources one-by-one","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":"font-size: 255 | 16px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 256 | 16px;","text":"- judge result in \"thought\" phase (I want to know at least 257 | Fail or Success)","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 258 | theme: yellow 259 | title: '' 260 | type: '' 261 | width: 568 262 | height: 140 263 | id: '1741669231167' 264 | position: 265 | x: 88.65225939320432 266 | y: 879.6113451588606 267 | positionAbsolute: 268 | x: 88.65225939320432 269 | y: 879.6113451588606 270 | selected: false 271 | sourcePosition: right 272 | targetPosition: left 273 | type: custom-note 274 | width: 568 275 | - data: 276 | author: 3Dify-developer 277 | desc: '' 278 | height: 128 279 | selected: false 280 | showAuthor: true 281 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 282 | 16px;","text":"MCP Prompt Test prompt","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 283 | 16px;","text":"Try mcp prompts one-by-one","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":"font-size: 284 | 16px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 285 | 16px;","text":"- judge result in \"thought\" phase (I want to know at least 286 | Fail or Success)","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 287 | theme: green 288 | title: '' 289 | type: '' 290 | width: 568 291 | height: 128 292 | id: '1741669352719' 293 | position: 294 | x: 87.25729350925093 295 | y: 1033.9889783833733 296 | positionAbsolute: 297 | x: 87.25729350925093 298 | y: 1033.9889783833733 299 | selected: false 300 | sourcePosition: right 301 | targetPosition: left 302 | type: custom-note 303 | width: 568 304 | - data: 305 | author: 3Dify-developer 306 | desc: '' 307 | height: 354 308 | selected: false 309 | showAuthor: true 310 | text: '{"root":{"children":[{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 311 | 16px;","text":"Windows ","type":"text","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 312 | 16px;","text":"config.json example","type":"text","version":1},{"type":"linebreak","version":1},{"detail":0,"format":0,"mode":"normal","style":"","text":"{","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"mcpServers\": 313 | {","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"everything\": 314 | {","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"command\": 315 | \"npx.cmd\",","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"args\": 316 | [","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"-y\",","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" \"@modelcontextprotocol/server-everything\"","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" ]","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" }","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":" }","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"","text":"}","type":"text","version":1}],"direction":null,"format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""},{"children":[{"detail":0,"format":1,"mode":"normal","style":"font-size: 317 | 16px;","text":"warnning","type":"text","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 318 | 16px;","text":": use ","type":"text","version":1},{"detail":0,"format":1,"mode":"normal","style":"font-size: 319 | 16px;","text":"npx.cmd","type":"text","version":1},{"detail":0,"format":0,"mode":"normal","style":"font-size: 320 | 16px;","text":" ","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":1,"textStyle":""},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 321 | 16px;","text":" instead of npx","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":""}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 322 | theme: pink 323 | title: '' 324 | type: '' 325 | width: 288 326 | height: 354 327 | id: '1741671860068' 328 | position: 329 | x: 675.9865130991246 330 | y: 610.9646610182883 331 | positionAbsolute: 332 | x: 675.9865130991246 333 | y: 610.9646610182883 334 | selected: false 335 | sourcePosition: right 336 | targetPosition: left 337 | type: custom-note 338 | width: 288 339 | - data: 340 | agent_parameters: 341 | config_json: 342 | type: constant 343 | value: "{\n \"mcpServers\": {\n \"everything\": {\n \"command\"\ 344 | : \"npx.cmd\",\n \"args\": [\n \"-y\",\n \"@modelcontextprotocol/server-everything\"\ 345 | \n ]\n }\n }\n}" 346 | instruction: 347 | type: constant 348 | value: You are helpful LLM agent. 349 | maximum_iterations: 350 | type: constant 351 | value: 3 352 | model: 353 | type: constant 354 | value: 355 | completion_params: {} 356 | mode: chat 357 | model: gemini-2.0-flash-exp 358 | model_type: llm 359 | provider: langgenius/gemini/google 360 | type: model-selector 361 | query: 362 | type: constant 363 | value: '{{#sys.query#}}' 364 | tools: 365 | type: constant 366 | value: [] 367 | agent_strategy_label: mcpReAct 368 | agent_strategy_name: mcpReAct 369 | agent_strategy_provider_name: 3dify-project/mcp_client/agent 370 | desc: '' 371 | output_schema: null 372 | plugin_unique_identifier: 3dify-project/mcp_client:0.0.1@eac12d97acd3c57819aae9ee561591b60708e1cf09bdaa3ba38fce60221ad50d 373 | selected: false 374 | title: Agent 375 | type: agent 376 | height: 145 377 | id: '1741765826830' 378 | position: 379 | x: 347.36834087720297 380 | y: 282 381 | positionAbsolute: 382 | x: 347.36834087720297 383 | y: 282 384 | selected: false 385 | sourcePosition: right 386 | targetPosition: left 387 | type: custom 388 | width: 243 389 | - data: 390 | author: 3Dify-developer 391 | desc: '' 392 | height: 93 393 | selected: false 394 | showAuthor: true 395 | text: '{"root":{"children":[{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 396 | 14px;","text":"TOOL LIST should not blank. Choose some plugin.","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":"font-size: 397 | 14px;"},{"children":[{"detail":0,"format":0,"mode":"normal","style":"font-size: 398 | 14px;","text":"I recommend official Dify plugin like \"Current Time\".","type":"text","version":1}],"direction":"ltr","format":"","indent":0,"type":"paragraph","version":1,"textFormat":0,"textStyle":"font-size: 399 | 14px;"}],"direction":"ltr","format":"","indent":0,"type":"root","version":1}}' 400 | theme: yellow 401 | title: '' 402 | type: '' 403 | width: 363 404 | height: 93 405 | id: '1741766414431' 406 | position: 407 | x: 289.33163665746144 408 | y: 178.12817056680163 409 | positionAbsolute: 410 | x: 289.33163665746144 411 | y: 178.12817056680163 412 | selected: true 413 | sourcePosition: right 414 | targetPosition: left 415 | type: custom-note 416 | width: 363 417 | viewport: 418 | x: 184.66257640044716 419 | y: -96.37762447627853 420 | zoom: 0.8615237662477786 421 | -------------------------------------------------------------------------------- /test/result/everything_mcp_server_prompt_log.txt: -------------------------------------------------------------------------------- 1 | ROUND 1 2 | { 3 | "action_input": {}, 4 | "action_name": "mcp_prompt_simple_prompt", 5 | "observation": "user: This is a simple prompt without arguments.", 6 | "thought": "Okay, I understand. I will try the MCP prompts one by one and judge the result in the \"thought\" phase, indicating either \"Success\" or \"Fail\" based on whether the tool call was successful and returned a value. I will ignore the \"Current Time Tool.\"\n\nFirst, I will try the `mcp_prompt_simple_prompt`." 7 | } 8 | 9 | ROUND 2 10 | { 11 | "action_input": { 12 | "style": "verbose", 13 | "temperature": "0.7" 14 | }, 15 | "action_name": "mcp_prompt_complex_prompt", 16 | "observation": "user: This is a complex prompt with arguments: temperature=0.7, style=verbose\nassistant: I understand. You've provided a complex prompt with temperature and style arguments. How would you like me to proceed?\nuser: type='image' data='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' mimeType='image/png'", 17 | "thought": "The `mcp_prompt_simple_prompt` executed successfully and returned a string. Result: Success.\n\nNow I will try the `mcp_prompt_complex_prompt`. This prompt requires arguments, specifically `temperature` and `style`. I will set `temperature` to \"0.7\" and `style` to \"verbose\"." 18 | } 19 | 20 | ROUND 3 21 | { 22 | "action_input": {}, 23 | "action_name": "mcp_prompt_tool_use", 24 | "observation": "there is not a tool named mcp_prompt_tool_use", 25 | "thought": "The `mcp_prompt_complex_prompt` appeared to execute successfully but returned an image as a string encoded as a base64. While not a typical text result, it's still a valid response. Result: Success.\n\nNow I will try the `mcp_prompt_tool_use`. This prompt is intended to test the agent's ability to use tools, and will likely require interaction." 26 | } 27 | -------------------------------------------------------------------------------- /test/result/everything_mcp_server_resource_log.txt: -------------------------------------------------------------------------------- 1 | ROUND 1 2 | { 3 | "action_input": {}, 4 | "action_name": "mcp_resource_Resource 1", 5 | "observation": "Resource content (text/plain): Resource 1: This is a plaintext resource", 6 | "thought": "Okay, I understand. I will attempt to read each of the MCP resources one by one, and after each attempt, I will indicate in my \"thought\" whether the attempt was a success or a failure. I will ignore the \"Current Time Tool\".\n\nFirst, I will attempt to read 'Resource 1'." 7 | } 8 | 9 | ROUND 2 10 | { 11 | "action_input": {}, 12 | "action_name": "mcp_resource_Resource 2", 13 | "observation": "Binary resource (application/octet-stream) of size 44 bytes", 14 | "thought": "Okay, reading 'Resource 1' was successful. Now I will attempt to read 'Resource 2'." 15 | } 16 | 17 | ROUND 3 18 | { 19 | "action_input": {}, 20 | "action_name": "mcp_resource_Resource 3", 21 | "observation": "Resource content (text/plain): Resource 3: This is a plaintext resource", 22 | "thought": "Reading 'Resource 2' also appears to have been successful since I received a binary resource. Now I will attempt to read 'Resource 3'." 23 | } 24 | 25 | ROUND 4 26 | { 27 | "action_input": {}, 28 | "action_name": "mcp_resource_Resource 4", 29 | "observation": "Binary resource (application/octet-stream) of size 44 bytes", 30 | "thought": "Okay, reading 'Resource 3' was successful. Now I will attempt to read 'Resource 4'." 31 | } 32 | 33 | ROUND 5 34 | { 35 | "action_input": {}, 36 | "action_name": "mcp_resource_Resource 5", 37 | "observation": "Resource content (text/plain): Resource 5: This is a plaintext resource", 38 | "thought": "Reading 'Resource 4' also appears to have been successful. Now I will attempt to read 'Resource 5'." 39 | } 40 | 41 | ROUND 6 42 | { 43 | "action_input": {}, 44 | "action_name": "mcp_resource_Resource 6", 45 | "observation": "Binary resource (application/octet-stream) of size 44 bytes", 46 | "thought": "Okay, reading 'Resource 5' was successful. Now I will attempt to read 'Resource 6'." 47 | } 48 | 49 | ROUND 7 50 | { 51 | "action_input": {}, 52 | "action_name": "mcp_resource_Resource 7", 53 | "observation": "Resource content (text/plain): Resource 7: This is a plaintext resource", 54 | "thought": "Reading 'Resource 6' also appears to have been successful. Now I will attempt to read 'Resource 7'." 55 | } 56 | 57 | ROUND 8 58 | { 59 | "action_input": {}, 60 | "action_name": "mcp_resource_Resource 8", 61 | "observation": "Binary resource (application/octet-stream) of size 44 bytes", 62 | "thought": "Okay, reading 'Resource 7' was successful. Now I will attempt to read 'Resource 8'." 63 | } 64 | 65 | ROUND 9 66 | { 67 | "action_input": {}, 68 | "action_name": "mcp_resource_Resource 9", 69 | "observation": "Resource content (text/plain): Resource 9: This is a plaintext resource", 70 | "thought": "Reading 'Resource 8' also appears to have been successful. Now I will attempt to read 'Resource 9'." 71 | } 72 | 73 | ROUND 10 74 | { 75 | "action_input": {}, 76 | "action_name": "mcp_resource_Resource 10", 77 | "observation": "Binary resource (application/octet-stream) of size 48 bytes", 78 | "thought": "Okay, reading 'Resource 9' was successful. Now I will attempt to read 'Resource 10'." 79 | } 80 | 81 | ROUND 11 82 | { 83 | "action_input": {}, 84 | "action_name": "mcp_resource_Resource 11", 85 | "observation": "there is not a tool named mcp_resource_Resource 11", 86 | "thought": "Reading 'Resource 10' also appears to have been successful. Now I will attempt to read 'Resource 11'." 87 | } 88 | -------------------------------------------------------------------------------- /test/result/everything_mcp_server_tool_log.txt: -------------------------------------------------------------------------------- 1 | ROUND 1 2 | { 3 | "action_input": { 4 | "message": "Hello, MCP server!" 5 | }, 6 | "action_name": "mcp_tool_echo", 7 | "observation": "Echo: Hello, MCP server!", 8 | "thought": "Okay, I understand. I will try the MCP tools one by one, judging the result in the \"thought\" phase and avoiding the \"sampleLLM\" tool. I will also ignore the \"Current Time Tool\". Let's start with the `mcp_tool_echo` tool.\n\n I will start with the `mcp_tool_echo` tool to check if I can communicate with the MCP server. This should be a simple test to see if the basic connection is working." 9 | } 10 | 11 | ROUND 2 12 | { 13 | "action_input": { 14 | "a": 5, 15 | "b": 7 16 | }, 17 | "action_name": "mcp_tool_add", 18 | "observation": "The sum of 5 and 7 is 12.", 19 | "thought": "The `mcp_tool_echo` tool was successful. Now, let's try the `mcp_tool_add` tool to see if basic arithmetic operations are working. This will help verify that the MCP server can process requests with numerical data." 20 | } 21 | 22 | ROUND 3 23 | { 24 | "action_input": {}, 25 | "action_name": "mcp_tool_printEnv", 26 | "observation": "{\n \"APPDATA\": \"C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Roaming\",\n \"COLOR\": \"0\",\n \"COMSPEC\": \"C:\\\\Windows\\\\system32\\\\cmd.exe\",\n \"EDITOR\": \"C:\\\\Windows\\\\notepad.exe\",\n \"HOME\": \"C:\\\\Users\\\\USER_NAME\",\n \"HOMEDRIVE\": \"C:\",\n \"HOMEPATH\": \"\\\\Users\\\\USER_NAME\",\n \"INIT_CWD\": \"C:\\\\Users\\\\USER_NAME\\\\dify_mcp_plugin\\\\mcp_client\",\n \"LOCALAPPDATA\": \"C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Local\",\n \"NODE\": \"C:\\\\Program Files\\\\nodejs\\\\node.exe\",\n \"NODE_EXE\": \"C:\\\\Program Files\\\\nodejs\\\\\\\\node.exe\",\n \"npm_command\": \"exec\",\n \"npm_config_cache\": \"C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Local\\\\npm-cache\",\n \"npm_config_globalconfig\": \"C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Roaming\\\\npm\\\\etc\\\\npmrc\",\n \"npm_config_global_prefix\": \"C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Roaming\\\\npm\",\n \"npm_config_init_module\": \"C:\\\\Users\\\\USER_NAME\\\\.npm-init.js\",\n \"npm_config_local_prefix\": \"C:\\\\Users\\\\USER_NAME\",\n \"npm_config_node_gyp\": \"C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Roaming\\\\npm\\\\node_modules\\\\npm\\\\node_modules\\\\node-gyp\\\\bin\\\\node-gyp.js\",\n \"npm_config_noproxy\": \"\",\n \"npm_config_npm_version\": \"10.8.1\",\n \"npm_config_prefix\": \"C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Roaming\\\\npm\",\n \"npm_config_userconfig\": \"C:\\\\Users\\\\USER_NAME\\\\.npmrc\",\n \"npm_config_user_agent\": \"npm/10.8.1 node/v20.15.0 win32 x64 workspaces/false\",\n \"npm_config_yes\": \"true\",\n \"npm_execpath\": \"C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Roaming\\\\npm\\\\node_modules\\\\npm\\\\bin\\\\npm-cli.js\",\n \"npm_lifecycle_event\": \"npx\",\n \"npm_lifecycle_script\": \"mcp-server-everything\",\n \"npm_node_execpath\": \"C:\\\\Program Files\\\\nodejs\\\\node.exe\",\n \"npm_package_json\": \"C:\\\\Users\\\\USER_NAME\\\\dify_mcp_plugin\\\\mcp_client\\\\package.json\",\n \"NPM_PREFIX_JS\": \"C:\\\\Program Files\\\\nodejs\\\\\\\\node_modules\\\\npm\\\\bin\\\\npm-prefix.js\",\n \"NPM_PREFIX_NPX_CLI_JS\": \"C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Roaming\\\\npm\\\\node_modules\\\\npm\\\\bin\\\\npx-cli.js\",\n \"NPX_CLI_JS\": \"C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Roaming\\\\npm\\\\node_modules\\\\npm\\\\bin\\\\npx-cli.js\",\n \"PATH\": \"C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Local\\\\npm-cache\\\\_npx\\\\5b2dd62b9d0bddd4\\\\node_modules\\\\.bin;C:\\\\Users\\\\USER_NAME\\\\dify_mcp_plugin\\\\mcp_client\\\\node_modules\\\\.bin;C:\\\\Users\\\\USER_NAME\\\\dify_mcp_plugin\\\\node_modules\\\\.bin;C:\\\\Users\\\\USER_NAME\\\\node_modules\\\\.bin;C:\\\\Users\\\\node_modules\\\\.bin;C:\\\\node_modules\\\\.bin;C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Roaming\\\\npm\\\\node_modules\\\\npm\\\\node_modules\\\\@npmcli\\\\run-script\\\\lib\\\\node-gyp-bin;C:\\\\Users\\\\USER_NAME\\\\dify_mcp_plugin\\\\.venv\\\\Scripts;C:\\\\Program Files (x86)\\\\Common Files\\\\Intel\\\\Shared Libraries\\\\redist\\\\intel64\\\\compiler;C:\\\\Windows\\\\system32;C:\\\\Windows;C:\\\\Windows\\\\System32\\\\Wbem;C:\\\\Windows\\\\System32\\\\WindowsPowerShell\\\\v1.0\\\\;C:\\\\Windows\\\\System32\\\\OpenSSH\\\\;C:\\\\Program Files\\\\NVIDIA Corporation\\\\NVIDIA NvDLISR;C:\\\\Program Files\\\\Git\\\\cmd;C:\\\\Program Files\\\\dotnet\\\\;C:\\\\Program Files (x86)\\\\NVIDIA Corporation\\\\PhysX\\\\Common;C:\\\\Program Files\\\\nodejs\\\\;C:\\\\Program Files (x86)\\\\Incredibuild;C:\\\\Program Files (x86)\\\\Windows Kits\\\\10\\\\Windows Performance Toolkit\\\\;C:\\\\Program Files\\\\Docker\\\\Docker\\\\resources\\\\bin;C:\\\\Program Files\\\\Graphviz\\\\bin;C:\\\\Users\\\\USER_NAME\\\\.local\\\\bin;C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Local\\\\Programs\\\\Python\\\\Python310\\\\Scripts\\\\;C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Local\\\\Programs\\\\Python\\\\Python310\\\\;C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Local\\\\Microsoft\\\\WindowsApps;C:\\\\Users\\\\USER_NAME\\\\.dotnet\\\\tools;C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Roaming\\\\npm;C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Local\\\\Programs\\\\Ollama\",\n \"PATHEXT\": \".COM;.EXE;.BAT;.CMD;.VBS;.JS;.WS;.MSC\",\n \"PROCESSOR_ARCHITECTURE\": \"AMD64\",\n \"PROMPT\": \"$P$G\",\n \"SYSTEMDRIVE\": \"C:\",\n \"SYSTEMROOT\": \"C:\\\\Windows\",\n \"TEMP\": \"C:\\\\Users\\\\USER_NAME\\\\AppData\\\\Local\\\\Temp\",\n \"USERNAME\": \"USER_NAME\",\n \"USERPROFILE\": \"C:\\\\Users\\\\USER_NAME\"\n}", 27 | "thought": "The `mcp_tool_add` tool was also successful. Now I'll try the `mcp_tool_printEnv` tool to see the environment variables on the MCP server. This might be helpful for understanding the server's configuration." 28 | } 29 | 30 | ROUND 4 31 | { 32 | "action_input": { 33 | "duration": 5, 34 | "steps": 5 35 | }, 36 | "action_name": "mcp_tool_longRunningOperation", 37 | "observation": "Long running operation completed. Duration: 5 seconds, Steps: 5.", 38 | "thought": "The `mcp_tool_printEnv` tool was successful. Now let's try `mcp_tool_longRunningOperation`. I'll set the duration to 5 seconds and 5 steps to observe the progress updates." 39 | } 40 | 41 | ROUND 5 42 | { 43 | "action_input": {}, 44 | "action_name": "mcp_tool_getTinyImage", 45 | "observation": "This is a tiny image:type='image' data='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' mimeType='image/png'The image above is the MCP tiny image.", 46 | "thought": "The `mcp_tool_longRunningOperation` was successful. Now let's try the `mcp_tool_getTinyImage` tool." 47 | } 48 | 49 | ROUND 6 50 | { 51 | "action_input": { 52 | "iterations": 1000 53 | }, 54 | "action_name": "mcp_tool_calculate_pi", 55 | "observation": "there is not a tool named mcp_tool_calculate_pi", 56 | "thought": "The `mcp_tool_getTinyImage` tool was successful. Now let's move on to the `mcp_tool_calculate_pi` tool. This tool is computationally intensive, and the MCP server may not have sufficient resources or the implementation might be flawed. I'll run it with a relatively small number of iterations to test its basic functionality." 57 | } 58 | --------------------------------------------------------------------------------