Modify AI class to support alternative LLMs/Models

sweep-ai[bot] commented 1 year ago

Description

This PR modifies the AI class in gpt_engineer/ai.py to support alternative LLMs/Models such as LiteLLM and Llama API, in addition to the existing support for GPT-4 and GPT-3.5. It also updates the documentation to reflect these changes.

Summary of Changes

Modified the create_chat_model function to handle LiteLLM and Llama API. The function now creates instances of the appropriate chat model based on the given model name and temperature.
Modified the get_tokenizer function to handle LiteLLM and Llama API. The function now returns the appropriate tokenizer for each model.
Added new functions to the AI class to handle specific features of LiteLLM and Llama API, if applicable.
Updated the documentation in docs/intro/quick_overview.md to mention the support for LiteLLM and Llama API in the AI class.
Updated the API reference in docs/api_reference.rst to include the new functions added for LiteLLM and Llama API.

Please review and merge this PR to enable support for alternative LLMs/Models in the AI class.

Fixes #1.

To checkout this PR branch, run the following command in your terminal:

git checkout sweep/modify-ai-class

To get Sweep to edit this pull request, leave a comment below or in the code. Leaving a comment in the code will only modify the file but commenting below can change the entire PR.

jontstaz commented 1 year ago

The implementation is not correct. Please refer to the LangChain documentation: https://python.langchain.com/docs/

Specifically the LiteLLM section in the docs: https://python.langchain.com/docs/integrations/chat/litellm

sweep-ai[bot] commented 1 year ago

The implementation is not correct. Please refer to the LangChain documentation: https://python.langchain.com/docs/

Specifically the LiteLLM section in the docs: https://python.langchain.com/docs/integrations/chat/litellm Hi @jontstaz,

I decided to make the following changes:

File Path Proposed Changes

gpt_engineer/ai.py Modify gpt_engineer/ai.py with contents:
* In the create_chat_model function, replace the current implementation for creating an instance of LiteLLMChatModel with the correct implementation as per the LangChain documentation. The LiteLLMChatModel should be initialized with the model parameter set to the path of the LiteLLM model file and the temperature parameter set to the desired temperature.
• In the get_tokenizer function, replace the current implementation for creating an instance of LiteLLMTokenizer with the correct implementation as per the LangChain documentation. The LiteLLMTokenizer should be initialized with the model parameter set to the path of the LiteLLM model file.

File Path	Proposed Changes
`gpt_engineer/ai.py`	Modify gpt_engineer/ai.py with contents: * In the `create_chat_model` function, replace the current implementation for creating an instance of `LiteLLMChatModel` with the correct implementation as per the LangChain documentation. The `LiteLLMChatModel` should be initialized with the `model` parameter set to the path of the LiteLLM model file and the `temperature` parameter set to the desired temperature. • In the `get_tokenizer` function, replace the current implementation for creating an instance of `LiteLLMTokenizer` with the correct implementation as per the LangChain documentation. The `LiteLLMTokenizer` should be initialized with the `model` parameter set to the path of the LiteLLM model file.

krrishdholakia commented 12 months ago

Hey @jontstaz thanks for using litellm. Would love to invite you to our community

happy to create a dedicated support channel for you as well, to make sure we can solve any issues/feature requests you have.

sweep-ai[bot] commented 12 months ago