fix(models): add Azure and other common provider patterns to LiteLLM supported_models #4434
Conversation
|
Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA). View this failed invocation of the CLA check for more information. For the most up to date status, view the checks section at the bottom of the pull request. |
Summary of ChangesHello @anmolg1997, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! This pull request significantly enhances the LiteLLM integration by expanding the list of supported model patterns. By registering a wider array of provider prefixes, it resolves critical model resolution failures, ensuring that various LLM providers can be correctly identified and utilized within the system. This change improves the robustness and flexibility of the LLM framework, particularly for evaluation and registry-dependent functionalities. Highlights
🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console. Changelog
Activity
Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request adds support for several new LiteLLM providers by including their model name patterns in the supported_models method of the LiteLlm class. This change fixes model resolution failures for various providers like Azure OpenAI, AWS Bedrock, and others. The added patterns ensure that the LLMRegistry.resolve() method can correctly identify and resolve models from these providers. The changes include adding regular expression patterns for each provider to the supported_models list.
| # For Azure OpenAI models (e.g., "azure/gpt-4o") | ||
| r"azure/.*", | ||
| # For Azure AI models (e.g., "azure_ai/command-r-plus") | ||
| r"azure_ai/.*", | ||
| # For Groq models via Groq API (e.g., "groq/llama3-70b-8192") | ||
| r"groq/.*", | ||
| # For Anthropic models (e.g., "anthropic/claude-3-opus-20240229") | ||
| r"anthropic/.*", | ||
| # For AWS Bedrock models (e.g., "bedrock/anthropic.claude-3-sonnet") | ||
| r"bedrock/.*", | ||
| # For Ollama models (e.g., "ollama/llama3") | ||
| r"ollama/.*", | ||
| # For Ollama chat models (e.g., "ollama_chat/llama3") | ||
| r"ollama_chat/.*", | ||
| # For Together AI models (e.g., "together_ai/meta-llama/Llama-3-70b") | ||
| r"together_ai/.*", | ||
| # For Vertex AI non-Gemini models (e.g., "vertex_ai/claude-3-sonnet") | ||
| r"vertex_ai/.*", | ||
| # For Mistral AI models (e.g., "mistral/mistral-large-latest") | ||
| r"mistral/.*", | ||
| # For DeepSeek models (e.g., "deepseek/deepseek-chat") | ||
| r"deepseek/.*", | ||
| # For Fireworks AI models (e.g., "fireworks_ai/llama-v3-70b") | ||
| r"fireworks_ai/.*", | ||
| # For Cohere models (e.g., "cohere/command-r-plus") | ||
| r"cohere/.*", | ||
| # For Databricks models (e.g., "databricks/dbrx-instruct") | ||
| r"databricks/.*", | ||
| # For AI21 models (e.g., "ai21/jamba-1.5-large") | ||
| r"ai21/.*", |
There was a problem hiding this comment.
The regular expression patterns for the supported models are added sequentially. Consider grouping them by category (e.g., Azure, AWS, etc.) to improve readability and maintainability.
…supported_models
The LiteLLM integration only registered three provider prefixes
(openai, groq, anthropic) in supported_models(), causing the model
registry to fail resolution for widely-used providers like Azure
OpenAI, AWS Bedrock, Ollama, Vertex AI, and others.
This meant that `registry.resolve("azure/gpt-4o")` raised
"Model azure/gpt-4o not found", breaking features like
LLM-as-a-Judge evaluation for Azure users.
Added the following provider patterns:
- azure, azure_ai (Azure OpenAI / Azure AI)
- bedrock (AWS Bedrock)
- ollama, ollama_chat (Ollama)
- together_ai (Together AI)
- vertex_ai (Vertex AI non-Gemini models)
- mistral (Mistral AI)
- deepseek (DeepSeek)
- fireworks_ai (Fireworks AI)
- cohere (Cohere)
- databricks (Databricks)
- ai21 (AI21)
Fixes google#4325
anmolg1997
force-pushed
the
fix/add-azure-litellm-supported-models
branch
from
8e4a3d0 to
55b825b
Compare
|
Response from ADK Triaging Agent Hello @anmolg1997, thank you for your contribution! Before we can merge this pull request, you'll need to sign the Contributor License Agreement (CLA). You can do so here: https://cla.developers.google.com/ Thank you! |
|
I have signed the CLA - @googlebot please verify. |
anmolg1997
force-pushed
the
fix/add-azure-litellm-supported-models
branch
from
55b825b to
533733b
Compare
2 participants
Link to Issue or Description of Change
Problem:
supported_models()inlite_llm.pyonly returns three provider patterns (openai/.*,groq/.*,anthropic/.*). When the model registry tries to resolve a model likeazure/gpt-4o, it doesn't match any pattern and throwsModel azure/gpt-4o not found. This breaks LLM-as-a-Judge eval and any other code path that usesLLMRegistry.resolve()for provider-prefixed models outside those three.Solution:
Added 13 additional well-known LiteLLM provider prefixes to
supported_models():azure,azure_ai,bedrock,ollama,ollama_chat,together_ai,vertex_ai,mistral,deepseek,fireworks_ai,cohere,databricks,ai21. All patterns sourced from https://docs.litellm.ai/docs/providers.Testing Plan
Unit Tests:
Also verified end-to-end that
LLMRegistry.resolve("azure/gpt-4o")now correctly resolves toLiteLlm, and same for bedrock, ollama, vertex_ai etc.