* add client interface, response protocol, and move code into openai client class
* add ability to register custom client
* tidy up code
* adding checks and errors, and more unit tests
* remove code
* fix error msg
* add use_docer False in notebook
* better error message
* add another example to custom model notebook
* rename and have register_client take model name too
* make Client protocol and remove inheritance
* renames
* update notebook
* add link
* rename and more error checking for registered agents
* adding message retrieval to client protocol for more flexible response
* fix failing openai test
* api_type req made model_client_cls requirement
* notebook cleanup and added blog
* remove raise error if client list is empty - client list will never be empty it will have placeholders
* rename Client -> ModelClient
* add forgotten file
* fix test by fetching internal client
* Update autogen/oai/client.py
Co-authored-by: Eric Zhu <ekzhu@users.noreply.github.com>
* don't add retrieval function to cache
* added placeholder cllient class during initial client init, and rewrote registration
* fix spelling
* Update autogen/agentchat/conversable_agent.py
Co-authored-by: Chi Wang <wang.chi@microsoft.com>
* type hints, small fixes, docstr comment
* fix api type checking
---------
Co-authored-by: Eric Zhu <ekzhu@users.noreply.github.com>
Co-authored-by: Chi Wang <wang.chi@microsoft.com>
* implement redis cache mode, if redis_url is set in the llm_config then
it will try to use this. also adds a test to validate both the existing
and the redis cache behavior.
* PR feedback, add unit tests
* more PR feedback, move the new style cache to a context manager
* Update agent_chat.md
* more PR feedback, remove tests from contrib and have them run with the normal jobs
* doc
* updated
* Update website/docs/Use-Cases/agent_chat.md
Co-authored-by: Chi Wang <wang.chi@microsoft.com>
* update docs
* update docs; let openaiwrapper to use cache object
* typo
* Update website/docs/Use-Cases/enhanced_inference.md
Co-authored-by: Chi Wang <wang.chi@microsoft.com>
* save previous client cache and reset it after send/a_send
* a_run_chat
---------
Co-authored-by: Vijay Ramesh <vijay@regrello.com>
Co-authored-by: Eric Zhu <ekzhu@users.noreply.github.com>
Co-authored-by: Chi Wang <wang.chi@microsoft.com>
* Switched to AzureOpenAI for api_type=="azure"
* Setting AzureOpenAI to empty object if no `openai`
* extra_ and openai_ kwargs
* test_client, support for Azure and "gpt-35-turbo-instruct"
* instruct/azure model in test_client_stream
* generalize aoai support (#1)
* generalize aoai support
* Null check, fixing tests
* cleanup test
---------
Co-authored-by: Maxim Saplin <smaxmail@gmail.com>
* Returning back model names for instruct
* process model in create
* None check
---------
Co-authored-by: Chi Wang <wang.chi@microsoft.com>
* Added the ability to add tags to the OAI_CONFIG_LIST, and filter on them.
* Update openai_utils.py
Co-authored-by: Chi Wang <wang.chi@microsoft.com>
---------
Co-authored-by: Chi Wang <wang.chi@microsoft.com>
* Throw an explicit and proper error when someone asks to load the OAI_CONFIG_LIST, and it is missing.
* Updated to use pytest.raises. Added docstring. Updated some tests to not try to load the config_list when skipping OAI tests.
* Fixed wrong indentation in config_list_from_json, and updated test_utils to work with non-empty lists.
* Read key location from global constants.
* Added missingpath.
* Moved config_list_from_json to inside a skip check.
* Improve config_list_from_json
The `env_or_file` variabe can point to an
environment variable of file path.
* Update autogen/oai/openai_utils.py
Co-authored-by: Chi Wang <wang.chi@microsoft.com>
* Use "with" to open config file
* Remove unused.
* Remove accidental added file
---------
Co-authored-by: Eric Zhu <ekzhu@users.noreply.github.com>
Co-authored-by: Chi Wang <wang.chi@microsoft.com>
* fix for #762 ,get_config_list/oia assetion, docs, test
* Improved docstrings
* test_utils using fake data and temp files
* "Black" formatting applied
* Fix build (KEY_LOC and OAI_CONFIG_LIST as consts in test_client)
* Ramping up openai_utils coverage
* Missing parameter doc.
---------
Co-authored-by: Eric Zhu <ekzhu@users.noreply.github.com>
* add function decorator to converasble agent
* polishing
* polishing
* added function decorator to the notebook with async function calls
* added support for return type hint and JSON encoding of returned value if needed
* polishing
* polishing
* refactored async case
* Python 3.8 support added
* polishing
* polishing
* missing docs added
* refacotring and changes as requested
* getLogger
* documentation added
* test fix
* test fix
* added testing of agentchat_function_call_currency_calculator.ipynb to test_notebook.py
* added support for Pydantic parameters in function decorator
* polishing
* Update website/docs/Use-Cases/agent_chat.md
Co-authored-by: Li Jiang <bnujli@gmail.com>
* Update website/docs/Use-Cases/agent_chat.md
Co-authored-by: Li Jiang <bnujli@gmail.com>
* fixes problem with logprob parameter in openai.types.chat.chat_completion.Choice added by openai version 1.5.0
* get 100% code coverage on code added
* updated docs
* default values added to JSON schema
* serialization using json.dump() add for values not string or BaseModel
* added limit to openai version because of breaking changes in 1.5.0
* added line-by-line comments in docs to explain the process
* polishing
---------
Co-authored-by: Eric Zhu <ekzhu@users.noreply.github.com>
Co-authored-by: Li Jiang <bnujli@gmail.com>
* bump version to 0.2.1
* remove example link
* update
---------
Co-authored-by: Qingyun Wu <qingyun.wu@psu.edu>
Co-authored-by: kevin666aa <yrwu000627@gmail.com>
* Refactor GPTAssistantAgent constructor to handle
instructions and overwrite_instructions flag
- Ensure that `system_message` is always consistent with `instructions`
- Ensure provided instructions are always used
- Add option to permanently modify the instructions of the assistant
* Improve default behavior
* Add a test; add method to delete assistant
* Add a new test for overwriting instructions
* Add test case for when no instructions are given for existing assistant
* Add pytest markers to test_gpt_assistant.py
* add test in workflow
* update
* fix test_client_stream
* comment out test_hierarchy_
---------
Co-authored-by: Chi Wang <wang.chi@microsoft.com>
Co-authored-by: kevin666aa <yrwu000627@gmail.com>
* FORMATTING
* UPDATE - OAI __init__.py
* ruff
* ADD - notebook covering oai API configuration options and their different purposes
* ADD openai util updates so that the function just assumes the same environment variable name for all models, also added functionality for adding API configurations like api_base etc.
* ADD - updates to config_list_from_dotenv and tests for openai_util testing, update example notebook
* UPDATE - added working config_list_from_dotenv() with passing tests, and updated notebook
* UPDATE - code and tests to potentially get around the window build permission error, used different method of producing temporary files
---------
Co-authored-by: Ward <award40@LAMU0CLP74YXVX6.uhc.com>
