├── .env_example ├── .gitignore ├── LICENSE ├── Model ├── __init__.py ├── invokemodel.py └── provider.py ├── README.md ├── agent ├── __init__.py └── web_agent.py ├── config ├── __init__.py ├── log.py ├── settings.py └── user_settings.py ├── configure ├── __init__.py ├── config_llm.py ├── llama.py └── vision.py ├── extras ├── __init__.py ├── pawelzmarlak-2025-01-22T16_41_02.675Z.png ├── pawelzmarlak-2025-02-13T06_51_12.773Z.png ├── safejsonload.py └── viewimage.py ├── main.py ├── memory ├── __init__.py └── research_mem.py ├── requirements.txt ├── source_reliable ├── __init__.py └── source_reliability_class.py ├── test ├── __init__.py ├── ollama.py ├── selenium.py └── test_model.py └── tools ├── __init__.py ├── capture_ss.py ├── create_vecstore.py ├── extract_urls.py ├── fetch_webpage.py ├── host_tracker.py ├── size_limit.py ├── split_doc.py ├── topic_into_sub.py └── vision_query.py /.env_example: -------------------------------------------------------------------------------- 1 | GROQ_API_KEY = "API_KEY" 2 | BRAVE_API_KEY = "API_KEY" -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- 1 | agent_memory.json 2 | HOSTS.txt 3 | 4 | # Byte-compiled / optimized / DLL files 5 | __pycache__/ 6 | *.py[cod] 7 | *$py.class 8 | 9 | # C extensions 10 | *.so 11 | 12 | # Distribution / packaging 13 | .Python 14 | build/ 15 | develop-eggs/ 16 | dist/ 17 | downloads/ 18 | eggs/ 19 | .eggs/ 20 | lib/ 21 | lib64/ 22 | parts/ 23 | sdist/ 24 | var/ 25 | wheels/ 26 | share/python-wheels/ 27 | *.egg-info/ 28 | .installed.cfg 29 | *.egg 30 | MANIFEST 31 | 32 | # PyInstaller 33 | # Usually these files are written by a python script from a template 34 | # before PyInstaller builds the exe, so as to inject date/other infos into it. 35 | *.manifest 36 | *.spec 37 | 38 | # Installer logs 39 | pip-log.txt 40 | pip-delete-this-directory.txt 41 | 42 | # Unit test / coverage reports 43 | htmlcov/ 44 | .tox/ 45 | .nox/ 46 | .coverage 47 | .coverage.* 48 | .cache 49 | nosetests.xml 50 | coverage.xml 51 | *.cover 52 | *.py,cover 53 | .hypothesis/ 54 | .pytest_cache/ 55 | cover/ 56 | 57 | # Translations 58 | *.mo 59 | *.pot 60 | 61 | # Django stuff: 62 | *.log 63 | local_settings.py 64 | db.sqlite3 65 | db.sqlite3-journal 66 | 67 | # Flask stuff: 68 | instance/ 69 | .webassets-cache 70 | 71 | # Scrapy stuff: 72 | .scrapy 73 | 74 | # Sphinx documentation 75 | docs/_build/ 76 | 77 | # PyBuilder 78 | .pybuilder/ 79 | target/ 80 | 81 | # Jupyter Notebook 82 | .ipynb_checkpoints 83 | 84 | # IPython 85 | profile_default/ 86 | ipython_config.py 87 | 88 | # pyenv 89 | # For a library or package, you might want to ignore these files since the code is 90 | # intended to run in multiple environments; otherwise, check them in: 91 | # .python-version 92 | 93 | # pipenv 94 | # According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control. 95 | # However, in case of collaboration, if having platform-specific dependencies or dependencies 96 | # having no cross-platform support, pipenv may install dependencies that don't work, or not 97 | # install all needed dependencies. 98 | #Pipfile.lock 99 | 100 | # UV 101 | # Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control. 102 | # This is especially recommended for binary packages to ensure reproducibility, and is more 103 | # commonly ignored for libraries. 104 | #uv.lock 105 | 106 | # poetry 107 | # Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control. 108 | # This is especially recommended for binary packages to ensure reproducibility, and is more 109 | # commonly ignored for libraries. 110 | # https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control 111 | #poetry.lock 112 | 113 | # pdm 114 | # Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control. 115 | #pdm.lock 116 | # pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it 117 | # in version control. 118 | # https://pdm.fming.dev/latest/usage/project/#working-with-version-control 119 | .pdm.toml 120 | .pdm-python 121 | .pdm-build/ 122 | 123 | # PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm 124 | __pypackages__/ 125 | 126 | # Celery stuff 127 | celerybeat-schedule 128 | celerybeat.pid 129 | 130 | # SageMath parsed files 131 | *.sage.py 132 | 133 | # Environments 134 | .env 135 | .venv 136 | env/ 137 | venv/ 138 | ENV/ 139 | env.bak/ 140 | venv.bak/ 141 | 142 | # Spyder project settings 143 | .spyderproject 144 | .spyproject 145 | 146 | # Rope project settings 147 | .ropeproject 148 | 149 | # mkdocs documentation 150 | /site 151 | 152 | # mypy 153 | .mypy_cache/ 154 | .dmypy.json 155 | dmypy.json 156 | 157 | # Pyre type checker 158 | .pyre/ 159 | 160 | # pytype static type analyzer 161 | .pytype/ 162 | 163 | # Cython debug symbols 164 | cython_debug/ 165 | 166 | # PyCharm 167 | # JetBrains specific template is maintained in a separate JetBrains.gitignore that can 168 | # be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore 169 | # and can be added to the global gitignore or merged into this file. For a more nuclear 170 | # option (not recommended) you can uncomment the following to ignore the entire idea folder. 171 | #.idea/ 172 | 173 | # PyPI configuration file 174 | .pypirc 175 | -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | GNU GENERAL PUBLIC LICENSE 2 | Version 3, 29 June 2007 3 | 4 | Copyright (C) 2007 Free Software Foundation, Inc. 5 | Everyone is permitted to copy and distribute verbatim copies 6 | of this license document, but changing it is not allowed. 7 | 8 | Preamble 9 | 10 | The GNU General Public License is a free, copyleft license for 11 | software and other kinds of works. 12 | 13 | The licenses for most software and other practical works are designed 14 | to take away your freedom to share and change the works. By contrast, 15 | the GNU General Public License is intended to guarantee your freedom to 16 | share and change all versions of a program--to make sure it remains free 17 | software for all its users. We, the Free Software Foundation, use the 18 | GNU General Public License for most of our software; it applies also to 19 | any other work released this way by its authors. You can apply it to 20 | your programs, too. 21 | 22 | When we speak of free software, we are referring to freedom, not 23 | price. Our General Public Licenses are designed to make sure that you 24 | have the freedom to distribute copies of free software (and charge for 25 | them if you wish), that you receive source code or can get it if you 26 | want it, that you can change the software or use pieces of it in new 27 | free programs, and that you know you can do these things. 28 | 29 | To protect your rights, we need to prevent others from denying you 30 | these rights or asking you to surrender the rights. Therefore, you have 31 | certain responsibilities if you distribute copies of the software, or if 32 | you modify it: responsibilities to respect the freedom of others. 33 | 34 | For example, if you distribute copies of such a program, whether 35 | gratis or for a fee, you must pass on to the recipients the same 36 | freedoms that you received. You must make sure that they, too, receive 37 | or can get the source code. And you must show them these terms so they 38 | know their rights. 39 | 40 | Developers that use the GNU GPL protect your rights with two steps: 41 | (1) assert copyright on the software, and (2) offer you this License 42 | giving you legal permission to copy, distribute and/or modify it. 43 | 44 | For the developers' and authors' protection, the GPL clearly explains 45 | that there is no warranty for this free software. For both users' and 46 | authors' sake, the GPL requires that modified versions be marked as 47 | changed, so that their problems will not be attributed erroneously to 48 | authors of previous versions. 49 | 50 | Some devices are designed to deny users access to install or run 51 | modified versions of the software inside them, although the manufacturer 52 | can do so. This is fundamentally incompatible with the aim of 53 | protecting users' freedom to change the software. The systematic 54 | pattern of such abuse occurs in the area of products for individuals to 55 | use, which is precisely where it is most unacceptable. Therefore, we 56 | have designed this version of the GPL to prohibit the practice for those 57 | products. If such problems arise substantially in other domains, we 58 | stand ready to extend this provision to those domains in future versions 59 | of the GPL, as needed to protect the freedom of users. 60 | 61 | Finally, every program is threatened constantly by software patents. 62 | States should not allow patents to restrict development and use of 63 | software on general-purpose computers, but in those that do, we wish to 64 | avoid the special danger that patents applied to a free program could 65 | make it effectively proprietary. To prevent this, the GPL assures that 66 | patents cannot be used to render the program non-free. 67 | 68 | The precise terms and conditions for copying, distribution and 69 | modification follow. 70 | 71 | TERMS AND CONDITIONS 72 | 73 | 0. Definitions. 74 | 75 | "This License" refers to version 3 of the GNU General Public License. 76 | 77 | "Copyright" also means copyright-like laws that apply to other kinds of 78 | works, such as semiconductor masks. 79 | 80 | "The Program" refers to any copyrightable work licensed under this 81 | License. Each licensee is addressed as "you". "Licensees" and 82 | "recipients" may be individuals or organizations. 83 | 84 | To "modify" a work means to copy from or adapt all or part of the work 85 | in a fashion requiring copyright permission, other than the making of an 86 | exact copy. The resulting work is called a "modified version" of the 87 | earlier work or a work "based on" the earlier work. 88 | 89 | A "covered work" means either the unmodified Program or a work based 90 | on the Program. 91 | 92 | To "propagate" a work means to do anything with it that, without 93 | permission, would make you directly or secondarily liable for 94 | infringement under applicable copyright law, except executing it on a 95 | computer or modifying a private copy. Propagation includes copying, 96 | distribution (with or without modification), making available to the 97 | public, and in some countries other activities as well. 98 | 99 | To "convey" a work means any kind of propagation that enables other 100 | parties to make or receive copies. Mere interaction with a user through 101 | a computer network, with no transfer of a copy, is not conveying. 102 | 103 | An interactive user interface displays "Appropriate Legal Notices" 104 | to the extent that it includes a convenient and prominently visible 105 | feature that (1) displays an appropriate copyright notice, and (2) 106 | tells the user that there is no warranty for the work (except to the 107 | extent that warranties are provided), that licensees may convey the 108 | work under this License, and how to view a copy of this License. If 109 | the interface presents a list of user commands or options, such as a 110 | menu, a prominent item in the list meets this criterion. 111 | 112 | 1. Source Code. 113 | 114 | The "source code" for a work means the preferred form of the work 115 | for making modifications to it. "Object code" means any non-source 116 | form of a work. 117 | 118 | A "Standard Interface" means an interface that either is an official 119 | standard defined by a recognized standards body, or, in the case of 120 | interfaces specified for a particular programming language, one that 121 | is widely used among developers working in that language. 122 | 123 | The "System Libraries" of an executable work include anything, other 124 | than the work as a whole, that (a) is included in the normal form of 125 | packaging a Major Component, but which is not part of that Major 126 | Component, and (b) serves only to enable use of the work with that 127 | Major Component, or to implement a Standard Interface for which an 128 | implementation is available to the public in source code form. A 129 | "Major Component", in this context, means a major essential component 130 | (kernel, window system, and so on) of the specific operating system 131 | (if any) on which the executable work runs, or a compiler used to 132 | produce the work, or an object code interpreter used to run it. 133 | 134 | The "Corresponding Source" for a work in object code form means all 135 | the source code needed to generate, install, and (for an executable 136 | work) run the object code and to modify the work, including scripts to 137 | control those activities. However, it does not include the work's 138 | System Libraries, or general-purpose tools or generally available free 139 | programs which are used unmodified in performing those activities but 140 | which are not part of the work. For example, Corresponding Source 141 | includes interface definition files associated with source files for 142 | the work, and the source code for shared libraries and dynamically 143 | linked subprograms that the work is specifically designed to require, 144 | such as by intimate data communication or control flow between those 145 | subprograms and other parts of the work. 146 | 147 | The Corresponding Source need not include anything that users 148 | can regenerate automatically from other parts of the Corresponding 149 | Source. 150 | 151 | The Corresponding Source for a work in source code form is that 152 | same work. 153 | 154 | 2. Basic Permissions. 155 | 156 | All rights granted under this License are granted for the term of 157 | copyright on the Program, and are irrevocable provided the stated 158 | conditions are met. This License explicitly affirms your unlimited 159 | permission to run the unmodified Program. The output from running a 160 | covered work is covered by this License only if the output, given its 161 | content, constitutes a covered work. This License acknowledges your 162 | rights of fair use or other equivalent, as provided by copyright law. 163 | 164 | You may make, run and propagate covered works that you do not 165 | convey, without conditions so long as your license otherwise remains 166 | in force. You may convey covered works to others for the sole purpose 167 | of having them make modifications exclusively for you, or provide you 168 | with facilities for running those works, provided that you comply with 169 | the terms of this License in conveying all material for which you do 170 | not control copyright. Those thus making or running the covered works 171 | for you must do so exclusively on your behalf, under your direction 172 | and control, on terms that prohibit them from making any copies of 173 | your copyrighted material outside their relationship with you. 174 | 175 | Conveying under any other circumstances is permitted solely under 176 | the conditions stated below. Sublicensing is not allowed; section 10 177 | makes it unnecessary. 178 | 179 | 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 180 | 181 | No covered work shall be deemed part of an effective technological 182 | measure under any applicable law fulfilling obligations under article 183 | 11 of the WIPO copyright treaty adopted on 20 December 1996, or 184 | similar laws prohibiting or restricting circumvention of such 185 | measures. 186 | 187 | When you convey a covered work, you waive any legal power to forbid 188 | circumvention of technological measures to the extent such circumvention 189 | is effected by exercising rights under this License with respect to 190 | the covered work, and you disclaim any intention to limit operation or 191 | modification of the work as a means of enforcing, against the work's 192 | users, your or third parties' legal rights to forbid circumvention of 193 | technological measures. 194 | 195 | 4. Conveying Verbatim Copies. 196 | 197 | You may convey verbatim copies of the Program's source code as you 198 | receive it, in any medium, provided that you conspicuously and 199 | appropriately publish on each copy an appropriate copyright notice; 200 | keep intact all notices stating that this License and any 201 | non-permissive terms added in accord with section 7 apply to the code; 202 | keep intact all notices of the absence of any warranty; and give all 203 | recipients a copy of this License along with the Program. 204 | 205 | You may charge any price or no price for each copy that you convey, 206 | and you may offer support or warranty protection for a fee. 207 | 208 | 5. Conveying Modified Source Versions. 209 | 210 | You may convey a work based on the Program, or the modifications to 211 | produce it from the Program, in the form of source code under the 212 | terms of section 4, provided that you also meet all of these conditions: 213 | 214 | a) The work must carry prominent notices stating that you modified 215 | it, and giving a relevant date. 216 | 217 | b) The work must carry prominent notices stating that it is 218 | released under this License and any conditions added under section 219 | 7. This requirement modifies the requirement in section 4 to 220 | "keep intact all notices". 221 | 222 | c) You must license the entire work, as a whole, under this 223 | License to anyone who comes into possession of a copy. This 224 | License will therefore apply, along with any applicable section 7 225 | additional terms, to the whole of the work, and all its parts, 226 | regardless of how they are packaged. This License gives no 227 | permission to license the work in any other way, but it does not 228 | invalidate such permission if you have separately received it. 229 | 230 | d) If the work has interactive user interfaces, each must display 231 | Appropriate Legal Notices; however, if the Program has interactive 232 | interfaces that do not display Appropriate Legal Notices, your 233 | work need not make them do so. 234 | 235 | A compilation of a covered work with other separate and independent 236 | works, which are not by their nature extensions of the covered work, 237 | and which are not combined with it such as to form a larger program, 238 | in or on a volume of a storage or distribution medium, is called an 239 | "aggregate" if the compilation and its resulting copyright are not 240 | used to limit the access or legal rights of the compilation's users 241 | beyond what the individual works permit. Inclusion of a covered work 242 | in an aggregate does not cause this License to apply to the other 243 | parts of the aggregate. 244 | 245 | 6. Conveying Non-Source Forms. 246 | 247 | You may convey a covered work in object code form under the terms 248 | of sections 4 and 5, provided that you also convey the 249 | machine-readable Corresponding Source under the terms of this License, 250 | in one of these ways: 251 | 252 | a) Convey the object code in, or embodied in, a physical product 253 | (including a physical distribution medium), accompanied by the 254 | Corresponding Source fixed on a durable physical medium 255 | customarily used for software interchange. 256 | 257 | b) Convey the object code in, or embodied in, a physical product 258 | (including a physical distribution medium), accompanied by a 259 | written offer, valid for at least three years and valid for as 260 | long as you offer spare parts or customer support for that product 261 | model, to give anyone who possesses the object code either (1) a 262 | copy of the Corresponding Source for all the software in the 263 | product that is covered by this License, on a durable physical 264 | medium customarily used for software interchange, for a price no 265 | more than your reasonable cost of physically performing this 266 | conveying of source, or (2) access to copy the 267 | Corresponding Source from a network server at no charge. 268 | 269 | c) Convey individual copies of the object code with a copy of the 270 | written offer to provide the Corresponding Source. This 271 | alternative is allowed only occasionally and noncommercially, and 272 | only if you received the object code with such an offer, in accord 273 | with subsection 6b. 274 | 275 | d) Convey the object code by offering access from a designated 276 | place (gratis or for a charge), and offer equivalent access to the 277 | Corresponding Source in the same way through the same place at no 278 | further charge. You need not require recipients to copy the 279 | Corresponding Source along with the object code. If the place to 280 | copy the object code is a network server, the Corresponding Source 281 | may be on a different server (operated by you or a third party) 282 | that supports equivalent copying facilities, provided you maintain 283 | clear directions next to the object code saying where to find the 284 | Corresponding Source. Regardless of what server hosts the 285 | Corresponding Source, you remain obligated to ensure that it is 286 | available for as long as needed to satisfy these requirements. 287 | 288 | e) Convey the object code using peer-to-peer transmission, provided 289 | you inform other peers where the object code and Corresponding 290 | Source of the work are being offered to the general public at no 291 | charge under subsection 6d. 292 | 293 | A separable portion of the object code, whose source code is excluded 294 | from the Corresponding Source as a System Library, need not be 295 | included in conveying the object code work. 296 | 297 | A "User Product" is either (1) a "consumer product", which means any 298 | tangible personal property which is normally used for personal, family, 299 | or household purposes, or (2) anything designed or sold for incorporation 300 | into a dwelling. In determining whether a product is a consumer product, 301 | doubtful cases shall be resolved in favor of coverage. For a particular 302 | product received by a particular user, "normally used" refers to a 303 | typical or common use of that class of product, regardless of the status 304 | of the particular user or of the way in which the particular user 305 | actually uses, or expects or is expected to use, the product. A product 306 | is a consumer product regardless of whether the product has substantial 307 | commercial, industrial or non-consumer uses, unless such uses represent 308 | the only significant mode of use of the product. 309 | 310 | "Installation Information" for a User Product means any methods, 311 | procedures, authorization keys, or other information required to install 312 | and execute modified versions of a covered work in that User Product from 313 | a modified version of its Corresponding Source. The information must 314 | suffice to ensure that the continued functioning of the modified object 315 | code is in no case prevented or interfered with solely because 316 | modification has been made. 317 | 318 | If you convey an object code work under this section in, or with, or 319 | specifically for use in, a User Product, and the conveying occurs as 320 | part of a transaction in which the right of possession and use of the 321 | User Product is transferred to the recipient in perpetuity or for a 322 | fixed term (regardless of how the transaction is characterized), the 323 | Corresponding Source conveyed under this section must be accompanied 324 | by the Installation Information. But this requirement does not apply 325 | if neither you nor any third party retains the ability to install 326 | modified object code on the User Product (for example, the work has 327 | been installed in ROM). 328 | 329 | The requirement to provide Installation Information does not include a 330 | requirement to continue to provide support service, warranty, or updates 331 | for a work that has been modified or installed by the recipient, or for 332 | the User Product in which it has been modified or installed. Access to a 333 | network may be denied when the modification itself materially and 334 | adversely affects the operation of the network or violates the rules and 335 | protocols for communication across the network. 336 | 337 | Corresponding Source conveyed, and Installation Information provided, 338 | in accord with this section must be in a format that is publicly 339 | documented (and with an implementation available to the public in 340 | source code form), and must require no special password or key for 341 | unpacking, reading or copying. 342 | 343 | 7. Additional Terms. 344 | 345 | "Additional permissions" are terms that supplement the terms of this 346 | License by making exceptions from one or more of its conditions. 347 | Additional permissions that are applicable to the entire Program shall 348 | be treated as though they were included in this License, to the extent 349 | that they are valid under applicable law. If additional permissions 350 | apply only to part of the Program, that part may be used separately 351 | under those permissions, but the entire Program remains governed by 352 | this License without regard to the additional permissions. 353 | 354 | When you convey a copy of a covered work, you may at your option 355 | remove any additional permissions from that copy, or from any part of 356 | it. (Additional permissions may be written to require their own 357 | removal in certain cases when you modify the work.) You may place 358 | additional permissions on material, added by you to a covered work, 359 | for which you have or can give appropriate copyright permission. 360 | 361 | Notwithstanding any other provision of this License, for material you 362 | add to a covered work, you may (if authorized by the copyright holders of 363 | that material) supplement the terms of this License with terms: 364 | 365 | a) Disclaiming warranty or limiting liability differently from the 366 | terms of sections 15 and 16 of this License; or 367 | 368 | b) Requiring preservation of specified reasonable legal notices or 369 | author attributions in that material or in the Appropriate Legal 370 | Notices displayed by works containing it; or 371 | 372 | c) Prohibiting misrepresentation of the origin of that material, or 373 | requiring that modified versions of such material be marked in 374 | reasonable ways as different from the original version; or 375 | 376 | d) Limiting the use for publicity purposes of names of licensors or 377 | authors of the material; or 378 | 379 | e) Declining to grant rights under trademark law for use of some 380 | trade names, trademarks, or service marks; or 381 | 382 | f) Requiring indemnification of licensors and authors of that 383 | material by anyone who conveys the material (or modified versions of 384 | it) with contractual assumptions of liability to the recipient, for 385 | any liability that these contractual assumptions directly impose on 386 | those licensors and authors. 387 | 388 | All other non-permissive additional terms are considered "further 389 | restrictions" within the meaning of section 10. If the Program as you 390 | received it, or any part of it, contains a notice stating that it is 391 | governed by this License along with a term that is a further 392 | restriction, you may remove that term. If a license document contains 393 | a further restriction but permits relicensing or conveying under this 394 | License, you may add to a covered work material governed by the terms 395 | of that license document, provided that the further restriction does 396 | not survive such relicensing or conveying. 397 | 398 | If you add terms to a covered work in accord with this section, you 399 | must place, in the relevant source files, a statement of the 400 | additional terms that apply to those files, or a notice indicating 401 | where to find the applicable terms. 402 | 403 | Additional terms, permissive or non-permissive, may be stated in the 404 | form of a separately written license, or stated as exceptions; 405 | the above requirements apply either way. 406 | 407 | 8. Termination. 408 | 409 | You may not propagate or modify a covered work except as expressly 410 | provided under this License. Any attempt otherwise to propagate or 411 | modify it is void, and will automatically terminate your rights under 412 | this License (including any patent licenses granted under the third 413 | paragraph of section 11). 414 | 415 | However, if you cease all violation of this License, then your 416 | license from a particular copyright holder is reinstated (a) 417 | provisionally, unless and until the copyright holder explicitly and 418 | finally terminates your license, and (b) permanently, if the copyright 419 | holder fails to notify you of the violation by some reasonable means 420 | prior to 60 days after the cessation. 421 | 422 | Moreover, your license from a particular copyright holder is 423 | reinstated permanently if the copyright holder notifies you of the 424 | violation by some reasonable means, this is the first time you have 425 | received notice of violation of this License (for any work) from that 426 | copyright holder, and you cure the violation prior to 30 days after 427 | your receipt of the notice. 428 | 429 | Termination of your rights under this section does not terminate the 430 | licenses of parties who have received copies or rights from you under 431 | this License. If your rights have been terminated and not permanently 432 | reinstated, you do not qualify to receive new licenses for the same 433 | material under section 10. 434 | 435 | 9. Acceptance Not Required for Having Copies. 436 | 437 | You are not required to accept this License in order to receive or 438 | run a copy of the Program. Ancillary propagation of a covered work 439 | occurring solely as a consequence of using peer-to-peer transmission 440 | to receive a copy likewise does not require acceptance. However, 441 | nothing other than this License grants you permission to propagate or 442 | modify any covered work. These actions infringe copyright if you do 443 | not accept this License. Therefore, by modifying or propagating a 444 | covered work, you indicate your acceptance of this License to do so. 445 | 446 | 10. Automatic Licensing of Downstream Recipients. 447 | 448 | Each time you convey a covered work, the recipient automatically 449 | receives a license from the original licensors, to run, modify and 450 | propagate that work, subject to this License. You are not responsible 451 | for enforcing compliance by third parties with this License. 452 | 453 | An "entity transaction" is a transaction transferring control of an 454 | organization, or substantially all assets of one, or subdividing an 455 | organization, or merging organizations. If propagation of a covered 456 | work results from an entity transaction, each party to that 457 | transaction who receives a copy of the work also receives whatever 458 | licenses to the work the party's predecessor in interest had or could 459 | give under the previous paragraph, plus a right to possession of the 460 | Corresponding Source of the work from the predecessor in interest, if 461 | the predecessor has it or can get it with reasonable efforts. 462 | 463 | You may not impose any further restrictions on the exercise of the 464 | rights granted or affirmed under this License. For example, you may 465 | not impose a license fee, royalty, or other charge for exercise of 466 | rights granted under this License, and you may not initiate litigation 467 | (including a cross-claim or counterclaim in a lawsuit) alleging that 468 | any patent claim is infringed by making, using, selling, offering for 469 | sale, or importing the Program or any portion of it. 470 | 471 | 11. Patents. 472 | 473 | A "contributor" is a copyright holder who authorizes use under this 474 | License of the Program or a work on which the Program is based. The 475 | work thus licensed is called the contributor's "contributor version". 476 | 477 | A contributor's "essential patent claims" are all patent claims 478 | owned or controlled by the contributor, whether already acquired or 479 | hereafter acquired, that would be infringed by some manner, permitted 480 | by this License, of making, using, or selling its contributor version, 481 | but do not include claims that would be infringed only as a 482 | consequence of further modification of the contributor version. For 483 | purposes of this definition, "control" includes the right to grant 484 | patent sublicenses in a manner consistent with the requirements of 485 | this License. 486 | 487 | Each contributor grants you a non-exclusive, worldwide, royalty-free 488 | patent license under the contributor's essential patent claims, to 489 | make, use, sell, offer for sale, import and otherwise run, modify and 490 | propagate the contents of its contributor version. 491 | 492 | In the following three paragraphs, a "patent license" is any express 493 | agreement or commitment, however denominated, not to enforce a patent 494 | (such as an express permission to practice a patent or covenant not to 495 | sue for patent infringement). To "grant" such a patent license to a 496 | party means to make such an agreement or commitment not to enforce a 497 | patent against the party. 498 | 499 | If you convey a covered work, knowingly relying on a patent license, 500 | and the Corresponding Source of the work is not available for anyone 501 | to copy, free of charge and under the terms of this License, through a 502 | publicly available network server or other readily accessible means, 503 | then you must either (1) cause the Corresponding Source to be so 504 | available, or (2) arrange to deprive yourself of the benefit of the 505 | patent license for this particular work, or (3) arrange, in a manner 506 | consistent with the requirements of this License, to extend the patent 507 | license to downstream recipients. "Knowingly relying" means you have 508 | actual knowledge that, but for the patent license, your conveying the 509 | covered work in a country, or your recipient's use of the covered work 510 | in a country, would infringe one or more identifiable patents in that 511 | country that you have reason to believe are valid. 512 | 513 | If, pursuant to or in connection with a single transaction or 514 | arrangement, you convey, or propagate by procuring conveyance of, a 515 | covered work, and grant a patent license to some of the parties 516 | receiving the covered work authorizing them to use, propagate, modify 517 | or convey a specific copy of the covered work, then the patent license 518 | you grant is automatically extended to all recipients of the covered 519 | work and works based on it. 520 | 521 | A patent license is "discriminatory" if it does not include within 522 | the scope of its coverage, prohibits the exercise of, or is 523 | conditioned on the non-exercise of one or more of the rights that are 524 | specifically granted under this License. You may not convey a covered 525 | work if you are a party to an arrangement with a third party that is 526 | in the business of distributing software, under which you make payment 527 | to the third party based on the extent of your activity of conveying 528 | the work, and under which the third party grants, to any of the 529 | parties who would receive the covered work from you, a discriminatory 530 | patent license (a) in connection with copies of the covered work 531 | conveyed by you (or copies made from those copies), or (b) primarily 532 | for and in connection with specific products or compilations that 533 | contain the covered work, unless you entered into that arrangement, 534 | or that patent license was granted, prior to 28 March 2007. 535 | 536 | Nothing in this License shall be construed as excluding or limiting 537 | any implied license or other defenses to infringement that may 538 | otherwise be available to you under applicable patent law. 539 | 540 | 12. No Surrender of Others' Freedom. 541 | 542 | If conditions are imposed on you (whether by court order, agreement or 543 | otherwise) that contradict the conditions of this License, they do not 544 | excuse you from the conditions of this License. If you cannot convey a 545 | covered work so as to satisfy simultaneously your obligations under this 546 | License and any other pertinent obligations, then as a consequence you may 547 | not convey it at all. For example, if you agree to terms that obligate you 548 | to collect a royalty for further conveying from those to whom you convey 549 | the Program, the only way you could satisfy both those terms and this 550 | License would be to refrain entirely from conveying the Program. 551 | 552 | 13. Use with the GNU Affero General Public License. 553 | 554 | Notwithstanding any other provision of this License, you have 555 | permission to link or combine any covered work with a work licensed 556 | under version 3 of the GNU Affero General Public License into a single 557 | combined work, and to convey the resulting work. The terms of this 558 | License will continue to apply to the part which is the covered work, 559 | but the special requirements of the GNU Affero General Public License, 560 | section 13, concerning interaction through a network will apply to the 561 | combination as such. 562 | 563 | 14. Revised Versions of this License. 564 | 565 | The Free Software Foundation may publish revised and/or new versions of 566 | the GNU General Public License from time to time. Such new versions will 567 | be similar in spirit to the present version, but may differ in detail to 568 | address new problems or concerns. 569 | 570 | Each version is given a distinguishing version number. If the 571 | Program specifies that a certain numbered version of the GNU General 572 | Public License "or any later version" applies to it, you have the 573 | option of following the terms and conditions either of that numbered 574 | version or of any later version published by the Free Software 575 | Foundation. If the Program does not specify a version number of the 576 | GNU General Public License, you may choose any version ever published 577 | by the Free Software Foundation. 578 | 579 | If the Program specifies that a proxy can decide which future 580 | versions of the GNU General Public License can be used, that proxy's 581 | public statement of acceptance of a version permanently authorizes you 582 | to choose that version for the Program. 583 | 584 | Later license versions may give you additional or different 585 | permissions. However, no additional obligations are imposed on any 586 | author or copyright holder as a result of your choosing to follow a 587 | later version. 588 | 589 | 15. Disclaimer of Warranty. 590 | 591 | THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 592 | APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT 593 | HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY 594 | OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, 595 | THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 596 | PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM 597 | IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF 598 | ALL NECESSARY SERVICING, REPAIR OR CORRECTION. 599 | 600 | 16. Limitation of Liability. 601 | 602 | IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING 603 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS 604 | THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY 605 | GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE 606 | USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF 607 | DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 608 | PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), 609 | EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF 610 | SUCH DAMAGES. 611 | 612 | 17. Interpretation of Sections 15 and 16. 613 | 614 | If the disclaimer of warranty and limitation of liability provided 615 | above cannot be given local legal effect according to their terms, 616 | reviewing courts shall apply local law that most closely approximates 617 | an absolute waiver of all civil liability in connection with the 618 | Program, unless a warranty or assumption of liability accompanies a 619 | copy of the Program in return for a fee. 620 | 621 | END OF TERMS AND CONDITIONS 622 | 623 | How to Apply These Terms to Your New Programs 624 | 625 | If you develop a new program, and you want it to be of the greatest 626 | possible use to the public, the best way to achieve this is to make it 627 | free software which everyone can redistribute and change under these terms. 628 | 629 | To do so, attach the following notices to the program. It is safest 630 | to attach them to the start of each source file to most effectively 631 | state the exclusion of warranty; and each file should have at least 632 | the "copyright" line and a pointer to where the full notice is found. 633 | 634 | 635 | Copyright (C) 636 | 637 | This program is free software: you can redistribute it and/or modify 638 | it under the terms of the GNU General Public License as published by 639 | the Free Software Foundation, either version 3 of the License, or 640 | (at your option) any later version. 641 | 642 | This program is distributed in the hope that it will be useful, 643 | but WITHOUT ANY WARRANTY; without even the implied warranty of 644 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 645 | GNU General Public License for more details. 646 | 647 | You should have received a copy of the GNU General Public License 648 | along with this program. If not, see . 649 | 650 | Also add information on how to contact you by electronic and paper mail. 651 | 652 | If the program does terminal interaction, make it output a short 653 | notice like this when it starts in an interactive mode: 654 | 655 | Copyright (C) 656 | This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. 657 | This is free software, and you are welcome to redistribute it 658 | under certain conditions; type `show c' for details. 659 | 660 | The hypothetical commands `show w' and `show c' should show the appropriate 661 | parts of the General Public License. Of course, your program's commands 662 | might be different; for a GUI interface, you would use an "about box". 663 | 664 | You should also get your employer (if you work as a programmer) or school, 665 | if any, to sign a "copyright disclaimer" for the program, if necessary. 666 | For more information on this, and how to apply and follow the GNU GPL, see 667 | . 668 | 669 | The GNU General Public License does not permit incorporating your program 670 | into proprietary programs. If your program is a subroutine library, you 671 | may consider it more useful to permit linking proprietary applications with 672 | the library. If this is what you want to do, use the GNU Lesser General 673 | Public License instead of this License. But first, please read 674 | . 675 | -------------------------------------------------------------------------------- /Model/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Haseebasif7/SurfAgent/e527e1bce6234e3b4456298811550bf3a7552375/Model/__init__.py -------------------------------------------------------------------------------- /Model/invokemodel.py: -------------------------------------------------------------------------------- 1 | from langchain.schema import HumanMessage, AIMessage 2 | 3 | def invoke_model(llm, prompt: str) -> AIMessage: 4 | """Helper to invoke LLM with a single prompt.""" 5 | response = llm([HumanMessage(content=prompt)]) 6 | return response -------------------------------------------------------------------------------- /Model/provider.py: -------------------------------------------------------------------------------- 1 | class ModelProvider: 2 | OLLAMA = "ollama" 3 | GROQ = "groq" 4 | 5 | @staticmethod 6 | def get_provider_choice() -> str: 7 | return ModelProvider.GROQ -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | 2 | # 🌐 **SurfAgent: The Web Search and Analysis Agent** 3 | 4 | SurfAgent is a sophisticated **CLI-based agent** built from scratch. Powered by **Selenium** and **Brave Search**, SurfAgent extracts relevant information and even analyzes images related to your queries and **quotes** Sources of that information in the form of links as well, leveraging state-of-the-art **Llama Vision Models**. These models are compatible with **GROQ** or **OLLAMA**, offering seamless integration. 5 | 6 | --- 7 | 8 | ## ✨ **Key Features** 9 | 10 | ### 📁 Contextual Memory 11 | SurfAgent remembers and learns from past interactions, storing details in a dynamic memory system. Memory snapshots are available in a JSON format, providing insights like: 12 | ```json 13 | { 14 | "query_types": "", 15 | "last_success": "", 16 | "last_failure": "", 17 | "total_attempts": "", 18 | "successful_attempts": "", 19 | "average_response_time": "", 20 | "notes": "" 21 | } 22 | ``` 23 | 24 | ### 🚫 Intelligent Host Management 25 | SurfAgent keeps a record of problematic hosts in a `HOSTS.txt` file, ensuring those hosts are avoided in future searches. 26 | 27 | ALT TEXT 28 | 29 | ### 🖼️ Image Analysis 30 | Integrates advanced image processing capabilities by fetching relevant images based on user input or context, then passing these images to **Llama Vision Models**. These models analyze the images to extract key information, such as text, objects, patterns, or any visual data present, and process it to generate meaningful insights or responses. 31 | 32 | ALT TEXT 33 | 34 | ### 🔗 LangChain Integration 35 | Utilizes **LangChain tools** to enhance automation and analytical capabilities. 36 | 37 | --- 38 | 39 | ## 🚀 **How to Run SurfAgent** 40 | 41 | To get started with SurfAgent, follow these simple steps: 42 | 43 | ### 🛠️ **Step 1: Clone the Repository** 44 | 45 | First, clone the SurfAgent repository from GitHub to your local machine. Open your terminal or command prompt and run the following command: 46 | 47 | ```bash 48 | git clone https://github.com/Haseebasif7/SurfAgent.git 49 | ``` 50 | 51 | ### 🛠️ **Step 2: Install the Required Dependencies** 52 | 53 | Navigate to the `SurfAgent` directory: 54 | 55 | ```bash 56 | cd SurfAgent 57 | ``` 58 | 59 | Then, install the required libraries by running: 60 | 61 | ```bash 62 | pip install -r requirements.txt 63 | ``` 64 | 65 | This will install the necessary packages, including **Selenium**, **Brave Search**, **Llama Vision Models**, **LangChain**, and more. 66 | 67 | ### 🔑 **Step 3: Set Up Environment Variables** 68 | 69 | Create a `.env` file in the root directory of the SurfAgent project and add your API keys for **GROQ** and **Brave Search**: 70 | 71 | ```env 72 | GROQ_API_KEY="your_groq_api_key" 73 | BRAVE_API_KEY="your_brave_api_key" 74 | ``` 75 | 76 | Make sure to replace `"your_groq_api_key"` and `"your_brave_api_key"` with your actual API keys. 77 | 78 | ### 🖥️ **Step 4: Run the SurfAgent** 79 | 80 | Once the dependencies are installed and environment variables are set, you're ready to run SurfAgent. Execute the following command to start the agent: 81 | 82 | ```bash 83 | python main.py 84 | ``` 85 | 86 | SurfAgent will now initialize, begin processing web searches, and provide results based on your queries. Enjoy enhanced web search capabilities with intelligent memory, host management, and image analysis! 87 | 88 | --- 89 | 90 | ### 🧑‍💻 **Troubleshooting && Contributions** 91 | 92 | If you encounter any issues during installation or want to add some new feature, feel free to : 93 | 94 | - Open an Issue or Open an Pull Request to Contribute 95 | 96 | --- 97 | -------------------------------------------------------------------------------- /agent/__init__.py: -------------------------------------------------------------------------------- 1 | ## Made by Haseebasif7 -------------------------------------------------------------------------------- /agent/web_agent.py: -------------------------------------------------------------------------------- 1 | from memory.research_mem import ResearchMemory 2 | from tools.host_tracker import host_tracker 3 | from config.log import logger 4 | from typing import Dict, List 5 | from datetime import datetime 6 | from urllib.parse import urlparse 7 | import requests 8 | import time 9 | import json 10 | import re 11 | from Model.invokemodel import invoke_model 12 | from extras.safejsonload import safe_json_loads 13 | from config.settings import BRAVE_API_KEY 14 | from tools.extract_urls import extract_urls_from_search_results 15 | from tools.fetch_webpage import fetch_webpage_content 16 | 17 | class WebAgent: 18 | def __init__(self, retriever, llm, prompt, brave_search, wikipedia, provider): 19 | self.retriever = retriever 20 | self.llm = llm 21 | self.prompt = prompt 22 | self.brave_search = brave_search 23 | self.wikipedia = wikipedia 24 | self.provider = provider 25 | self.max_retries = 3 26 | self.retry_delay = 2 27 | self.research_memory = {} 28 | self.confidence_threshold = 0.5 29 | self.host_tracker = host_tracker 30 | self.current_topic = None 31 | self.memory = ResearchMemory() 32 | self.current_assessment = None 33 | 34 | def assess_content_relevance(self, content: str, topic: str) -> Dict: 35 | assessment_prompt = f"""You are a content assessment expert. Analyze this content's relevance and completeness for the given topic. 36 | Consider: 37 | 1. How directly it answers the topic/question 38 | 2. The specificity and accuracy of information 39 | 3. Whether it provides context and supporting details 40 | 4. The currentness and reliability of the information 41 | 42 | Topic: {topic} 43 | Content length: {len(content)} characters 44 | First 1000 chars: {content[:1000]} 45 | 46 | You must respond with ONLY a JSON object in this exact format: 47 | {{ 48 | "relevance": , 49 | "is_complete": , 50 | "found_data": "", 51 | "needs_verification": , 52 | "needs_context": , 53 | "confidence": 54 | }}""" 55 | 56 | try: 57 | response = invoke_model(self.llm, assessment_prompt) 58 | response_text = response.content.strip() 59 | json_match = re.search(r'\{[\s\S]*?\}', response_text) 60 | if not json_match: 61 | return { 62 | 'relevance': 0.5 if len(content) > 100 else 0.0, 63 | 'is_complete': False, 64 | 'found_data': content[:200] if len(content) > 0 else '', 65 | 'needs_verification': True, 66 | 'needs_context': True, 67 | 'confidence': 0.3 68 | } 69 | json_str = json_match.group(0) 70 | fallback = { 71 | 'relevance': 0.5 if len(content) > 100 else 0.0, 72 | 'is_complete': False, 73 | 'found_data': content[:200] if len(content) > 0 else '', 74 | 'needs_verification': True, 75 | 'needs_context': True, 76 | 'confidence': 0.3 77 | } 78 | result = safe_json_loads(json_str, fallback, content) 79 | return { 80 | 'relevance': float(result.get('relevance', 0)), 81 | 'is_complete': bool(result.get('is_complete', False)), 82 | 'found_data': str(result.get('found_data', '')), 83 | 'needs_verification': bool(result.get('needs_verification', True)), 84 | 'needs_context': bool(result.get('needs_context', True)), 85 | 'confidence': float(result.get('confidence', 0)) 86 | } 87 | except Exception as e: 88 | logger.error(f"Error in content assessment: {str(e)}") 89 | return { 90 | 'relevance': 0.0, 91 | 'is_complete': False, 92 | 'found_data': '', 93 | 'needs_verification': True, 94 | 'needs_context': True, 95 | 'confidence': 0.0 96 | } 97 | 98 | def extract_key_information(self, content: str, topic: str) -> Dict: 99 | extraction_prompt = f"""You are a precise information extractor. Extract key information from the content that is relevant to the topic. 100 | You must respond in valid JSON format with exactly these fields: 101 | {{ 102 | "main_facts": [list of key facts as strings], 103 | "confidence": number between 0.0-1.0, 104 | "timestamp": string or null, 105 | "source_quality": number between 0.0-1.0 106 | }} 107 | 108 | Topic: {topic} 109 | Content: {content} 110 | 111 | Respond ONLY with the JSON object, no other text:""" 112 | 113 | try: 114 | response = invoke_model(self.llm, extraction_prompt) 115 | response_text = response.content.strip() 116 | json_match = re.search(r'\{[\s\S]*?\}', response_text) 117 | if not json_match: 118 | return { 119 | "main_facts": ["Unable to extract structured information from source"], 120 | "confidence": 0.0, 121 | "timestamp": None, 122 | "source_quality": 0.0 123 | } 124 | json_str = json_match.group(0) 125 | fallback = { 126 | "main_facts": ["Unable to extract structured information from source"], 127 | "confidence": 0.0, 128 | "timestamp": None, 129 | "source_quality": 0.0 130 | } 131 | info = safe_json_loads(json_str, fallback, content) 132 | 133 | if not isinstance(info.get('main_facts', []), list): 134 | info['main_facts'] = [str(info.get('main_facts', ''))] 135 | 136 | return { 137 | 'main_facts': info.get('main_facts', []), 138 | 'confidence': min(max(info.get('confidence', 0.0), 0.0), 1.0), 139 | 'timestamp': info.get('timestamp'), 140 | 'source_quality': min(max(info.get('source_quality', 0.0), 0.0), 1.0) 141 | } 142 | except Exception as e: 143 | logger.error(f"Error extracting information: {str(e)}") 144 | return { 145 | "main_facts": ["Unable to extract structured information from source"], 146 | "confidence": 0.0, 147 | "timestamp": None, 148 | "source_quality": 0.0 149 | } 150 | 151 | def assess_question_complexity(self, topic: str) -> float: 152 | complexity_prompt = f""" 153 | Analyze the complexity of this research topic/question. 154 | Rate from 0.0 to 1.0, where: 155 | - 0.0: Very simple 156 | - 0.3: Basic fact-finding 157 | - 0.6: Moderate complexity 158 | - 1.0: Complex analysis 159 | 160 | Topic: {topic} 161 | 162 | Respond with only a number between 0.0 and 1.0:""" 163 | 164 | try: 165 | response = invoke_model(self.llm, complexity_prompt) 166 | matches = re.findall(r"0?\.[0-9]+", response.content) 167 | if matches: 168 | rating = float(matches[0]) 169 | else: 170 | rating = 0.5 171 | return min(max(rating, 0.0), 1.0) 172 | except Exception as e: 173 | logger.error(f"Error assessing question complexity: {str(e)}") 174 | return 0.5 175 | 176 | def _check_information_consistency(self, facts: List[str]) -> bool: 177 | return True 178 | 179 | def should_continue_research(self, topic: str, current_source: Dict) -> Dict: 180 | if topic not in self.research_memory: 181 | return {"continue": True, "reason": "No research started yet"} 182 | 183 | findings = self.research_memory[topic] 184 | sources_count = len(findings['sources']) 185 | complexity = self.assess_question_complexity(topic) 186 | 187 | min_sources = max(2, int(complexity * 5)) 188 | quality_threshold = 0.7 + (complexity * 0.2) 189 | high_quality_sources = sum(1 for s in findings['sources'] 190 | if s.get('relevance', 0) > quality_threshold 191 | and s.get('confidence', 0) > quality_threshold) 192 | 193 | if high_quality_sources >= min_sources: 194 | return {"continue": False, "reason": "Sufficient high-quality sources found"} 195 | 196 | max_sources = min_sources * 2 197 | if sources_count >= max_sources: 198 | return {"continue": False, "reason": "Maximum sources reached"} 199 | 200 | if sources_count > 1: 201 | info_consistent = self._check_information_consistency(findings['main_facts']) 202 | if not info_consistent: 203 | return {"continue": True, "reason": "Found inconsistent information", "priority": "verification"} 204 | 205 | if sources_count > 0: 206 | latest_source = findings['sources'][-1] 207 | if latest_source.get('needs_verification', True): 208 | return {"continue": True, "reason": "Need verification", "priority": "verification"} 209 | if latest_source.get('needs_context', True) and complexity > 0.5: 210 | return {"continue": True, "reason": "Need context", "priority": "context"} 211 | 212 | return {"continue": True, "reason": "Need more information"} 213 | 214 | def brave_search_run(self, query: str, retries: int = 3) -> str: 215 | if not BRAVE_API_KEY: 216 | logger.error("Brave Search API key not set. Unable to perform search.") 217 | return "" 218 | for i in range(retries): 219 | try: 220 | return self.brave_search.run(query) 221 | except requests.HTTPError as e: 222 | if e.response.status_code == 429: 223 | logger.warning("Hit rate limit. Waiting before retry...") 224 | time.sleep((i+1)*2) 225 | else: 226 | logger.error(f"HTTP Error during Brave search: {str(e)}") 227 | time.sleep(2) 228 | except Exception as ex: 229 | logger.error(f"Error in Brave search: {str(ex)}") 230 | time.sleep(2) 231 | return "" 232 | 233 | def fetch_additional_info(self, topic: str) -> str: 234 | self.current_topic = topic 235 | query_type = self.memory.categorize_query(topic) 236 | 237 | if topic not in self.research_memory: 238 | self.research_memory[topic] = { 239 | 'sources': [], 240 | 'main_facts': [], 241 | 'last_update': time.time(), 242 | 'visited_urls': set() 243 | } 244 | 245 | all_research = [] 246 | research_status = {"continue": True, "reason": "Initial research"} 247 | 248 | if query_type == 'stock_price': 249 | priority_domains = [ 250 | 'marketwatch.com', 251 | 'finance.yahoo.com', 252 | 'bloomberg.com', 253 | 'reuters.com' 254 | ] 255 | 256 | for domain in priority_domains: 257 | if any(domain in s.get('url', '') for s in self.research_memory[topic]['sources']): 258 | continue 259 | 260 | search_query = f"site:{domain} {topic}" 261 | search_results = self.brave_search_run(search_query) 262 | urls = extract_urls_from_search_results(search_results) 263 | 264 | if urls: 265 | url = urls[0] 266 | if url not in self.research_memory[topic]['visited_urls']: 267 | self.research_memory[topic]['visited_urls'].add(url) 268 | content = fetch_webpage_content(url, self.provider, topic) 269 | 270 | assessment = self.assess_content_relevance(content, topic) 271 | if assessment['relevance'] > 0.7: 272 | info = self.extract_key_information(content, topic) 273 | current_source = {**assessment, **info} 274 | 275 | self.research_memory[topic]['sources'].append({ 276 | 'url': url, 277 | 'content': content, 278 | **current_source 279 | }) 280 | self.research_memory[topic]['main_facts'].extend(info['main_facts']) 281 | 282 | if assessment['relevance'] > 0.8 and assessment['confidence'] > 0.8: 283 | research_status = {"continue": False, "reason": "Found reliable stock price"} 284 | break 285 | 286 | search_attempts = 0 287 | max_search_attempts = 3 288 | 289 | while research_status["continue"] and search_attempts < max_search_attempts: 290 | try: 291 | if search_attempts == 0: 292 | search_query = topic 293 | elif search_attempts == 1: 294 | search_query = f"{topic} latest information" 295 | else: 296 | search_query = f"{topic} current data {datetime.now().strftime('%Y')}" 297 | 298 | if research_status.get("priority") == "verification": 299 | search_query += " facts verify source" 300 | elif research_status.get("priority") == "context": 301 | search_query += " background context" 302 | 303 | logger.info(f"Searching with query: {search_query}") 304 | results = self.brave_search_run(search_query) 305 | urls = extract_urls_from_search_results(results) 306 | 307 | urls = [url for url in urls if url not in self.research_memory[topic]['visited_urls']] 308 | 309 | if not urls: 310 | search_attempts += 1 311 | continue 312 | 313 | urls = self.memory.prioritize_urls(urls, topic) 314 | 315 | for url in urls[:2]: 316 | if url in self.research_memory[topic]['visited_urls']: 317 | continue 318 | 319 | self.research_memory[topic]['visited_urls'].add(url) 320 | start_time = time.time() 321 | content = fetch_webpage_content(url, self.provider, topic) 322 | response_time = time.time() - start_time 323 | 324 | assessment = self.assess_content_relevance(content, topic) 325 | domain = urlparse(url).netloc 326 | 327 | success = assessment['relevance'] > 0.5 328 | self.memory.update_source_reliability( 329 | domain=domain, 330 | query_type=query_type, 331 | success=success, 332 | response_time=response_time, 333 | content_quality=assessment['relevance'] 334 | ) 335 | 336 | if success: 337 | info = self.extract_key_information(content, topic) 338 | current_source = {**assessment, **info} 339 | 340 | self.research_memory[topic]['sources'].append({ 341 | 'url': url, 342 | 'content': content, 343 | **current_source 344 | }) 345 | self.research_memory[topic]['main_facts'].extend(info['main_facts']) 346 | 347 | research_status = self.should_continue_research(topic, current_source) 348 | logger.info(f"Research status: {research_status['reason']}") 349 | 350 | if not research_status["continue"]: 351 | break 352 | 353 | if not research_status["continue"]: 354 | break 355 | 356 | search_attempts += 1 357 | 358 | except Exception as e: 359 | logger.error(f"Error in research iteration: {str(e)}") 360 | search_attempts += 1 361 | 362 | all_research.append(f""" 363 | === Research Summary === 364 | Query Type: {query_type} 365 | Total Sources: {len(self.research_memory[topic]['sources'])} 366 | Key Facts Found: {json.dumps(self.research_memory[topic]['main_facts'], indent=2)} 367 | Sources: {json.dumps([{ 368 | 'url': s['url'], 369 | 'relevance': s.get('relevance', 0), 370 | 'confidence': s.get('confidence', 0), 371 | 'found_data': s.get('found_data', '') 372 | } for s in self.research_memory[topic]['sources']], indent=2)} 373 | """) 374 | 375 | return "\n\n".join(all_research) 376 | 377 | def generate_report(self, topic: str) -> str: 378 | additional_info = self.fetch_additional_info(topic) 379 | 380 | enhanced_prompt = f""" 381 | Generate a report based on the research findings. 382 | Focus on the most relevant and current information. 383 | 384 | Topic: {topic} 385 | Research Findings: {additional_info} 386 | 387 | Guidelines: 388 | 1. Prioritize information from high-quality sources 389 | 2. Include specific, factual information 390 | 3. Note any significant gaps or uncertainties 391 | 4. Cite sources where appropriate 392 | 393 | Report:""" 394 | 395 | for attempt in range(self.max_retries): 396 | try: 397 | report = invoke_model(self.llm, enhanced_prompt) 398 | return report.content 399 | except Exception as e: 400 | if attempt == self.max_retries - 1: 401 | return f"Error generating report: {str(e)}" 402 | time.sleep(self.retry_delay) 403 | 404 | def assess_research_accuracy(self, topic: str, research_data: Dict) -> Dict: 405 | assessment_prompt = f"""Analyze the research results for accuracy and completeness. 406 | Consider: 407 | 1. Consistency across sources 408 | 2. Data freshness and relevance 409 | 3. Source reliability 410 | 4. Information completeness 411 | 412 | Topic: {topic} 413 | Research Data: {json.dumps(research_data, indent=2)} 414 | 415 | Respond with JSON: 416 | {{ 417 | "is_accurate": boolean, 418 | "confidence": float, 419 | "completeness": float, 420 | "concerns": [string], 421 | "verification_needed": boolean 422 | }}""" 423 | 424 | try: 425 | response = invoke_model(self.llm, assessment_prompt) 426 | json_match = re.search(r'\{[\s\S]*?\}', response.content) 427 | if json_match: 428 | assessment = safe_json_loads(json_match.group(0), { 429 | "is_accurate": False, 430 | "confidence": 0.0, 431 | "completeness": 0.0, 432 | "concerns": ["Assessment failed"], 433 | "verification_needed": True 434 | }) 435 | else: 436 | assessment = { 437 | "is_accurate": False, 438 | "confidence": 0.0, 439 | "completeness": 0.0, 440 | "concerns": ["No JSON returned"], 441 | "verification_needed": True 442 | } 443 | self.current_assessment = assessment 444 | return assessment 445 | except Exception as e: 446 | logger.error(f"Error in research assessment: {str(e)}") 447 | return { 448 | "is_accurate": False, 449 | "confidence": 0.0, 450 | "completeness": 0.0, 451 | "concerns": ["Assessment failed"], 452 | "verification_needed": True 453 | } 454 | 455 | def record_human_feedback(self, topic: str, is_accurate: bool, notes: str = None): 456 | if not self.current_assessment: 457 | logger.error("No current research assessment available") 458 | return 459 | 460 | sources = [s['url'] for s in self.research_memory.get(topic, {}).get('sources', [])] 461 | self.memory.record_feedback( 462 | topic=topic, 463 | sources=sources, 464 | agent_assessment=self.current_assessment, 465 | human_feedback=is_accurate, 466 | notes=notes 467 | ) 468 | 469 | self.current_assessment = None -------------------------------------------------------------------------------- /config/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Haseebasif7/SurfAgent/e527e1bce6234e3b4456298811550bf3a7552375/config/__init__.py -------------------------------------------------------------------------------- /config/log.py: -------------------------------------------------------------------------------- 1 | import logging 2 | 3 | logging.basicConfig(level=logging.INFO) 4 | logger = logging.getLogger(__name__) -------------------------------------------------------------------------------- /config/settings.py: -------------------------------------------------------------------------------- 1 | import os 2 | from dotenv import load_dotenv 3 | 4 | load_dotenv() 5 | 6 | GROQ_API_KEY = os.getenv("GROQ_API_KEY") 7 | BRAVE_API_KEY = os.getenv("BRAVE_API_KEY") -------------------------------------------------------------------------------- /config/user_settings.py: -------------------------------------------------------------------------------- 1 | from datetime import datetime 2 | import pytz 3 | import locale 4 | import geocoder 5 | from dataclasses import dataclass 6 | from config.log import logger 7 | 8 | @dataclass 9 | class UserSettings: 10 | country: str 11 | timezone: str 12 | locale: str 13 | current_time: datetime 14 | currency: str 15 | 16 | @classmethod 17 | def auto_detect(cls) -> 'UserSettings': 18 | """Automatically detect user settings dynamically.""" 19 | try: 20 | # Detect country using geolocation 21 | location = geocoder.ip('me') 22 | country = location.country if location and location.country else "Unknown" 23 | 24 | # Detect timezone using geolocation 25 | if location and location.latlng: 26 | country_timezones = pytz.country_timezones.get(location.country, []) 27 | local_tz = pytz.timezone(country_timezones[0]) if country_timezones else pytz.UTC 28 | else: 29 | local_tz = pytz.UTC 30 | 31 | # Get current time in detected timezone 32 | current_time = datetime.now(local_tz) 33 | 34 | # Detect locale based on country 35 | if country != "Unknown": 36 | locale_str = f'en_{country.upper()}.UTF-8' 37 | else: 38 | locale_str = 'en_US.UTF-8' 39 | locale.setlocale(locale.LC_ALL, locale_str) 40 | 41 | # Set currency based on country 42 | country_currency_map = { 43 | 'US': 'USD', 44 | 'CA': 'CAD', 45 | 'GB': 'GBP', 46 | 'AU': 'AUD', 47 | 'IN': 'INR', 48 | # Add more mappings as needed 49 | } 50 | currency = country_currency_map.get(country, 'USD') # Default to USD 51 | 52 | logger.info(f"Detected user settings: Country={country}, Timezone={local_tz}, Current time={current_time}") 53 | 54 | return cls( 55 | country=country, 56 | timezone=str(local_tz), 57 | locale=locale_str, 58 | current_time=current_time, 59 | currency=currency 60 | ) 61 | except Exception as e: 62 | logger.error(f"Error detecting user settings: {str(e)}") 63 | return cls( 64 | country="Unknown", 65 | timezone="UTC", 66 | locale="en_US.UTF-8", 67 | current_time=datetime.now(pytz.UTC), 68 | currency="USD" 69 | ) -------------------------------------------------------------------------------- /configure/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Haseebasif7/SurfAgent/e527e1bce6234e3b4456298811550bf3a7552375/configure/__init__.py -------------------------------------------------------------------------------- /configure/config_llm.py: -------------------------------------------------------------------------------- 1 | from langchain_groq import ChatGroq 2 | from langchain_ollama import ChatOllama 3 | from Model.provider import ModelProvider 4 | from typing import Union 5 | from config.settings import GROQ_API_KEY 6 | 7 | def configure_llm(provider: str) -> Union[ChatOllama, ChatGroq]: 8 | """Configure LLM based on selected provider.""" 9 | if provider == ModelProvider.OLLAMA: 10 | return ChatOllama( 11 | model="llama3.2:3b-instruct-q8_0", 12 | base_url="http://localhost:11434", 13 | temperature=0.5, 14 | num_gpu=1, 15 | num_thread=8 16 | ) 17 | else: 18 | return ChatGroq( 19 | model="llama-3.3-70b-specdec", 20 | temperature=0.5, 21 | groq_api_key=GROQ_API_KEY 22 | ) -------------------------------------------------------------------------------- /configure/llama.py: -------------------------------------------------------------------------------- 1 | from Model.provider import ModelProvider 2 | from langchain.prompts import PromptTemplate 3 | from configure.config_llm import configure_llm 4 | from config.log import logger 5 | import sys 6 | from test.test_model import test_model_provider 7 | 8 | def configure_llama(): 9 | """Configure model and prompt based on user's choice.""" 10 | provider = ModelProvider.get_provider_choice() 11 | if not test_model_provider(provider): 12 | logger.error(f"Failed to initialize {provider} models") 13 | sys.exit(1) 14 | llm = configure_llm(provider) 15 | prompt = PromptTemplate( 16 | template="""You are an assistant for research tasks. Use the following documents to provide a comprehensive and concise report on the topic. Ensure the report is self-contained with all necessary information. 17 | 18 | Topic: {topic} 19 | Documents: {documents} 20 | Report: """, 21 | input_variables=["topic", "documents"], 22 | ) 23 | return llm, prompt, provider -------------------------------------------------------------------------------- /configure/vision.py: -------------------------------------------------------------------------------- 1 | from langchain_groq import ChatGroq 2 | from langchain_ollama import ChatOllama 3 | from Model.provider import ModelProvider 4 | from typing import Union 5 | from config.settings import GROQ_API_KEY 6 | 7 | def configure_vision_model(provider: str) -> Union[ChatOllama, ChatGroq]: 8 | """Configure vision model based on selected provider.""" 9 | if provider == ModelProvider.OLLAMA: 10 | return ChatOllama( 11 | model="llama3.2-vision:11b", 12 | base_url="http://localhost:11434", 13 | temperature=0.5, 14 | num_gpu=1, 15 | num_thread=8, 16 | madvise=True, 17 | f16=True 18 | ) 19 | else: 20 | return ChatGroq( 21 | model="llama-3.2-90b-vision-preview", 22 | temperature=0.5, 23 | groq_api_key=GROQ_API_KEY 24 | ) -------------------------------------------------------------------------------- /extras/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Haseebasif7/SurfAgent/e527e1bce6234e3b4456298811550bf3a7552375/extras/__init__.py -------------------------------------------------------------------------------- /extras/pawelzmarlak-2025-01-22T16_41_02.675Z.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Haseebasif7/SurfAgent/e527e1bce6234e3b4456298811550bf3a7552375/extras/pawelzmarlak-2025-01-22T16_41_02.675Z.png -------------------------------------------------------------------------------- /extras/pawelzmarlak-2025-02-13T06_51_12.773Z.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Haseebasif7/SurfAgent/e527e1bce6234e3b4456298811550bf3a7552375/extras/pawelzmarlak-2025-02-13T06_51_12.773Z.png -------------------------------------------------------------------------------- /extras/safejsonload.py: -------------------------------------------------------------------------------- 1 | from typing import Dict 2 | import json 3 | from config.log import logger 4 | 5 | def safe_json_loads(json_str: str, fallback: Dict, content: str = "") -> Dict: 6 | try: 7 | return json.loads(json_str) 8 | except json.JSONDecodeError: 9 | fixed = json_str.replace('""', '"') 10 | try: 11 | return json.loads(fixed) 12 | except json.JSONDecodeError: 13 | logger.error("Still can't parse JSON after fix.") 14 | return fallback -------------------------------------------------------------------------------- /extras/viewimage.py: -------------------------------------------------------------------------------- 1 | from typing import Dict 2 | import json 3 | from config.log import logger 4 | 5 | def safe_json_loads(json_str: str, fallback: Dict, content: str = "") -> Dict: 6 | try: 7 | return json.loads(json_str) 8 | except json.JSONDecodeError: 9 | fixed = json_str.replace('""', '"') 10 | try: 11 | return json.loads(fixed) 12 | except json.JSONDecodeError: 13 | logger.error("Still can't parse JSON after fix.") 14 | return fallback -------------------------------------------------------------------------------- /main.py: -------------------------------------------------------------------------------- 1 | from config.log import logger 2 | from test.selenium import test_selenium 3 | import sys 4 | from Model.provider import ModelProvider 5 | from test.test_model import test_model_provider 6 | from configure.llama import configure_llama 7 | from config.settings import BRAVE_API_KEY 8 | from langchain_community.tools import BraveSearch, WikipediaQueryRun 9 | from langchain_community.utilities.wikipedia import WikipediaAPIWrapper 10 | from agent.web_agent import WebAgent 11 | import time 12 | import warnings 13 | 14 | warnings.filterwarnings("ignore") 15 | 16 | def print_banner(): 17 | banner = """ 18 | \033[92m 19 | ____ _ _ ____ ____ __ ___ ____ __ _ ____ 20 | / ___)/ )( \( _ \( __) / _\ / __)( __)( ( \(_ _) 21 | \___ \) \/ ( ) / ) _) / \( (_ \ ) _) / / )( 22 | (____/\____/(__\_)(__) \_/\_/ \___/(____)\_)__) (__) 23 | \033[0m 24 | """ 25 | print(banner) 26 | print("\033[92mSurf: Your Intelligent Web Companion!\033[0m\n") 27 | 28 | def print_separator(): 29 | print("=" * 80) 30 | 31 | def main(): 32 | print_banner() 33 | print("Initializing SurfAgent...") 34 | time.sleep(1) 35 | 36 | if not test_selenium(): 37 | logger.error("Selenium service check failed") 38 | print("Error: Selenium service is not working. Exiting.") 39 | sys.exit(1) 40 | 41 | provider = input("Choose model provider (ollama/groq): ").lower().strip() 42 | if provider not in [ModelProvider.OLLAMA, ModelProvider.GROQ]: 43 | logger.error("Invalid provider choice") 44 | print("Error: Invalid provider. Exiting.") 45 | sys.exit(1) 46 | 47 | if not test_model_provider(provider): 48 | logger.error(f"Model provider {provider} check failed") 49 | print(f"Error: Model provider {provider} is not functioning correctly. Exiting.") 50 | sys.exit(1) 51 | 52 | llm, prompt, provider = configure_llama() 53 | 54 | if not BRAVE_API_KEY: 55 | logger.warning("Brave Search API key not set. Searches will not return results.") 56 | print("Warning: Brave Search API key not found. Limited functionality.") 57 | 58 | brave_search = BraveSearch.from_api_key( 59 | api_key=BRAVE_API_KEY, 60 | search_kwargs={"count": 6} 61 | ) 62 | wikipedia = WikipediaQueryRun(api_wrapper=WikipediaAPIWrapper()) 63 | retriever = None 64 | agent = WebAgent(retriever, llm, prompt, brave_search, wikipedia, provider) 65 | 66 | print("SurfAgent is ready to assist you! 🚀\n") 67 | print_separator() 68 | 69 | while True: 70 | try: 71 | topic = input("\n🌐 Enter a topic for web search (or type 'quit' to exit): ").strip() 72 | if topic.lower() == 'quit': 73 | print("\n👋 Thank you for using SurfAgent. Goodbye!") 74 | break 75 | 76 | if not topic: 77 | print("⚠️ Please enter a valid topic.") 78 | continue 79 | logger.info(f"Starting research for topic: {topic}") 80 | print(f"\n🔍 Researching: {topic}...") 81 | report = agent.generate_report(topic) 82 | 83 | print("\n📜 Response:") 84 | print_separator() 85 | print(report) 86 | print_separator() 87 | 88 | feedback = input("\n✅ Was this information accurate? (y/n): ").lower().strip() 89 | if feedback in ['y', 'n']: 90 | is_accurate = feedback == 'y' 91 | notes = input("📝 Any additional notes? (Enter to skip): ").strip() 92 | agent.record_human_feedback(topic, is_accurate, notes if notes else None) 93 | print("🙏 Thank you for your feedback!") 94 | 95 | except KeyboardInterrupt: 96 | print("\n🛑 Research interrupted by user.") 97 | break 98 | except Exception as e: 99 | logger.error(f"Error during research: {str(e)}") 100 | print("⚠️ An error occurred. Please try again.") 101 | continue 102 | 103 | if __name__ == "__main__": 104 | main() 105 | -------------------------------------------------------------------------------- /memory/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Haseebasif7/SurfAgent/e527e1bce6234e3b4456298811550bf3a7552375/memory/__init__.py -------------------------------------------------------------------------------- /memory/research_mem.py: -------------------------------------------------------------------------------- 1 | import os 2 | import json 3 | import re 4 | from datetime import datetime, timezone 5 | from urllib.parse import urlparse 6 | from source_reliable.source_reliability_class import SourceReliability 7 | from config.log import logger 8 | from typing import Dict, List 9 | 10 | 11 | class ResearchMemory: 12 | def __init__(self, memory_file="agent_memory.json"): 13 | self.memory_file = memory_file 14 | self.source_reliability = {} 15 | self.query_patterns = {} 16 | self.feedback_history = {} 17 | self.load_memory() 18 | 19 | def load_memory(self): 20 | try: 21 | if os.path.exists(self.memory_file): 22 | with open(self.memory_file, 'r') as f: 23 | data = json.load(f) 24 | 25 | for domain, info in data.get('sources', {}).items(): 26 | self.source_reliability[domain] = SourceReliability( 27 | domain=domain, 28 | query_types=info.get('query_types', {}), 29 | last_success=datetime.fromisoformat(info['last_success']) if info.get('last_success') else None, 30 | last_failure=datetime.fromisoformat(info['last_failure']) if info.get('last_failure') else None, 31 | total_attempts=info.get('total_attempts', 0), 32 | successful_attempts=info.get('successful_attempts', 0), 33 | average_response_time=info.get('average_response_time', 0.0), 34 | notes=info.get('notes', []) 35 | ) 36 | 37 | self.query_patterns = data.get('query_patterns', {}) 38 | self.feedback_history = data.get('feedback_history', {}) 39 | 40 | logger.info(f"Loaded research memory with {len(self.source_reliability)} sources and {len(self.feedback_history)} feedback entries") 41 | except Exception as e: 42 | logger.error(f"Error loading research memory: {str(e)}") 43 | self.source_reliability = {} 44 | self.query_patterns = {} 45 | self.feedback_history = {} 46 | 47 | def save_memory(self): 48 | try: 49 | data = { 50 | 'sources': { 51 | domain: { 52 | 'query_types': info.query_types, 53 | 'last_success': info.last_success.isoformat() if info.last_success else None, 54 | 'last_failure': info.last_failure.isoformat() if info.last_failure else None, 55 | 'total_attempts': info.total_attempts, 56 | 'successful_attempts': info.successful_attempts, 57 | 'average_response_time': info.average_response_time, 58 | 'notes': info.notes 59 | } 60 | for domain, info in self.source_reliability.items() 61 | }, 62 | 'query_patterns': self.query_patterns, 63 | 'feedback_history': self.feedback_history 64 | } 65 | 66 | with open(self.memory_file, 'w') as f: 67 | json.dump(data, f, indent=2) 68 | 69 | logger.info("Successfully saved research memory") 70 | except Exception as e: 71 | logger.error(f"Error saving research memory: {str(e)}") 72 | 73 | def categorize_query(self, query: str) -> str: 74 | categories = { 75 | 'stock_price': r'(?i)(stock|share)\s+price|price\s+of\s+stock', 76 | 'financial_data': r'(?i)financial|revenue|earnings|profit|market\s+cap', 77 | 'company_info': r'(?i)headquarters|ceo|founded|employees|about', 78 | 'news': r'(?i)news|latest|recent|update|announce', 79 | 'technical': r'(?i)technology|software|product|service|api', 80 | 'general': r'.*' 81 | } 82 | 83 | for category, pattern in categories.items(): 84 | if re.search(pattern, query): 85 | return category 86 | return 'general' 87 | 88 | def update_source_reliability(self, domain: str, query_type: str, success: bool, response_time: float, content_quality: float): 89 | if domain not in self.source_reliability: 90 | self.source_reliability[domain] = SourceReliability( 91 | domain=domain, 92 | query_types={}, 93 | last_success=None, 94 | last_failure=None, 95 | total_attempts=0, 96 | successful_attempts=0, 97 | average_response_time=0.0, 98 | notes=[] 99 | ) 100 | 101 | source = self.source_reliability[domain] 102 | current_time = datetime.now(timezone.utc) 103 | 104 | if query_type not in source.query_types: 105 | source.query_types[query_type] = 0.0 106 | 107 | source.total_attempts += 1 108 | if success: 109 | source.successful_attempts += 1 110 | source.last_success = current_time 111 | source.query_types[query_type] = ( 112 | source.query_types[query_type] * 0.9 + 113 | content_quality * 0.1 114 | ) 115 | else: 116 | source.last_failure = current_time 117 | source.query_types[query_type] *= 0.9 118 | 119 | source.average_response_time = ( 120 | source.average_response_time * 0.9 + 121 | response_time * 0.1 122 | ) 123 | 124 | self.save_memory() 125 | 126 | def get_best_sources(self, query_type: str, min_reliability: float = 0.3) -> List[str]: 127 | relevant_sources = [] 128 | 129 | for domain, source in self.source_reliability.items(): 130 | reliability = source.query_types.get(query_type, 0.0) 131 | if reliability >= min_reliability: 132 | relevant_sources.append((domain, reliability)) 133 | 134 | relevant_sources.sort(key=lambda x: x[1], reverse=True) 135 | return [domain for domain, _ in relevant_sources] 136 | 137 | def prioritize_urls(self, urls: List[str], query: str) -> List[str]: 138 | query_type = self.categorize_query(query) 139 | self.get_best_sources(query_type) 140 | 141 | scored_urls = [] 142 | for url in urls: 143 | domain = urlparse(url).netloc 144 | source = self.source_reliability.get(domain) 145 | 146 | if source: 147 | reliability = source.query_types.get(query_type, 0.0) 148 | success_rate = source.successful_attempts / max(1, source.total_attempts) 149 | response_speed = 1.0 / (1.0 + source.average_response_time) 150 | score = (reliability * 0.5 + 151 | success_rate * 0.3 + 152 | response_speed * 0.2) 153 | else: 154 | score = 0.1 155 | 156 | scored_urls.append((url, score)) 157 | 158 | scored_urls.sort(key=lambda x: x[1], reverse=True) 159 | return [url for url, _ in scored_urls] 160 | 161 | def record_feedback(self, topic: str, sources: List[str], agent_assessment: Dict, human_feedback: bool, notes: str = None): 162 | current_time = datetime.now(timezone.utc) 163 | query_type = self.categorize_query(topic) 164 | 165 | feedback_entry = { 166 | 'timestamp': current_time.isoformat(), 167 | 'topic': topic, 168 | 'sources': sources, 169 | 'agent_assessment': agent_assessment, 170 | 'human_feedback': human_feedback, 171 | 'query_type': query_type, 172 | 'notes': notes 173 | } 174 | 175 | if topic not in self.feedback_history: 176 | self.feedback_history[topic] = [] 177 | self.feedback_history[topic].append(feedback_entry) 178 | 179 | agent_confidence = agent_assessment.get('confidence', 0.0) 180 | agent_correct = agent_assessment.get('is_accurate', False) 181 | 182 | for source in sources: 183 | domain = urlparse(source).netloc 184 | if domain not in self.source_reliability: 185 | continue 186 | 187 | source_info = self.source_reliability[domain] 188 | 189 | if human_feedback: 190 | if agent_correct == human_feedback: 191 | self._update_source_confidence(domain, query_type, True, 1.0) 192 | source_info.notes.append(f"[{current_time.isoformat()}] Accurate assessment confirmed by human feedback") 193 | else: 194 | self._update_source_confidence(domain, query_type, False, 1.0) 195 | source_info.notes.append(f"[{current_time.isoformat()}] Assessment contradicted by human feedback") 196 | else: 197 | self._update_source_confidence(domain, query_type, agent_correct, agent_confidence) 198 | 199 | self.save_memory() 200 | 201 | def _update_source_confidence(self, domain: str, query_type: str, success: bool, confidence: float): 202 | source = self.source_reliability[domain] 203 | 204 | if query_type not in source.query_types: 205 | source.query_types[query_type] = 0.0 206 | 207 | current_reliability = source.query_types[query_type] 208 | 209 | if success: 210 | new_reliability = current_reliability + (1 - current_reliability) * confidence * 0.1 211 | else: 212 | new_reliability = current_reliability * 0.8 213 | 214 | source.query_types[query_type] = max(0.0, min(1.0, new_reliability)) 215 | 216 | def get_feedback_stats(self, domain: str = None, query_type: str = None) -> Dict: 217 | stats = { 218 | 'total_entries': 0, 219 | 'agent_accuracy': 0.0, 220 | 'human_agreement': 0.0, 221 | 'query_type_performance': {}, 222 | 'recent_trends': [] 223 | } 224 | 225 | relevant_entries = [] 226 | 227 | for topic_entries in self.feedback_history.values(): 228 | for entry in topic_entries: 229 | if domain and not any(domain in s for s in entry['sources']): 230 | continue 231 | if query_type and entry['query_type'] != query_type: 232 | continue 233 | relevant_entries.append(entry) 234 | 235 | if not relevant_entries: 236 | return stats 237 | 238 | stats['total_entries'] = len(relevant_entries) 239 | 240 | correct_assessments = sum(1 for e in relevant_entries 241 | if e['agent_assessment'].get('is_accurate') == e['human_feedback']) 242 | human_agreements = sum(1 for e in relevant_entries if e['human_feedback']) 243 | 244 | stats['agent_accuracy'] = correct_assessments / len(relevant_entries) 245 | stats['human_agreement'] = human_agreements / len(relevant_entries) 246 | 247 | query_types = {} 248 | for entry in relevant_entries: 249 | qt = entry['query_type'] 250 | if qt not in query_types: 251 | query_types[qt] = {'total': 0, 'successful': 0} 252 | query_types[qt]['total'] += 1 253 | if entry['human_feedback']: 254 | query_types[qt]['successful'] += 1 255 | 256 | stats['query_type_performance'] = { 257 | qt: {'success_rate': data['successful'] / data['total']} 258 | for qt, data in query_types.items() 259 | } 260 | 261 | recent = relevant_entries[-10:] 262 | stats['recent_trends'] = [ 263 | { 264 | 'timestamp': e['timestamp'], 265 | 'query_type': e['query_type'], 266 | 'success': e['human_feedback'] 267 | } 268 | for e in recent 269 | ] 270 | 271 | return stats -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | langchain 2 | langchain_community 3 | langchain_ollama 4 | langchain_groq 5 | requests 6 | beautifulsoup4 7 | pytz 8 | selenium 9 | webdriver_manager 10 | pillow 11 | geocoder -------------------------------------------------------------------------------- /source_reliable/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Haseebasif7/SurfAgent/e527e1bce6234e3b4456298811550bf3a7552375/source_reliable/__init__.py -------------------------------------------------------------------------------- /source_reliable/source_reliability_class.py: -------------------------------------------------------------------------------- 1 | from dataclasses import dataclass 2 | from datetime import datetime 3 | from typing import Dict, List, Optional 4 | 5 | @dataclass 6 | class SourceReliability: 7 | domain: str 8 | query_types: Dict[str, float] 9 | last_success: Optional[datetime] 10 | last_failure: Optional[datetime] 11 | total_attempts: int 12 | successful_attempts: int 13 | average_response_time: float 14 | notes: List[str] -------------------------------------------------------------------------------- /test/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Haseebasif7/SurfAgent/e527e1bce6234e3b4456298811550bf3a7552375/test/__init__.py -------------------------------------------------------------------------------- /test/ollama.py: -------------------------------------------------------------------------------- 1 | from config.log import logger 2 | from langchain_ollama import ChatOllama 3 | from langchain.schema import HumanMessage, AIMessage 4 | 5 | def test_ollama() -> bool: 6 | """Test if Ollama is running and accessible.""" 7 | try: 8 | test_llm = ChatOllama( 9 | model="llama3.2:3b-instruct-q8_0", 10 | base_url="http://localhost:11434", 11 | temperature=0, 12 | num_gpu=1, 13 | num_thread=8 14 | ) 15 | resp = test_llm([HumanMessage(content="Hello")]) 16 | if isinstance(resp, AIMessage) and len(resp.content) > 0: 17 | logger.info("✅ Ollama is accessible") 18 | return True 19 | else: 20 | logger.error("❌ Ollama did not return a valid response") 21 | return False 22 | except Exception as e: 23 | logger.error(f"❌ Ollama test failed: {str(e)}") 24 | return False -------------------------------------------------------------------------------- /test/selenium.py: -------------------------------------------------------------------------------- 1 | from selenium import webdriver 2 | from selenium.webdriver.chrome.service import Service 3 | from selenium.webdriver.chrome.options import Options 4 | from webdriver_manager.chrome import ChromeDriverManager 5 | import time 6 | import io 7 | from PIL import Image 8 | from tools.size_limit import ensure_size_within_limits 9 | from config.log import logger 10 | 11 | def test_selenium() -> bool: 12 | """Test if Selenium can run and capture a screenshot of a test page using Chrome.""" 13 | try: 14 | chrome_options = Options() 15 | chrome_options.add_argument('--headless') 16 | chrome_options.add_argument('--window-size=1920,1080') 17 | chrome_options.add_argument('--disable-gpu') # To avoid potential issues with headless mode 18 | 19 | driver = webdriver.Chrome( 20 | service=Service(ChromeDriverManager().install()), 21 | options=chrome_options 22 | ) 23 | 24 | driver.set_page_load_timeout(20) 25 | driver.get("https://example.com") 26 | 27 | # Set zoom level for better text legibility 28 | driver.execute_script("document.body.style.zoom = '200%'") # Increased from 150% 29 | 30 | # Ensure text is readable 31 | driver.execute_script(""" 32 | document.querySelectorAll('*').forEach(function(el) { 33 | let style = window.getComputedStyle(el); 34 | if (parseInt(style.fontSize) < 16) { // Increased minimum font size 35 | el.style.fontSize = '16px'; 36 | } 37 | // Improve contrast 38 | if (style.color && style.backgroundColor) { 39 | let textColor = style.color; 40 | let bgColor = style.backgroundColor; 41 | if (textColor === bgColor || textColor === 'rgba(0, 0, 0, 0)') { 42 | el.style.color = '#000000'; 43 | } 44 | } 45 | }); 46 | """) 47 | 48 | # Additional wait for text scaling 49 | time.sleep(1) 50 | 51 | # Get page dimensions with padding for better quality 52 | total_height = driver.execute_script("return Math.max(document.documentElement.scrollHeight, document.body.scrollHeight);") 53 | total_width = driver.execute_script("return Math.max(document.documentElement.scrollWidth, document.body.scrollWidth);") 54 | 55 | # Add padding and ensure minimum dimensions 56 | total_width = max(total_width, 1920) 57 | total_height = int(total_height * 1.1) 58 | 59 | # Ensure dimensions are within pixel limit 60 | final_width, final_height = ensure_size_within_limits(total_width, total_height) 61 | 62 | # Set window size with the adjusted dimensions 63 | driver.set_window_size(final_width, final_height) 64 | 65 | # Wait for any dynamic content to load 66 | time.sleep(1) 67 | 68 | # Capture full screenshot in memory with high quality 69 | screenshot_png = driver.get_screenshot_as_png() 70 | driver.quit() 71 | 72 | # Decode and verify image 73 | img = Image.open(io.BytesIO(screenshot_png)) 74 | img.verify() 75 | logger.info(f"✅ Selenium is running with Chrome and captured screenshot ({img.size[0]}x{img.size[1]} px)") 76 | return True 77 | except Exception as e: 78 | logger.error(f"❌ Selenium test failed: {str(e)}") 79 | return False -------------------------------------------------------------------------------- /test/test_model.py: -------------------------------------------------------------------------------- 1 | from config.log import logger 2 | from Model.provider import ModelProvider 3 | from config.settings import GROQ_API_KEY 4 | import requests 5 | from test.ollama import test_ollama 6 | 7 | def test_model_provider(provider: str) -> bool: 8 | """Test if the selected model provider is accessible.""" 9 | try: 10 | if provider == ModelProvider.OLLAMA: 11 | return test_ollama() 12 | else: 13 | if not GROQ_API_KEY: 14 | logger.error("GROQ_API_KEY not set.") 15 | return False 16 | headers = {"Authorization": f"Bearer {GROQ_API_KEY}"} 17 | response = requests.get("https://api.groq.com/openai/v1/models", headers=headers) 18 | response.raise_for_status() 19 | logger.info("✅ Groq API is accessible") 20 | return True 21 | except Exception as e: 22 | logger.error(f"❌ {provider.capitalize()} test failed: {str(e)}") 23 | return False -------------------------------------------------------------------------------- /tools/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Haseebasif7/SurfAgent/e527e1bce6234e3b4456298811550bf3a7552375/tools/__init__.py -------------------------------------------------------------------------------- /tools/capture_ss.py: -------------------------------------------------------------------------------- 1 | from config.log import logger 2 | from PIL import Image 3 | import io 4 | import math 5 | import time 6 | from tools.size_limit import ensure_size_within_limits 7 | 8 | def capture_full_page_screenshot(driver, url: str) -> bytes: 9 | """Capture a full page screenshot by scrolling and stitching.""" 10 | try: 11 | # Get initial dimensions 12 | total_height = driver.execute_script("return Math.max(document.documentElement.scrollHeight, document.body.scrollHeight);") 13 | total_width = driver.execute_script("return Math.max(document.documentElement.scrollWidth, document.body.scrollWidth);") 14 | 15 | # Calculate viewport height 16 | viewport_height = driver.execute_script("return window.innerHeight;") 17 | 18 | # Pre-calculate final dimensions to ensure they're within limits 19 | MAX_PIXELS = 33177600 * 0.9 # 10% safety margin 20 | 21 | # If the page is very long, we'll split it into sections 22 | if total_height > 15000 or (total_width * total_height) > MAX_PIXELS: 23 | # Calculate maximum height that would fit within pixel limit 24 | max_safe_height = int(MAX_PIXELS / total_width) 25 | 26 | # Adjust section size based on max safe height 27 | section_height = min(viewport_height, max_safe_height // 4) # Use quarter of max safe height per section 28 | 29 | sections = [] 30 | offset = 0 31 | while offset < total_height: 32 | # Scroll to position 33 | driver.execute_script(f"window.scrollTo(0, {offset});") 34 | time.sleep(0.5) # Wait for scroll and content to load 35 | 36 | # Capture viewport 37 | section_png = driver.get_screenshot_as_png() 38 | section = Image.open(io.BytesIO(section_png)) 39 | 40 | # Ensure section is within limits 41 | if section.height > section_height: 42 | section = section.crop((0, 0, section.width, section_height)) 43 | 44 | sections.append(section) 45 | offset += section_height 46 | 47 | # Calculate final dimensions ensuring they're within limits 48 | final_width = min(total_width, 1920) # Cap width at 1920px 49 | final_height = min(total_height, int(MAX_PIXELS / final_width)) 50 | 51 | # Create new image with calculated dimensions 52 | final_image = Image.new('RGB', (final_width, final_height)) 53 | y_offset = 0 54 | 55 | for section in sections: 56 | if y_offset + section.height > final_height: 57 | # Crop section if it would exceed final height 58 | remaining_height = final_height - y_offset 59 | if remaining_height <= 0: 60 | break 61 | section = section.crop((0, 0, section.width, remaining_height)) 62 | 63 | final_image.paste(section, (0, y_offset)) 64 | y_offset += section.height 65 | if y_offset >= final_height: 66 | break 67 | 68 | # Verify final size 69 | if final_image.width * final_image.height > MAX_PIXELS: 70 | # Resize if somehow still too large 71 | scale = math.sqrt(MAX_PIXELS / (final_image.width * final_image.height)) 72 | new_width = int(final_image.width * scale) 73 | new_height = int(final_image.height * scale) 74 | final_image = final_image.resize((new_width, new_height), Image.Resampling.LANCZOS) 75 | 76 | # Convert to PNG 77 | output = io.BytesIO() 78 | final_image.save(output, format='PNG', optimize=True) 79 | return output.getvalue() 80 | else: 81 | # For shorter pages, still ensure we're within limits 82 | final_width, final_height = ensure_size_within_limits(total_width, total_height) 83 | driver.set_window_size(final_width, final_height) 84 | time.sleep(0.5) 85 | return driver.get_screenshot_as_png() 86 | 87 | except Exception as e: 88 | logger.error(f"Error in full page capture: {str(e)}") 89 | # Fallback to a safe capture 90 | safe_width, safe_height = ensure_size_within_limits(1920, 1080) 91 | driver.set_window_size(safe_width, safe_height) 92 | return driver.get_screenshot_as_png() -------------------------------------------------------------------------------- /tools/create_vecstore.py: -------------------------------------------------------------------------------- 1 | from langchain_community.vectorstores import SKLearnVectorStore 2 | from langchain_ollama import OllamaEmbeddings 3 | 4 | def create_vectorstore(docs_splits): 5 | embeddings = OllamaEmbeddings( 6 | model="all-minilm", 7 | base_url="http://localhost:11434/v1" 8 | ) 9 | vectorstore = SKLearnVectorStore.from_documents( 10 | documents=docs_splits, 11 | embedding=embeddings 12 | ) 13 | return vectorstore.as_retriever(k=4) -------------------------------------------------------------------------------- /tools/extract_urls.py: -------------------------------------------------------------------------------- 1 | import re 2 | from urllib.parse import urlparse 3 | from typing import List 4 | from config.log import logger 5 | from tools.host_tracker import host_tracker 6 | 7 | def extract_urls_from_search_results(search_text: str) -> List[str]: 8 | urls = re.findall(r'(https?://[^\s\'"]+)', search_text) 9 | valid_urls = [] 10 | for url in urls: 11 | url = re.sub(r'[.,)\]]+$', '', url) 12 | if url.startswith(('http://', 'https://')): 13 | if not host_tracker.is_problematic_host(url): 14 | valid_urls.append(url) 15 | else: 16 | logger.info(f"Filtered out problematic host: {urlparse(url).netloc}") 17 | return list(set(valid_urls)) -------------------------------------------------------------------------------- /tools/fetch_webpage.py: -------------------------------------------------------------------------------- 1 | from config.log import logger 2 | from tools.host_tracker import host_tracker 3 | from PIL import Image 4 | from selenium import webdriver 5 | from selenium.webdriver.chrome.service import Service 6 | from selenium.webdriver.chrome.options import Options 7 | from webdriver_manager.chrome import ChromeDriverManager 8 | from urllib.parse import urlparse 9 | import time 10 | import io 11 | import base64 12 | from tools.size_limit import ensure_size_within_limits 13 | from tools.capture_ss import capture_full_page_screenshot 14 | from configure.vision import configure_vision_model 15 | from configure.config_llm import configure_llm 16 | from tools.vision_query import generate_vision_query 17 | 18 | 19 | def fetch_webpage_content(url: str, provider: str, original_query: str) -> str: 20 | """Fetch webpage content by capturing a screenshot via Selenium and processing it with a vision model.""" 21 | if host_tracker .is_problematic_host(url): 22 | logger.info(f"Skipping known problematic host: {urlparse(url).netloc}") 23 | return f"Skipped: Known problematic host" 24 | 25 | try: 26 | # Set up Selenium (headless Chrome) 27 | chrome_options = Options() 28 | chrome_options.add_argument('--headless') 29 | chrome_options.add_argument('--window-size=1920,1080') 30 | chrome_options.add_argument('--disable-gpu') # To avoid potential issues with headless mode 31 | chrome_options.add_argument('--disable-blink-features=AutomationControlled') # Hide automation 32 | chrome_options.add_argument('--disable-notifications') 33 | 34 | # Add headers to appear more like a real browser 35 | chrome_options.add_argument('user-agent=Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/122.0.0.0 Safari/537.36') 36 | 37 | driver = webdriver.Chrome( 38 | service=Service(ChromeDriverManager().install()), 39 | options=chrome_options 40 | ) 41 | driver.set_page_load_timeout(60) 42 | 43 | # Set cookies and localStorage to bypass some anti-bot measures 44 | driver.execute_script("Object.defineProperty(navigator, 'webdriver', {get: () => undefined})") 45 | 46 | # Attempt to load the page 47 | driver.get(url) 48 | time.sleep(2) # Give time for dynamic content to load 49 | 50 | # Check for and handle CAPTCHA/cookie popups 51 | try: 52 | driver.execute_script(""" 53 | // Remove common overlay elements 54 | document.querySelectorAll('[class*="cookie"], [class*="popup"], [class*="modal"], [id*="cookie"], [id*="popup"], [id*="modal"]') 55 | .forEach(el => el.remove()); 56 | // Remove fixed position elements that might overlay content 57 | document.querySelectorAll('*').forEach(el => { 58 | const style = window.getComputedStyle(el); 59 | if (style.position === 'fixed' || style.position === 'sticky') { 60 | el.remove(); 61 | } 62 | }); 63 | """) 64 | except Exception as e: 65 | logger.warning(f"Error handling overlays: {str(e)}") 66 | 67 | # Set text size and ensure readability with special handling for financial data 68 | driver.execute_script(""" 69 | // Set base zoom 70 | document.body.style.zoom = '200%'; // Increased from 125% 71 | 72 | // Function to check if text might be financial data 73 | function isFinancialData(text) { 74 | return /\\$|\\d+\\.\\d+|\\d+%|price|stock|market|share/i.test(text); 75 | } 76 | 77 | // Ensure text is readable with special handling for financial data 78 | document.querySelectorAll('*').forEach(function(el) { 79 | let style = window.getComputedStyle(el); 80 | let text = el.textContent || ''; 81 | 82 | // Special handling for financial data 83 | if (isFinancialData(text)) { 84 | el.style.fontSize = '24px'; // Larger size for financial data 85 | el.style.fontWeight = 'bold'; 86 | el.style.color = '#000000'; // Ensure high contrast 87 | } else if (parseInt(style.fontSize) < 16) { // Increased minimum font size 88 | el.style.fontSize = '16px'; 89 | } 90 | 91 | // Improve contrast 92 | if (style.color && style.backgroundColor) { 93 | let textColor = style.color; 94 | let bgColor = style.backgroundColor; 95 | if (textColor === bgColor || textColor === 'rgba(0, 0, 0, 0)' || 96 | textColor === 'rgb(255, 255, 255)' || textColor === '#ffffff') { 97 | el.style.color = '#000000'; 98 | } 99 | } 100 | 101 | // Improve visibility of links 102 | if (el.tagName.toLowerCase() === 'a') { 103 | el.style.textDecoration = 'underline'; 104 | } 105 | }); 106 | 107 | // Additional handling for table cells (common in financial data) 108 | document.querySelectorAll('td, th').forEach(function(el) { 109 | let text = el.textContent || ''; 110 | if (isFinancialData(text)) { 111 | el.style.padding = '10px'; 112 | el.style.fontSize = '24px'; 113 | el.style.fontWeight = 'bold'; 114 | } 115 | }); 116 | """) 117 | 118 | # Additional wait for text adjustments 119 | time.sleep(2) # Increased wait time 120 | 121 | # Get dimensions and ensure they're within limits 122 | total_height = driver.execute_script("return Math.max(document.documentElement.scrollHeight, document.body.scrollHeight);") 123 | total_width = driver.execute_script("return Math.max(document.documentElement.scrollWidth, document.body.scrollWidth);") 124 | 125 | final_width, final_height = ensure_size_within_limits(total_width, total_height) 126 | logger.info(f"Adjusted dimensions to {final_width}x{final_height} to stay within pixel limit") 127 | 128 | # Set final window size 129 | driver.set_window_size(final_width, final_height) 130 | time.sleep(1) 131 | 132 | # Capture the screenshot using our improved method 133 | screenshot_png = capture_full_page_screenshot(driver, url) 134 | driver.quit() 135 | 136 | # Process the image 137 | img = Image.open(io.BytesIO(screenshot_png)) 138 | 139 | # Convert to RGB and enhance readability 140 | if img.mode in ('RGBA', 'P'): 141 | img = img.convert('RGB') 142 | 143 | # Enhance image quality with specified values 144 | from PIL import ImageEnhance 145 | enhancer = ImageEnhance.Sharpness(img) 146 | img = enhancer.enhance(1.25) # Modified sharpness value 147 | enhancer = ImageEnhance.Contrast(img) 148 | img = enhancer.enhance(1.25) # Modified contrast value 149 | 150 | # Save with high quality 151 | output = io.BytesIO() 152 | img.save(output, format='JPEG', quality=100, optimize=True) # Maximum quality 153 | screenshot_data = output.getvalue() 154 | 155 | # Convert to base64 156 | base64_image = base64.b64encode(screenshot_data).decode('utf-8') 157 | 158 | vision_llm = configure_vision_model(provider) 159 | text_llm = configure_llm(provider) 160 | 161 | vision_query = generate_vision_query(text_llm, original_query) 162 | logger.info(f"Using vision query: {vision_query}") 163 | 164 | messages = [ 165 | { 166 | "role": "user", 167 | "content": [ 168 | {"type": "text", "text": vision_query}, 169 | { 170 | "type": "image_url", 171 | "image_url": { 172 | "url": f"data:image/jpeg;base64,{base64_image}", 173 | "detail": "high" 174 | } 175 | } 176 | ] 177 | } 178 | ] 179 | 180 | logger.info(f"Processing screenshot from {url} with vision model ({provider})") 181 | vision_response = vision_llm.invoke(messages) 182 | 183 | extracted_text = vision_response.content.strip() 184 | 185 | print("\n" + "="*80) 186 | print(f"Vision Model Description for {url}:") 187 | print("-"*80) 188 | print(extracted_text) 189 | print("="*80 + "\n") 190 | 191 | logger.info(f"Successfully processed content from {url}") 192 | return extracted_text 193 | 194 | except Exception as e: 195 | host_tracker .add_failed_host(url) 196 | logger.error(f"Error processing {url}: {str(e)}") 197 | return f"Error processing {url}: {str(e)}" -------------------------------------------------------------------------------- /tools/host_tracker.py: -------------------------------------------------------------------------------- 1 | from config.log import logger 2 | from urllib.parse import urlparse 3 | import os 4 | 5 | class HostTracker: 6 | def __init__(self, filename="HOSTS.txt"): 7 | self.filename = filename 8 | self.failed_hosts = set() 9 | self.load_failed_hosts() 10 | 11 | def load_failed_hosts(self): 12 | """Load failed hosts from file.""" 13 | try: 14 | if os.path.exists(self.filename): 15 | with open(self.filename, 'r') as f: 16 | self.failed_hosts = set(line.strip() for line in f if line.strip()) 17 | logger.info(f"Loaded {len(self.failed_hosts)} problematic hosts from {self.filename}") 18 | except Exception as e: 19 | logger.error(f"Error loading failed hosts: {str(e)}") 20 | self.failed_hosts = set() 21 | 22 | def add_failed_host(self, url: str): 23 | """Add a failed host to the tracking list.""" 24 | try: 25 | host = urlparse(url).netloc 26 | if host and host not in self.failed_hosts: 27 | self.failed_hosts.add(host) 28 | with open(self.filename, 'a') as f: 29 | f.write(f"{host}\n") 30 | logger.info(f"Added {host} to problematic hosts list") 31 | except Exception as e: 32 | logger.error(f"Error adding failed host: {str(e)}") 33 | 34 | def is_problematic_host(self, url: str) -> bool: 35 | """Check if a URL's host is in the problematic list.""" 36 | try: 37 | host = urlparse(url).netloc 38 | return host in self.failed_hosts 39 | except Exception: 40 | return False 41 | 42 | host_tracker = HostTracker() -------------------------------------------------------------------------------- /tools/size_limit.py: -------------------------------------------------------------------------------- 1 | import math 2 | 3 | def ensure_size_within_limits(width: int, height: int, max_pixels: int = 33177600) -> tuple: 4 | """Ensure dimensions are within the pixel limit while maintaining aspect ratio.""" 5 | total_pixels = width * height 6 | 7 | max_pixels = int(max_pixels * 0.9) 8 | 9 | if total_pixels <= max_pixels: 10 | return width, height 11 | 12 | # Calculate scaling factor to fit within limit 13 | scale = math.sqrt(max_pixels / total_pixels) 14 | new_width = int(width * scale) 15 | new_height = int(height * scale) 16 | 17 | if new_width * new_height > max_pixels: 18 | scale *= 0.95 19 | new_width = int(width * scale) 20 | new_height = int(height * scale) 21 | 22 | return new_width, new_height -------------------------------------------------------------------------------- /tools/split_doc.py: -------------------------------------------------------------------------------- 1 | from langchain.text_splitter import RecursiveCharacterTextSplitter 2 | 3 | def split_documents(docs): 4 | text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=0) 5 | return text_splitter.split_documents(docs) -------------------------------------------------------------------------------- /tools/topic_into_sub.py: -------------------------------------------------------------------------------- 1 | from Model.invokemodel import invoke_model 2 | 3 | def decompose_topic_into_subtopics(llm, topic): 4 | decomposition_prompt = f"""You are a research assistant. 5 | You will be given a research topic. If the topic is broad or complex, break it down into a list of more specific subtopics or sub-questions that would help in researching it thoroughly. 6 | If the topic is simple or already focused, just return it as is. 7 | Format your response as a simple list with one subtopic per line. 8 | 9 | Topic: {topic} 10 | 11 | Subtopics:""" 12 | response = invoke_model(llm, decomposition_prompt) 13 | response_text = response.content 14 | subtopics = [line.strip("- ").strip() for line in response_text.split("\n") if line.strip()] 15 | subtopics = [s for s in subtopics if s and not s.lower().startswith(("subtopic", "topic"))] 16 | return subtopics if subtopics else [topic] -------------------------------------------------------------------------------- /tools/vision_query.py: -------------------------------------------------------------------------------- 1 | from Model.invokemodel import invoke_model 2 | from config.log import logger 3 | import re 4 | 5 | def generate_vision_query(llm, original_query: str) -> str: 6 | """Generate a focused vision query based on the original research question.""" 7 | prompt = f"""You are a tool assisting in generating natural and concise vision model queries. 8 | Your task is to transform research questions into specific and actionable prompts that guide the model in analyzing webpage screenshots. 9 | The queries should: 10 | - Start with "Describe the image in detail, focusing on". 11 | - Be natural, concise, and no longer than 15 words. 12 | - Highlight the most relevant information to answer the research question. 13 | - Avoid mechanical or vague phrases like "extract X from the image." 14 | 15 | Examples: 16 | Research question: What is the current Tesla stock price? 17 | Vision query: Describe the image in detail, focusing on the specific Tesla stock price. 18 | 19 | Research question: What are the iPhone 15 specs? 20 | Vision query: Describe the image in detail, focusing on the iPhone 15 specifications and features. 21 | 22 | Research question: How much does the latest MacBook Pro cost? 23 | Vision query: Describe the image in detail, focusing on the price of the latest MacBook Pro. 24 | 25 | Research question: {original_query} 26 | 27 | Vision query:""" 28 | 29 | try: 30 | response = invoke_model(llm, prompt) 31 | vision_query = response.content.strip() 32 | 33 | # Clean up and standardize the query 34 | vision_query = vision_query.replace('"', '').replace("'", '') 35 | if not vision_query.lower().startswith("describe the image"): 36 | vision_query = f"Describe the image in detail, focusing on {vision_query}" 37 | 38 | # Remove mechanical phrases 39 | vision_query = vision_query.replace("extract from the image", "") 40 | vision_query = vision_query.replace("from the image", "") 41 | vision_query = re.sub(r'\s+', ' ', vision_query).strip() 42 | 43 | # Ensure it ends properly 44 | if vision_query.endswith("focusing on"): 45 | vision_query = vision_query[:-11].strip() 46 | 47 | return vision_query 48 | except Exception as e: 49 | logger.error(f"Error generating vision query: {str(e)}") 50 | return "Describe the image in detail, focusing on the main content and key information." 51 | --------------------------------------------------------------------------------