├── .dockerignore ├── images ├── app.png └── diagram.png ├── .gitignore ├── utils.py ├── requirements.txt ├── bot.Dockerfile ├── pull_model.Dockerfile ├── .example.env ├── CONTRIBUTING.md ├── docker-compose.yml ├── db.py ├── README.md ├── chains.py ├── agent.py ├── LICENSE └── bot.py /.dockerignore: -------------------------------------------------------------------------------- 1 | * 2 | !*.py 3 | !requirements.txt -------------------------------------------------------------------------------- /images/app.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/dockersamples/CodeExplorer/HEAD/images/app.png -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- 1 | .env 2 | data/ 3 | embedding_model/ 4 | .DS_Store 5 | __pycache__ 6 | .venv 7 | test -------------------------------------------------------------------------------- /images/diagram.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/dockersamples/CodeExplorer/HEAD/images/diagram.png -------------------------------------------------------------------------------- /utils.py: -------------------------------------------------------------------------------- 1 | class BaseLogger: 2 | def __init__(self) -> None: 3 | self.info = print 4 | -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | openai==0.27.10 2 | python-dotenv==1.0.0 3 | wikipedia==1.4.0 4 | tiktoken==0.5.1 5 | neo4j==5.2.1 6 | streamlit==1.27.2 7 | # sentence_transformers==2.2.2 8 | transformers==4.34.1 9 | Pillow==10.1.0 10 | fastapi==0.103.2 11 | PyPDF2==3.0.1 12 | # torch==2.0.1 13 | pydantic==1.10.13 14 | uvicorn==0.23.2 15 | sse-starlette==1.6.5 16 | boto3==1.28.69 17 | numexpr==2.8.7 18 | # langchain==0.0.323 19 | -------------------------------------------------------------------------------- /bot.Dockerfile: -------------------------------------------------------------------------------- 1 | FROM langchain/langchain:0.0.323 2 | 3 | WORKDIR /app 4 | 5 | RUN apt-get update && apt-get install -y \ 6 | build-essential \ 7 | curl \ 8 | software-properties-common \ 9 | && rm -rf /var/lib/apt/lists/* 10 | 11 | COPY requirements.txt . 12 | 13 | RUN pip install -r requirements.txt 14 | 15 | # set email for streamlit so it doesn't ask in docker container 16 | RUN mkdir -p ~/.streamlit/ 17 | RUN echo "[general]" > ~/.streamlit/credentials.toml 18 | RUN echo "email = \"\"" >> ~/.streamlit/credentials.toml 19 | 20 | COPY *.py . 21 | 22 | EXPOSE 8501 23 | 24 | HEALTHCHECK CMD curl --fail http://localhost:8501/_stcore/health 25 | 26 | ENTRYPOINT ["streamlit", "run", "bot.py", "--server.port=8501", "--server.address=0.0.0.0"] 27 | -------------------------------------------------------------------------------- /pull_model.Dockerfile: -------------------------------------------------------------------------------- 1 | #syntax = docker/dockerfile:1.4 2 | 3 | FROM ollama/ollama:latest AS ollama 4 | FROM babashka/babashka:latest 5 | 6 | # just using as a client - never as a server 7 | COPY --from=ollama /bin/ollama ./bin/ollama 8 | 9 | COPY <!! done :stop)) 31 | 32 | (println "OLLAMA model only pulled if both LLM and OLLAMA_BASE_URL are set and the LLM model is not gpt"))) 33 | (catch Throwable _ (System/exit 1))) 34 | EOF 35 | 36 | ENTRYPOINT ["bb", "-f", "pull_model.clj"] 37 | 38 | -------------------------------------------------------------------------------- /.example.env: -------------------------------------------------------------------------------- 1 | #***************************************************************** 2 | # LLM and Embedding Model 3 | #***************************************************************** 4 | LLM=codellama:7b-instruct #or any Ollama model tag, gpt-4, gpt-3.5, or claudev2 5 | EMBEDDING_MODEL=ollama #or sentence_transformer, openai, ollama, or aws 6 | 7 | #***************************************************************** 8 | # Neo4j 9 | #***************************************************************** 10 | NEO4J_URI=neo4j://database:7687 11 | NEO4J_USERNAME=neo4j 12 | NEO4J_PASSWORD=password 13 | 14 | #***************************************************************** 15 | # Langchain 16 | #***************************************************************** 17 | # Optional for enabling Langchain Smith API 18 | 19 | #LANGCHAIN_TRACING_V2=true # false 20 | #LANGCHAIN_ENDPOINT="https://api.smith.langchain.com" 21 | #LANGCHAIN_PROJECT=#your-project-name 22 | #LANGCHAIN_API_KEY=#your-api-key ls_... 23 | 24 | #***************************************************************** 25 | # Ollama 26 | #***************************************************************** 27 | OLLAMA_BASE_URL=http://host.docker.internal:11434 28 | # OLLAMA_BASE_URL=http://llm:11434 29 | 30 | #***************************************************************** 31 | # OpenAI 32 | #***************************************************************** 33 | # Only required when using OpenAI LLM or embedding model 34 | 35 | #OPENAI_API_KEY=sk-... 36 | 37 | #***************************************************************** 38 | # AWS 39 | #***************************************************************** 40 | # Only required when using AWS Bedrock LLM or embedding model 41 | 42 | #AWS_ACCESS_KEY_ID= 43 | #AWS_SECRET_ACCESS_KEY= 44 | #AWS_DEFAULT_REGION=us-east-1 -------------------------------------------------------------------------------- /CONTRIBUTING.md: -------------------------------------------------------------------------------- 1 | # Contributing to Code Explorer 2 | 3 | First off, thank you for considering contributing to Code Explorer! We're excited to have you on board. 4 | 5 | Please take a moment to review this document in order to make the contribution process easy and effective for everyone involved. 6 | 7 | ## Code of Conduct 8 | 9 | This project and everyone participating in it is governed by the [Code Explorer Code of Conduct](CODE_OF_CONDUCT.md). By participating, you are expected to uphold this code. Please report unacceptable behavior to the project maintainers. 10 | 11 | ## How Can I Contribute? 12 | 13 | ### Reporting Bugs 14 | 15 | If you find a bug in Code Explorer, please open an issue on the [GitHub issue tracker](https://github.com/your-username/CodeExplorer/issues). Be sure to include a clear title and description of the problem, as well as any relevant details such as the version of Code Explorer you're using. 16 | 17 | ### Suggesting Enhancements 18 | 19 | If you have an idea for a new feature or improvement to Code Explorer, please open an issue on the [GitHub issue tracker](https://github.com/dockersamples/CodeExplorer/issues). Be sure to provide a clear title and description of your suggestion, as well as any relevant details or examples. 20 | 21 | ### Pull Requests 22 | 23 | If you'd like to contribute code to Code Explorer, please follow these steps: 24 | 25 | 1. Fork the repository and create a new branch for your changes. 26 | 2. Make your changes and ensure that your code is well-formatted and follows the project's coding style guidelines. 27 | 3. Write clear and concise commit messages that describe your changes. 28 | 4. Open a pull request on the [GitHub repository](https://github.com/dockersamples/CodeExplorer) and provide a clear title and description of your changes. 29 | 5. Wait for feedback from the project maintainers and make any necessary adjustments. 30 | 6. Once your pull request is approved, it will be merged into the main branch. 31 | 32 | ## Styleguides 33 | 34 | ### Git Commit Messages 35 | 36 | * Use the present tense ("Add feature" not "Added feature") 37 | * Use the imperative mood ("Move cursor to..." not "Moves cursor to...") 38 | * Limit the first line to 72 characters or less 39 | * Reference issues and pull requests using the `#` symbol 40 | 41 | ### Python Code 42 | 43 | * Follow the PEP 8 style guide 44 | * Use 4 spaces per indentation level 45 | * Use blank lines to separate functions and classes, and larger blocks of code inside functions 46 | * Add comments to explain complex sections of code 47 | 48 | ### Documentation 49 | 50 | * Use Markdown for documentation 51 | * Keep documentation up-to-date with any changes to the code 52 | * Use clear and concise language 53 | 54 | Thank you for your interest in contributing to Code Explorer! We're excited to see what you can bring to the project. 55 | -------------------------------------------------------------------------------- /docker-compose.yml: -------------------------------------------------------------------------------- 1 | services: 2 | llm: 3 | image: ollama/ollama:latest 4 | profiles: ['linux'] 5 | networks: 6 | - net 7 | # deploy: 8 | # resources: 9 | # reservations: 10 | # devices: 11 | # - driver: nvidia 12 | # count: all 13 | # capabilities: [gpu] 14 | 15 | pull-model: 16 | build: 17 | dockerfile: pull_model.Dockerfile 18 | environment: 19 | - OLLAMA_BASE_URL=${OLLAMA_BASE_URL-http://host.docker.internal:11434} 20 | - LLM=${LLM-codellama:7b-instruct} 21 | networks: 22 | - net 23 | depends_on: 24 | - llm 25 | 26 | database: 27 | image: neo4j:5.11 28 | ports: 29 | - 7687:7687 30 | - 7474:7474 31 | volumes: 32 | - ./data:/data 33 | environment: 34 | - NEO4J_AUTH=${NEO4J_USERNAME-neo4j}/${NEO4J_PASSWORD-password} 35 | - NEO4J_PLUGINS=["apoc"] 36 | - NEO4J_db_tx__log_rotation_retention__policy=false 37 | healthcheck: 38 | test: 39 | [ 40 | 'CMD-SHELL', 41 | 'wget --no-verbose --tries=1 --spider localhost:7474 || exit 1' 42 | ] 43 | interval: 5s 44 | timeout: 3s 45 | retries: 5 46 | networks: 47 | - net 48 | 49 | bot: 50 | build: 51 | dockerfile: bot.Dockerfile 52 | volumes: 53 | - ./embedding_model:/embedding_model 54 | - type: bind 55 | source: /home/${whoami} 56 | target: /home/${whoami} 57 | environment: 58 | - NEO4J_URI=${NEO4J_URI-neo4j://database:7687} 59 | - NEO4J_PASSWORD=${NEO4J_PASSWORD-password} 60 | - NEO4J_USERNAME=${NEO4J_USERNAME-neo4j} 61 | - OPENAI_API_KEY=${OPENAI_API_KEY-} 62 | - OLLAMA_BASE_URL=${OLLAMA_BASE_URL-http://host.docker.internal:11434} 63 | - LLM=${LLM-codellama:7b-instruct} 64 | - EMBEDDING_MODEL=${EMBEDDING_MODEL-ollama} 65 | - LANGCHAIN_ENDPOINT=${LANGCHAIN_ENDPOINT-"https://api.smith.langchain.com"} 66 | - LANGCHAIN_TRACING_V2=${LANGCHAIN_TRACING_V2-false} 67 | - LANGCHAIN_PROJECT=${LANGCHAIN_PROJECT} 68 | - LANGCHAIN_API_KEY=${LANGCHAIN_API_KEY} 69 | - AWS_ACCESS_KEY_ID=${AWS_ACCESS_KEY_ID} 70 | - AWS_SECRET_ACCESS_KEY=${AWS_SECRET_ACCESS_KEY} 71 | - AWS_DEFAULT_REGION=${AWS_DEFAULT_REGION} 72 | networks: 73 | - net 74 | depends_on: 75 | database: 76 | condition: service_healthy 77 | pull-model: 78 | condition: service_completed_successfully 79 | develop: 80 | watch: 81 | - action: rebuild 82 | path: . 83 | ports: 84 | - 8501:8501 85 | 86 | networks: 87 | net: 88 | -------------------------------------------------------------------------------- /db.py: -------------------------------------------------------------------------------- 1 | import os 2 | from streamlit.logger import get_logger 3 | from dotenv import load_dotenv 4 | from chains import ( 5 | load_embedding_model 6 | ) 7 | from langchain.text_splitter import RecursiveCharacterTextSplitter 8 | from langchain.vectorstores.neo4j_vector import Neo4jVector 9 | from langchain.text_splitter import Language 10 | from langchain.document_loaders.generic import GenericLoader 11 | from langchain.document_loaders.parsers import LanguageParser 12 | from langchain.agents.conversational_chat.prompt import FORMAT_INSTRUCTIONS 13 | 14 | load_dotenv(".env") 15 | 16 | url = os.getenv("NEO4J_URI") 17 | username = os.getenv("NEO4J_USERNAME") 18 | password = os.getenv("NEO4J_PASSWORD") 19 | ollama_base_url = os.getenv("OLLAMA_BASE_URL") 20 | embedding_model_name = os.getenv("EMBEDDING_MODEL") 21 | llm_name = os.getenv("LLM") 22 | # Remapping for Langchain Neo4j integration 23 | os.environ["NEO4J_URL"] = url 24 | 25 | logger = get_logger(__name__) 26 | 27 | embeddings, dimension = load_embedding_model( 28 | embedding_model_name, config={"ollama_base_url": ollama_base_url}, logger=logger 29 | ) 30 | 31 | def process_documents(language, directory) -> (str, Neo4jVector): 32 | print("File chunking begins...", language, directory) 33 | 34 | # Create a dictionary mapping languages to file extensions 35 | language_suffix_mapping = { 36 | Language.CPP: ".cpp", 37 | Language.GO: ".go", 38 | Language.JAVA: ".java", 39 | Language.KOTLIN: ".kt", 40 | Language.JS: ".js", 41 | Language.TS: ".ts", 42 | Language.PHP: ".php", 43 | Language.PROTO: ".proto", 44 | Language.PYTHON: ".py", 45 | Language.RST: ".rst", 46 | Language.RUBY: ".rb", 47 | Language.RUST: ".rs", 48 | Language.SCALA: ".scala", 49 | Language.SWIFT: ".swift", 50 | Language.MARKDOWN: ".md", 51 | Language.LATEX: ".tex", 52 | Language.HTML: ".html", 53 | Language.SOL: ".sol", 54 | Language.CSHARP: ".cs", 55 | } 56 | # Get the corresponding suffix based on the selected language 57 | suffix = language_suffix_mapping.get(language, "") 58 | print("language file extension:", suffix) 59 | 60 | loader = GenericLoader.from_filesystem( 61 | path=directory, 62 | glob="**/*", 63 | suffixes=[suffix], 64 | parser=LanguageParser(language=language, parser_threshold=500) 65 | ) 66 | documents = loader.load() 67 | print("Total documents:", len(documents)) 68 | if len(documents) == 0: 69 | return ("0 documents found", None) 70 | 71 | text_splitter = RecursiveCharacterTextSplitter.from_language(language=language, 72 | chunk_size=5000, 73 | chunk_overlap=500) 74 | 75 | chunks = text_splitter.split_documents(documents) 76 | print("Chunks:", len(chunks)) 77 | 78 | hashStr = "myHash" # str(abs(hash(directory))) 79 | 80 | # Store the chunks part in db (vector) 81 | vectorstore = Neo4jVector.from_documents( 82 | chunks, 83 | url=url, 84 | username=username, 85 | password=password, 86 | embedding=embeddings, 87 | index_name=f"index_{hashStr}", 88 | node_label=f"node_{hashStr}", 89 | pre_delete_collection=True, # Delete existing data 90 | ) 91 | 92 | print("Files are now chunked up") 93 | 94 | return (None, vectorstore) -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # Code Explorer 2 | 3 | This app allows you to ask questions and get answers regarding your code provided the folder location of your code. It is a RAG LLM chain. The app was based on the GenAI stack from https://github.com/docker/genai-stack. 4 | 5 |

6 | App 7 |
8 |
9 | Diagram 10 |

11 | 12 | # Configure 13 | 14 | Create a `.env` file from the environment template file `.example.env` 15 | 16 | Available variables: 17 | | Variable Name | Default value | Description | 18 | |------------------------|------------------------------------|-------------------------------------------------------------------------| 19 | | OLLAMA_BASE_URL | http://host.docker.internal:11434 | REQUIRED - URL to Ollama LLM API | 20 | | NEO4J_URI | neo4j://database:7687 | REQUIRED - URL to Neo4j database | 21 | | NEO4J_USERNAME | neo4j | REQUIRED - Username for Neo4j database | 22 | | NEO4J_PASSWORD | password | REQUIRED - Password for Neo4j database | 23 | | LLM | codellama:7b-instruct | REQUIRED - Can be any Ollama model tag, or gpt-4 or gpt-3.5 or claudev2 | 24 | | EMBEDDING_MODEL | ollama | REQUIRED - Can be sentence_transformer, openai, aws or ollama | 25 | | AWS_ACCESS_KEY_ID | | REQUIRED - Only if LLM=claudev2 or embedding_model=aws | 26 | | AWS_SECRET_ACCESS_KEY | | REQUIRED - Only if LLM=claudev2 or embedding_model=aws | 27 | | AWS_DEFAULT_REGION | | REQUIRED - Only if LLM=claudev2 or embedding_model=aws | 28 | | OPENAI_API_KEY | | REQUIRED - Only if LLM=gpt-4 or LLM=gpt-3.5 or embedding_model=openai | 29 | | LANGCHAIN_ENDPOINT | "https://api.smith.langchain.com" | OPTIONAL - URL to Langchain Smith API | 30 | | LANGCHAIN_TRACING_V2 | false | OPTIONAL - Enable Langchain tracing v2 | 31 | | LANGCHAIN_PROJECT | | OPTIONAL - Langchain project name | 32 | | LANGCHAIN_API_KEY | | OPTIONAL - Langchain API key | 33 | 34 | NOTE: If using `EMBEDDING_MODEL=sentence_transformer`, uncomment code in `requirements.txt` & `chains.py`. It was commented out to reduce code size. 35 | 36 | NOTE: Make sure to set the `OLLAMA_BASE_URL=http://llm:11434` in the `.env` file when using Ollama docker container. 37 | 38 | ## Project Structure 39 | 40 | * `LICENSE`: Contains the project's license information. 41 | * `README.md`: The main README file, providing an introduction and overview of the project. 42 | * `CONTRIBUTING.md`: Guidelines for contributing to the project. 43 | * `CODE_OF_CONDUCT.md`: The code of conduct for the project. 44 | * `.env` and `.example.env`: Configuration files for environment variables. 45 | * `agent.py`, `bot.py`, `chains.py`, `db.py`, and `utils.py`: Core logic and functionality of the project. 46 | * `docker-compose.yml`: Defines and manages multi-container Docker applications. 47 | * `Dockerfile` and `bot.Dockerfile`: Instructions for building Docker images. 48 | * `requirements.txt`: Lists the project's dependencies and their versions. 49 | * `images`: Contains visual assets used in the README and other documentation. 50 | 51 | 52 | # Docker (Linux only) 53 | 54 | **Build only** 55 | 56 | ``` 57 | docker compose --profile linux build 58 | ``` 59 | 60 | **To start everything (Linux)** 61 | 62 | ``` 63 | docker compose --profile linux up --build 64 | ``` 65 | 66 | To enter **watch mode** (auto rebuild on file changes). 67 | First start everything, then in new terminal: 68 | 69 | ``` 70 | docker compose --profile linux alpha watch 71 | ``` 72 | 73 | **Shutdown** 74 | If health check fails or containers don't start up as expected, shutdown 75 | completely to start up again. 76 | 77 | ``` 78 | docker compose --profile linux down 79 | ``` 80 | 81 | # Application 82 | 83 | Access the app at http://localhost:8501. In the sidebar, enter the path to your code folder and click "Process files". Then start asking questions about your code in the main chat. The detailed mode toggle switches between asking the QA agent only (detailedMode=true) and asking an agent which uses the QA agent as a tool (detailedMode=false). In testing, the agent appears to summarize rather than giving a technical response as opposed to the QA agent only. 84 | 85 | The Neo4j vector database can be explored at http://localhost:7474. 86 | 87 | ## Query: 88 | 89 | What is the purpose of LLMSingleActionAgent() function? 90 | -------------------------------------------------------------------------------- /chains.py: -------------------------------------------------------------------------------- 1 | from langchain.embeddings.openai import OpenAIEmbeddings 2 | from langchain.embeddings import ( 3 | OllamaEmbeddings, 4 | # SentenceTransformerEmbeddings, 5 | BedrockEmbeddings, 6 | ) 7 | from langchain.chat_models import ChatOpenAI, ChatOllama, BedrockChat 8 | from langchain.chains import RetrievalQAWithSourcesChain 9 | from langchain.chains.qa_with_sources import load_qa_with_sources_chain 10 | from langchain.prompts.chat import ( 11 | ChatPromptTemplate, 12 | SystemMessagePromptTemplate, 13 | HumanMessagePromptTemplate, 14 | MessagesPlaceholder 15 | ) 16 | from typing import List, Any 17 | from utils import BaseLogger 18 | from langchain.chains import LLMChain 19 | from langchain.memory import ConversationBufferMemory 20 | 21 | 22 | def load_embedding_model(embedding_model_name: str, logger=BaseLogger(), config={}): 23 | if embedding_model_name == "ollama": 24 | embeddings = OllamaEmbeddings( 25 | base_url=config["ollama_base_url"], model="codellama:7b-instruct" 26 | ) 27 | dimension = 4096 28 | logger.info("Embedding: Using Ollama") 29 | elif embedding_model_name == "openai": 30 | embeddings = OpenAIEmbeddings() 31 | dimension = 1536 32 | logger.info("Embedding: Using OpenAI") 33 | elif embedding_model_name == "aws": 34 | embeddings = BedrockEmbeddings() 35 | dimension = 1536 36 | logger.info("Embedding: Using AWS") 37 | # else: 38 | # embeddings = SentenceTransformerEmbeddings( 39 | # model_name="all-MiniLM-L6-v2", cache_folder="./embedding_model" 40 | # ) 41 | # dimension = 384 42 | # logger.info("Embedding: Using SentenceTransformer") 43 | return embeddings, dimension 44 | 45 | 46 | def load_llm(llm_name: str, logger=BaseLogger(), config={}): 47 | if llm_name == "gpt-4": 48 | logger.info("LLM: Using GPT-4") 49 | return ChatOpenAI(temperature=0, model_name="gpt-4", streaming=True) 50 | elif llm_name == "gpt-3.5": 51 | logger.info("LLM: Using GPT-3.5") 52 | return ChatOpenAI(temperature=0, model_name="gpt-3.5-turbo", streaming=True) 53 | elif llm_name == "claudev2": 54 | logger.info("LLM: ClaudeV2") 55 | return BedrockChat( 56 | model_id="anthropic.claude-v2", 57 | model_kwargs={"temperature": 0.0, "max_tokens_to_sample": 1024}, 58 | streaming=True, 59 | ) 60 | elif len(llm_name): 61 | logger.info(f"LLM: Using Ollama: {llm_name}") 62 | return ChatOllama( 63 | temperature=0, 64 | base_url=config["ollama_base_url"], 65 | model=llm_name, 66 | streaming=True, 67 | # seed=2, 68 | top_k=10, # A higher value (100) will give more diverse answers, while a lower value (10) will be more conservative. 69 | top_p=0.3, # Higher value (0.95) will lead to more diverse text, while a lower value (0.5) will generate more focused text. 70 | num_ctx=3072, # Sets the size of the context window used to generate the next token. 71 | ) 72 | logger.info("LLM: Using GPT-3.5") 73 | return ChatOpenAI(temperature=0, model_name="gpt-3.5-turbo", streaming=True) 74 | 75 | 76 | def configure_llm_only_chain(llm): 77 | # LLM only response 78 | template = """ 79 | You are a helpful assistant that helps a support agent with answering programming questions. 80 | If you don't know the answer, just say that you don't know, you must not make up an answer. 81 | """ 82 | human_template = "{question}" 83 | 84 | chat_prompt = ChatPromptTemplate.from_messages([ 85 | SystemMessagePromptTemplate.from_template(template), # The persistent system prompt 86 | MessagesPlaceholder(variable_name="chat_history"), # Where the memory will be stored. 87 | HumanMessagePromptTemplate.from_template(human_template) # Where the human input will injected 88 | ]) 89 | 90 | memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True) 91 | chain = LLMChain( 92 | llm=llm, 93 | prompt=chat_prompt, 94 | verbose=False, 95 | memory=memory, 96 | ) 97 | 98 | def generate_llm_output( 99 | user_input: str, callbacks: List[Any] 100 | ) -> str: 101 | answer = chain.invoke(user_input, config={"callbacks": callbacks})["text"] 102 | return answer 103 | 104 | return generate_llm_output 105 | 106 | 107 | def get_qa_rag_chain(_vectorstore, llm): 108 | # Create qa RAG chain 109 | system_template = """ 110 | Use the following pieces of context to answer the question at the end. 111 | The context contains code source files which can be used to answer the question as well as be used as references. 112 | If you don't know the answer, just say that you don't know, don't try to make up an answer. 113 | ---- 114 | {summaries} 115 | ---- 116 | Generate concise answers with references to code source files at the end of every answer. 117 | """ 118 | user_template = "Question:```{question}```" 119 | chat_prompt = ChatPromptTemplate.from_messages([ 120 | SystemMessagePromptTemplate.from_template(system_template), # The persistent system prompt 121 | HumanMessagePromptTemplate.from_template(user_template), # Where the human input will injected 122 | ]) 123 | qa_chain = load_qa_with_sources_chain( 124 | llm, 125 | chain_type="stuff", 126 | prompt=chat_prompt, 127 | ) 128 | qa = RetrievalQAWithSourcesChain( 129 | combine_documents_chain=qa_chain, 130 | retriever=_vectorstore.as_retriever(search_kwargs={"k": 2}), 131 | reduce_k_below_max_tokens=False, 132 | max_tokens_limit=3375, 133 | return_source_documents=True 134 | ) 135 | 136 | return qa -------------------------------------------------------------------------------- /agent.py: -------------------------------------------------------------------------------- 1 | from langchain.memory import ConversationBufferWindowMemory 2 | from langchain.chains import LLMChain 3 | from langchain.agents import Tool 4 | from langchain.agents import Tool, AgentExecutor, LLMSingleActionAgent, AgentOutputParser 5 | from typing import List, Union 6 | from langchain.schema import AgentAction, AgentFinish, HumanMessage 7 | import re 8 | from langchain.prompts import BaseChatPromptTemplate 9 | from langchain.chains import LLMMathChain 10 | 11 | 12 | # Set up the base template 13 | template = """ 14 | Answer the following question as best you can. You have access to the following tools: 15 | 16 | {tools} 17 | 18 | Use the following format: 19 | 20 | \nQuestion: the input question you must answer 21 | \nThought: you should always think about what to do. if you have an answer to the question, then submit the final answer. 22 | \nAction: the action to take, should be one of [{tool_names}] 23 | \nAction Input: the input to the action 24 | \nObservation: the result of the action 25 | \n... (this Thought/Action/Action Input/Observation can repeat N times) 26 | \nThought: I now know the final answer 27 | \nFinal Answer: the final answer to the originals input question 28 | 29 | Begin! 30 | 31 | Previous conversation history: 32 | {chat_history} 33 | 34 | New Question: {input} 35 | {agent_scratchpad}""" 36 | 37 | 38 | # Set up a prompt template 39 | class CustomPromptTemplate(BaseChatPromptTemplate): 40 | # The template to use 41 | template: str 42 | # The list of tools available 43 | tools: List[Tool] 44 | 45 | def format_messages(self, **kwargs) -> str: 46 | # Get the intermediate steps (AgentAction, Observation tuples) 47 | # Format them in a particular way 48 | intermediate_steps = kwargs.pop("intermediate_steps") 49 | thoughts = "" 50 | for action, observation in intermediate_steps: 51 | thoughts += action.log 52 | thoughts += f"\nObservation: {observation}\nThought: " 53 | # Set the agent_scratchpad variable to that value 54 | kwargs["agent_scratchpad"] = thoughts 55 | # Create a tools variable from the list of tools provided 56 | kwargs["tools"] = "\n".join([f"{tool.name}: {tool.description}" for tool in self.tools]) 57 | # Create a list of tool names for the tools provided 58 | kwargs["tool_names"] = ", ".join([tool.name for tool in self.tools]) 59 | formatted = self.template.format(**kwargs) 60 | return [HumanMessage(content=formatted)] 61 | 62 | class CustomOutputParser(AgentOutputParser): 63 | def parse(self, llm_output: str) -> Union[AgentAction, AgentFinish]: 64 | # Check if agent should finish 65 | if "Final Answer:" in llm_output: 66 | return AgentFinish( 67 | # Return values is generally always a dictionary with a single `output` key 68 | # It is not recommended to try anything else at the moment :) 69 | return_values={"output": llm_output.split("Final Answer:")[-1].strip()}, 70 | log=llm_output, 71 | ) 72 | # Parse out the action and action input 73 | regex = r"Action\s*\d*\s*:(.*?)\nAction\s*\d*\s*Input\s*\d*\s*:[\s]*(.*)" 74 | match = re.search(regex, llm_output, re.DOTALL) 75 | if not match: 76 | # raise ValueError(f"Could not parse LLM output: `{llm_output}`") 77 | print("Could not parse LLM output, but finishing agent anyways: `{llm_output}`") 78 | return AgentFinish( 79 | # Return values is generally always a dictionary with a single `output` key 80 | # It is not recommended to try anything else at the moment :) 81 | return_values={"output": llm_output.split("Final Answer:")[-1].strip()}, 82 | log=llm_output, 83 | ) 84 | action = match.group(1).strip() 85 | action_input = match.group(2) 86 | # Return the action and action input 87 | return AgentAction(tool=action, tool_input=action_input.strip(" ").strip('"'), log=llm_output) 88 | 89 | def get_agent_executor(_qa, llm): 90 | # Create agent executor and use qa and math as tools 91 | llm_math_chain = LLMMathChain.from_llm(llm=llm) 92 | tools = [ 93 | Tool.from_function( 94 | name = "Code", 95 | func=_qa.run, 96 | description="useful for when you need to answer questions about code" 97 | ), 98 | Tool.from_function( 99 | func=llm_math_chain.run, 100 | name="Calculator", 101 | description="useful for when you need to answer questions about math" 102 | ) 103 | ] 104 | 105 | prompt = CustomPromptTemplate( 106 | template=template, 107 | tools=tools, 108 | # This omits the `agent_scratchpad`, `tools`, and `tool_names` variables because those are generated dynamically 109 | # This includes the `intermediate_steps` variable because that is needed 110 | input_variables=["input", "chat_history", "intermediate_steps"] 111 | ) 112 | output_parser = CustomOutputParser() 113 | llm_chain = LLMChain(llm=llm, prompt=prompt) 114 | tool_names = [tool.name for tool in tools] 115 | agent = LLMSingleActionAgent( 116 | llm_chain=llm_chain, 117 | handle_parsing_errors=False, 118 | output_parser=output_parser, 119 | stop=["\nObservation:", "Observation:", "\nObservation", "Observation"], 120 | allowed_tools=tool_names, 121 | max_iterations=2, 122 | ) 123 | # only retain the last message in the chat history 124 | memory = ConversationBufferWindowMemory(k=1, memory_key="chat_history", return_messages=True) 125 | agent_executor = AgentExecutor.from_agent_and_tools( 126 | agent=agent, 127 | tools=tools, 128 | verbose=True, 129 | memory=memory, 130 | max_iterations=2, 131 | ) 132 | 133 | # agent_executor = initialize_agent( 134 | # tools, 135 | # llm, 136 | # agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION, 137 | # verbose=True, 138 | # handle_parsing_errors=False, 139 | # agent_kwargs={ 140 | # 'output_parser': output_parser 141 | # }, 142 | # memory=memory, 143 | # max_iterations=2, 144 | # ) 145 | 146 | return agent_executor -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | Creative Commons Legal Code 2 | 3 | CC0 1.0 Universal 4 | 5 | CREATIVE COMMONS CORPORATION IS NOT A LAW FIRM AND DOES NOT PROVIDE 6 | LEGAL SERVICES. DISTRIBUTION OF THIS DOCUMENT DOES NOT CREATE AN 7 | ATTORNEY-CLIENT RELATIONSHIP. CREATIVE COMMONS PROVIDES THIS 8 | INFORMATION ON AN "AS-IS" BASIS. CREATIVE COMMONS MAKES NO WARRANTIES 9 | REGARDING THE USE OF THIS DOCUMENT OR THE INFORMATION OR WORKS 10 | PROVIDED HEREUNDER, AND DISCLAIMS LIABILITY FOR DAMAGES RESULTING FROM 11 | THE USE OF THIS DOCUMENT OR THE INFORMATION OR WORKS PROVIDED 12 | HEREUNDER. 13 | 14 | Statement of Purpose 15 | 16 | The laws of most jurisdictions throughout the world automatically confer 17 | exclusive Copyright and Related Rights (defined below) upon the creator 18 | and subsequent owner(s) (each and all, an "owner") of an original work of 19 | authorship and/or a database (each, a "Work"). 20 | 21 | Certain owners wish to permanently relinquish those rights to a Work for 22 | the purpose of contributing to a commons of creative, cultural and 23 | scientific works ("Commons") that the public can reliably and without fear 24 | of later claims of infringement build upon, modify, incorporate in other 25 | works, reuse and redistribute as freely as possible in any form whatsoever 26 | and for any purposes, including without limitation commercial purposes. 27 | These owners may contribute to the Commons to promote the ideal of a free 28 | culture and the further production of creative, cultural and scientific 29 | works, or to gain reputation or greater distribution for their Work in 30 | part through the use and efforts of others. 31 | 32 | For these and/or other purposes and motivations, and without any 33 | expectation of additional consideration or compensation, the person 34 | associating CC0 with a Work (the "Affirmer"), to the extent that he or she 35 | is an owner of Copyright and Related Rights in the Work, voluntarily 36 | elects to apply CC0 to the Work and publicly distribute the Work under its 37 | terms, with knowledge of his or her Copyright and Related Rights in the 38 | Work and the meaning and intended legal effect of CC0 on those rights. 39 | 40 | 1. Copyright and Related Rights. A Work made available under CC0 may be 41 | protected by copyright and related or neighboring rights ("Copyright and 42 | Related Rights"). Copyright and Related Rights include, but are not 43 | limited to, the following: 44 | 45 | i. the right to reproduce, adapt, distribute, perform, display, 46 | communicate, and translate a Work; 47 | ii. moral rights retained by the original author(s) and/or performer(s); 48 | iii. publicity and privacy rights pertaining to a person's image or 49 | likeness depicted in a Work; 50 | iv. rights protecting against unfair competition in regards to a Work, 51 | subject to the limitations in paragraph 4(a), below; 52 | v. rights protecting the extraction, dissemination, use and reuse of data 53 | in a Work; 54 | vi. database rights (such as those arising under Directive 96/9/EC of the 55 | European Parliament and of the Council of 11 March 1996 on the legal 56 | protection of databases, and under any national implementation 57 | thereof, including any amended or successor version of such 58 | directive); and 59 | vii. other similar, equivalent or corresponding rights throughout the 60 | world based on applicable law or treaty, and any national 61 | implementations thereof. 62 | 63 | 2. Waiver. To the greatest extent permitted by, but not in contravention 64 | of, applicable law, Affirmer hereby overtly, fully, permanently, 65 | irrevocably and unconditionally waives, abandons, and surrenders all of 66 | Affirmer's Copyright and Related Rights and associated claims and causes 67 | of action, whether now known or unknown (including existing as well as 68 | future claims and causes of action), in the Work (i) in all territories 69 | worldwide, (ii) for the maximum duration provided by applicable law or 70 | treaty (including future time extensions), (iii) in any current or future 71 | medium and for any number of copies, and (iv) for any purpose whatsoever, 72 | including without limitation commercial, advertising or promotional 73 | purposes (the "Waiver"). Affirmer makes the Waiver for the benefit of each 74 | member of the public at large and to the detriment of Affirmer's heirs and 75 | successors, fully intending that such Waiver shall not be subject to 76 | revocation, rescission, cancellation, termination, or any other legal or 77 | equitable action to disrupt the quiet enjoyment of the Work by the public 78 | as contemplated by Affirmer's express Statement of Purpose. 79 | 80 | 3. Public License Fallback. Should any part of the Waiver for any reason 81 | be judged legally invalid or ineffective under applicable law, then the 82 | Waiver shall be preserved to the maximum extent permitted taking into 83 | account Affirmer's express Statement of Purpose. In addition, to the 84 | extent the Waiver is so judged Affirmer hereby grants to each affected 85 | person a royalty-free, non transferable, non sublicensable, non exclusive, 86 | irrevocable and unconditional license to exercise Affirmer's Copyright and 87 | Related Rights in the Work (i) in all territories worldwide, (ii) for the 88 | maximum duration provided by applicable law or treaty (including future 89 | time extensions), (iii) in any current or future medium and for any number 90 | of copies, and (iv) for any purpose whatsoever, including without 91 | limitation commercial, advertising or promotional purposes (the 92 | "License"). The License shall be deemed effective as of the date CC0 was 93 | applied by Affirmer to the Work. Should any part of the License for any 94 | reason be judged legally invalid or ineffective under applicable law, such 95 | partial invalidity or ineffectiveness shall not invalidate the remainder 96 | of the License, and in such case Affirmer hereby affirms that he or she 97 | will not (i) exercise any of his or her remaining Copyright and Related 98 | Rights in the Work or (ii) assert any associated claims and causes of 99 | action with respect to the Work, in either case contrary to Affirmer's 100 | express Statement of Purpose. 101 | 102 | 4. Limitations and Disclaimers. 103 | 104 | a. No trademark or patent rights held by Affirmer are waived, abandoned, 105 | surrendered, licensed or otherwise affected by this document. 106 | b. Affirmer offers the Work as-is and makes no representations or 107 | warranties of any kind concerning the Work, express, implied, 108 | statutory or otherwise, including without limitation warranties of 109 | title, merchantability, fitness for a particular purpose, non 110 | infringement, or the absence of latent or other defects, accuracy, or 111 | the present or absence of errors, whether or not discoverable, all to 112 | the greatest extent permissible under applicable law. 113 | c. Affirmer disclaims responsibility for clearing rights of other persons 114 | that may apply to the Work or any use thereof, including without 115 | limitation any person's Copyright and Related Rights in the Work. 116 | Further, Affirmer disclaims responsibility for obtaining any necessary 117 | consents, permissions or other rights required for any use of the 118 | Work. 119 | d. Affirmer understands and acknowledges that Creative Commons is not a 120 | party to this document and has no duty or obligation with respect to 121 | this CC0 or use of the Work. 122 | -------------------------------------------------------------------------------- /bot.py: -------------------------------------------------------------------------------- 1 | import os 2 | import streamlit as st 3 | from streamlit.logger import get_logger 4 | # import tkinter as tk 5 | # from tkinter import filedialog 6 | from langchain.callbacks.base import BaseCallbackHandler 7 | from dotenv import load_dotenv 8 | from chains import ( 9 | load_llm, 10 | configure_llm_only_chain, 11 | get_qa_rag_chain 12 | ) 13 | from langchain.vectorstores.neo4j_vector import Neo4jVector 14 | from langchain.text_splitter import Language 15 | from agent import get_agent_executor 16 | from db import process_documents 17 | 18 | # set page title 19 | st.set_page_config( 20 | page_title="Code Explorer", 21 | page_icon="👨‍💻", 22 | layout="centered", 23 | initial_sidebar_state="expanded", 24 | menu_items={ 25 | "About": "GitHub: https://github.com/tobyloki/CodeExplorer" 26 | } 27 | ) 28 | 29 | load_dotenv(".env") 30 | 31 | url = os.getenv("NEO4J_URI") 32 | username = os.getenv("NEO4J_USERNAME") 33 | password = os.getenv("NEO4J_PASSWORD") 34 | ollama_base_url = os.getenv("OLLAMA_BASE_URL") 35 | embedding_model_name = os.getenv("EMBEDDING_MODEL") 36 | llm_name = os.getenv("LLM") 37 | # Remapping for Langchain Neo4j integration 38 | os.environ["NEO4J_URL"] = url 39 | 40 | logger = get_logger(__name__) 41 | 42 | @st.cache_resource 43 | def initLLM(): 44 | # create llm 45 | llm = load_llm(llm_name, logger=logger, config={"ollama_base_url": ollama_base_url}) 46 | 47 | return llm 48 | 49 | llm = initLLM() 50 | 51 | @st.cache_resource 52 | def get_llm_chain(): 53 | chain = configure_llm_only_chain(llm) 54 | return chain 55 | 56 | @st.cache_resource 57 | def process_directory(language, directory, count) -> (str, Neo4jVector): 58 | error, vectorstore = process_documents(language, directory) 59 | return (error, vectorstore) 60 | 61 | @st.cache_resource 62 | def get_qa_chain(_vectorstore, count): 63 | qa = get_qa_rag_chain(_vectorstore, llm) 64 | return qa 65 | 66 | @st.cache_resource 67 | def get_agent(_qa, count): 68 | qa = get_agent_executor(_qa, llm) 69 | return qa 70 | 71 | class StreamHandler(BaseCallbackHandler): 72 | def __init__(self, container, initial_text=""): 73 | self.container = container 74 | self.text = initial_text 75 | 76 | def on_llm_new_token(self, token: str, **kwargs) -> None: 77 | # if token.endswith('?'): 78 | # token += '\n\n\n' 79 | # token = token.replace('"', '') 80 | self.text += token 81 | self.container.markdown(self.text) 82 | 83 | def main(): 84 | qa = None 85 | agent = None 86 | llm_chain = get_llm_chain() 87 | 88 | if "language" not in st.session_state: 89 | st.session_state[f"language"] = None 90 | if "directory" not in st.session_state: 91 | st.session_state[f"directory"] = None 92 | if "detailedMode" not in st.session_state: 93 | st.session_state[f"detailedMode"] = True 94 | if "vectorstoreCount" not in st.session_state: # only incremented to reset cache for processDocuments() 95 | st.session_state[f"vectorstoreCount"] = 0 96 | if "qaCount" not in st.session_state: # only incremented to reset cache for get_qa_rag_chain() 97 | st.session_state[f"qaCount"] = 0 98 | if "user_input" not in st.session_state: 99 | st.session_state[f"user_input"] = [] 100 | if "generated" not in st.session_state: 101 | st.session_state[f"generated"] = [] 102 | 103 | # # Set up tkinter 104 | # root = tk.Tk() 105 | # root.withdraw() 106 | 107 | # # Make folder picker dialog appear on top of other windows 108 | # root.wm_attributes('-topmost', 1) 109 | 110 | # sidebar 111 | with st.sidebar: 112 | # Convert enum values to a list of strings 113 | languages_list = [lang.value for lang in Language] 114 | default_index = languages_list.index(Language.PYTHON) 115 | languageSelected = st.selectbox( 116 | 'Select language', 117 | languages_list, 118 | index=default_index 119 | ) 120 | 121 | # show folder picker dialog 122 | # st.title('Select Folder') 123 | # folderClicked = st.button('Folder Picker') 124 | 125 | currentPath = os.getcwd() 126 | directory = st.text_input('Enter folder path', currentPath) 127 | directory = directory.strip() 128 | 129 | processBtnClicked = st.button('Process files') 130 | if processBtnClicked: 131 | if not os.path.exists(directory): 132 | st.error("Path doesn't exist!") 133 | else: 134 | # directory = filedialog.askdirectory(master=root) 135 | if isinstance(directory, str) and directory: 136 | st.session_state[f"language"] = languageSelected 137 | st.session_state[f"directory"] = directory 138 | st.session_state[f"vectorstoreCount"] += 1 139 | st.session_state[f"qaCount"] += 1 140 | st.session_state[f"user_input"] = [] 141 | st.session_state[f"generated"] = [] 142 | 143 | # show folder selected 144 | if st.session_state[f"directory"]: 145 | st.code(st.session_state[f"directory"]) 146 | 147 | error, vectorstore = process_directory(st.session_state[f"language"], st.session_state[f"directory"], st.session_state[f"vectorstoreCount"]) 148 | 149 | if error: 150 | st.error(error) 151 | elif vectorstore: 152 | qa = get_qa_chain(vectorstore, st.session_state[f"qaCount"]) 153 | agent = get_agent(qa, st.session_state[f"qaCount"]) 154 | 155 | # show clear chat history button 156 | clearMemoryClicked = st.button("🧹 Reset chat history") 157 | if clearMemoryClicked: 158 | st.session_state[f"qaCount"] += 1 159 | st.session_state[f"user_input"] = [] 160 | st.session_state[f"generated"] = [] 161 | 162 | qa = get_qa_rag_chain(vectorstore, st.session_state[f"qaCount"]) 163 | agent = get_agent(qa, st.session_state[f"qaCount"]) 164 | 165 | # show toggle to switch between qa and agent mode 166 | detailedMode = st.toggle('Detailed mode', value=True) 167 | st.session_state[f"detailedMode"] = detailedMode 168 | 169 | # load previous chat history 170 | if st.session_state[f"generated"]: 171 | size = len(st.session_state[f"generated"]) 172 | # Display all exchanges 173 | for i in range(0, size): 174 | with st.chat_message("user"): 175 | st.write(st.session_state[f"user_input"][i]) 176 | with st.chat_message("assistant"): 177 | st.write(st.session_state[f"generated"][i]) 178 | 179 | # user chat 180 | user_input = st.chat_input("What coding issue can I help you resolve today?") 181 | if user_input: 182 | with st.chat_message("user"): 183 | st.write(user_input) 184 | st.session_state[f"user_input"].append(user_input) 185 | with st.chat_message("assistant"): 186 | with st.spinner("Generating..."): 187 | stream_handler = StreamHandler(st.empty()) 188 | if qa: 189 | if st.session_state[f"detailedMode"]: 190 | print("Using QA") 191 | result = qa( 192 | {"question": user_input}, 193 | callbacks=[stream_handler] 194 | ) 195 | answer = result["answer"] 196 | else: 197 | print("Using Agent") 198 | result = agent( 199 | {"input": user_input}, 200 | callbacks=[stream_handler] 201 | ) 202 | answer = result["output"] 203 | 204 | # print("result:", result) 205 | else: 206 | print("Using LLM only") 207 | answer = llm_chain( 208 | {"question": user_input}, 209 | callbacks=[stream_handler] 210 | ) 211 | 212 | # answer = answer.replace('"', '') 213 | st.session_state[f"generated"].append(answer) 214 | 215 | 216 | if __name__ == "__main__": 217 | main() 218 | 219 | 220 | 221 | --------------------------------------------------------------------------------