feat: Add thought parts from litellm response

[saravanabalagi]

Deepseek and Qwen3 models produce thought / reasoning in

<think>Okay, the user said "hi". I need to...</think>
Hello! How can I assist you

Litellm already parses these think tags and the response messages contain these thought parts in the reasoning_content field. This can be parsed and added as ADK text part with thought True.

Further, this should automatically be displayed in UI and be treated the same as that of Gemini models

---

Fixes #1788 and #2982

[google-cla]

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.

[gemini-code-assist]

Summary of Changes

Hello @saravanabalagi, 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 enhances the system's ability to process and display the internal reasoning or 'thought' processes of certain AI models, specifically Deepseek and Qwen3, when their responses are routed through LiteLLM. By extracting the reasoning_content and marking it as a distinct 'thought' part, the change aims to provide users with greater transparency into how these models arrive at their answers, ensuring a consistent user experience with other models like Gemini.

Highlights

• Thought Part Integration: Implemented the parsing and integration of 'thought' or 'reasoning_content' from LiteLLM responses into the ADK's types.Part structure, explicitly marking them with thought=True.
• Enhanced Model Output Display: Ensured that the reasoning steps from models like Deepseek and Qwen3 (when processed via LiteLLM) are captured and prepared for display in the UI, aligning with how Gemini models' thought processes are presented.
• Bug Fixes: This change addresses and fixes issues QUESTION: It seems that LiteLLM does not handle the reasoning_content of the model. #1788 and Thought / Reasoning messages missing in events #2982, likely related to the display or handling of model reasoning.

Using Gemini Code Assist

The 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 /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description

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 .gemini/ folder in the base of the repository. Detailed instructions can be found here.

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

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[adk-bot]

Hello @saravanabalagi, thank you for your contribution!

Before we can review your PR, please address the following items from our contribution guidelines:

• Sign our Contributor License Agreement: It appears you have not yet signed the CLA. You can do so at https://cla.developers.google.com/.
• Testing Plan: Please include a testing plan section in your PR description to describe how you tested your changes.
• Unit Tests: Please add or update unit tests for your changes and include a summary of passed pytest results in your PR description.

You can find more details in our contribution guidelines.

Response from ADK Triaging Agent

[gemini-code-assist]

Code Review

This pull request introduces support for 'thought' parts from litellm responses by processing the reasoning_content field. The implementation is correct, but I have a suggestion to make the code more concise and efficient. More importantly, the new functionality is not covered by unit tests. I've added a comment highlighting the need to add a test case to ensure the new logic is working as expected and to prevent future regressions.

[bastiandg]

This will probably work for most llms. It doesn't work on gpt-5 because thoughts are only emitted on the new responses api.

Hopefully this will be merged soon on litellms side, so this will also work for gpt-5: BerriAI/litellm#14765

[abelatyou]

Hi @saravanabalagi: we are also encountering the same issue and noticed that your changes would only work for non-sse responses

I think we also need to make some changes to generate_content_async and _model_response_to_chunk to return and process a new ThoughtChunk so that thoughts are also passed downstream