fix(agent_tool): wrap input_schema payload in ReAct prompt and propagate tool_choice to LiteLLM by ecanlar · Pull Request #5924 · google/adk-python

ecanlar · 2026-06-01T09:21:52Z

Summary

Fixes two issues that prevent Claude-family models from entering the ReAct tool-calling loop when used via LiteLLM inside a nested AgentTool:

AgentTool with input_schema — the serialized JSON payload sent as the first message causes Claude to interpret the request as already complete and respond directly without calling any tools.
tool_choice not propagated — llm_request.config.tool_config.function_calling_config.mode was not forwarded to LiteLLM's completion_args, so callers could not enforce tool use at the request level.

Changes

`src/google/adk/tools/agent_tool.py`

Wrap the serialized input_schema JSON in a natural-language instruction that explicitly asks the inner agent to use its available tools before producing a response. This keeps Claude in ReAct mode regardless of the message content format.

`src/google/adk/models/lite_llm.py`

Read llm_request.config.tool_config.function_calling_config.mode and map it to LiteLLM's tool_choice parameter:

ANY → "required"
NONE → "none"
AUTO → provider default (unchanged, key omitted from completion_args)

_get_completion_inputs now returns a 5-tuple (messages, tools, response_format, generation_params, tool_choice).

Unit Tests Added

`tests/unittests/tools/test_agent_tool.py`

test_run_async_no_input_schema_passes_request_unchanged: without input_schema, the content passed to the inner runner is args['request'] verbatim.
test_run_async_with_input_schema_wraps_in_natural_language: with input_schema, the text begins with "Process the following structured request", contains "Request:\n" followed by the JSON payload, and is not a bare JSON blob.
test_run_async_with_input_schema_text_not_raw_json: asserts the text does not start with {.

`tests/unittests/models/test_litellm.py`

test_get_completion_inputs_tool_choice_none_without_tool_config: tool_choice is None with no tool_config.
test_get_completion_inputs_tool_choice_required_for_any_mode: returns "required" for ANY mode.
test_get_completion_inputs_tool_choice_none_for_none_mode: returns "none" for NONE mode.
test_get_completion_inputs_tool_choice_none_for_auto_mode: returns None for AUTO mode.
test_generate_content_async_propagates_tool_choice_required: acompletion receives tool_choice="required" for ANY.
test_generate_content_async_propagates_tool_choice_none_mode: acompletion receives tool_choice="none" for NONE.
test_generate_content_async_omits_tool_choice_for_auto_mode: tool_choice key absent from completion_args for AUTO.
test_generate_content_async_omits_tool_choice_without_tool_config: tool_choice key absent when no tool_config.

Also updated all existing _get_completion_inputs call sites (10 occurrences) to unpack the new 5-tuple.

Pytest Results

tests/unittests/tools/test_agent_tool.py + tests/unittests/models/test_litellm.py
1 failed (pre-existing: test_custom_schema[GOOGLE_AI] — unrelated to this PR),
302 passed, 1 skipped in 2.98s

New tests: 11 passed (3 agent_tool + 8 litellm)

The 1 pre-existing failure (test_custom_schema[GOOGLE_AI]) is a pydantic.ValidationError that reproduces on the base branch before any of these changes and is unrelated to this fix.

Fixes fix: AgentTool with input_schema + LiteLLM/Claude skips inner agent tool calls #5926
Addresses Model Settings: Add Tool Choice ["auto", "required", "none"] #773 (expose tool_choice to callers via FunctionCallingConfig)
Related input_schema does not work, even with the simplest possible BaseModel class #1063 (fixed FunctionDeclaration description in v1.20, different issue)

…ol_choice to LiteLLM When AgentTool uses input_schema, the inner agent receives a raw JSON blob that causes Claude models to skip the tool-calling loop (ReAct). Fix by wrapping the payload in a natural-language instruction. Also propagate tool_config.function_calling_config.mode to LiteLLM's tool_choice parameter so callers can enforce tool_choice='required'. Addresses google#773. Fixes: google#5926

adk-bot · 2026-06-01T09:55:38Z

Response from ADK Triaging Agent

Hello @ecanlar, thank you for creating this PR!

We appreciate your contribution to fixing the LiteLLM/Claude AgentTool issue. To help our reviewers process this pull request more efficiently, please ensure it complies with our contribution guidelines.

Currently, the following items are missing:

Unit Tests: The PR modifies behavior in lite_llm.py and agent_tool.py, but does not add or update any unit tests (e.g., under tests/unittests/). Please add unit tests covering the new tool wrapping and tool_choice behavior.
Pytest Results Summary: Please run the tests and include a summary of the passed pytest or tox results in the PR description.
Logs/Screenshots: Please provide console logs or screenshots demonstrating that the inner agent successfully enters the ReAct loop and calls tools after applying the fix.

Providing this information helps ensure high code quality, prevents future regressions, and makes the review process smoother. Thank you!

…tool_choice propagation - Update existing _get_completion_inputs call sites to handle new 5-tuple return value (adds tool_choice as 5th element) - Add 3 tests for AgentTool.run_async: verifies message is passed verbatim without input_schema, and is wrapped in a natural-language instruction with input_schema (PR google#5924 fix) - Add 8 tests for LiteLLM tool_choice propagation: covers _get_completion_inputs returning correct tool_choice for ANY/NONE/AUTO modes and None when no tool_config; covers generate_content_async correctly including/omitting tool_choice in completion_args

- Fix: only apply the ReAct wrapper in agent_tool.py when output_schema is not set on the inner agent, preventing breaking of single-shot structured output mode - Add regression test test_run_async_with_input_and_output_schema_passes_raw_json documenting that raw JSON is passed when both input_schema and output_schema are set - Apply pre-commit formatting fixes (isort + pyink)

…st HTTP call Move the ValueError from module import time to the lazy _get_session() factory so that importing the package (e.g. pytest collection) does not fail in environments where GITHUB_TOKEN is not set.

ecanlar · 2026-06-05T13:19:05Z

Good catch — confirmed the guard was not in place. tool_choice and tools were derived from two independent config fields (config.tool_config and config.tools) with no cross-check, so a caller setting mode=ANY with no function_declarations would reach acompletion() with tools=None, tool_choice="required" — exactly the combination LiteLLM rejects.

Fixed in the latest commit:

_get_completion_inputs: coerce tool_choice to None when tools is falsy before returning.
generate_content_async: null tool_choice alongside tools inside the "functions" additional-arg branch.

Two new tests cover both paths; 10 tool_choice tests passing.

rohityan · 2026-06-05T18:26:40Z

Hi @ecanlar , Thank you for your contribution! We appreciate you taking the time to submit this pull request.

rohityan · 2026-06-05T18:26:47Z

/adk-pr-analyze

rohityan · 2026-06-09T18:23:35Z

Hi @ecanlar , Your PR has been received by the team and is currently under review. We will provide feedback as soon as we have an update to share.

rohityan · 2026-06-09T18:23:56Z

Hi @GWeale , can you please review this.

rohityan · 2026-06-11T23:47:45Z

Hi @ecanlar , can you please fix formatting errors.

LiteLLM rejects tool_choice="required"/"none" when the tools list is falsy. tool_choice and tools were derived from independent config fields with no cross-check, so a caller setting mode=ANY with no function_declarations would reach acompletion() with tools=None and tool_choice="required". Add a guard in _get_completion_inputs to coerce tool_choice to None when tools is falsy, and mirror the same coercion in generate_content_async after the "functions" additional-arg override that already nulls tools. Add two tests covering both paths; update four existing tool_choice tests to include a FunctionDeclaration so they remain valid under the new guard.

ecanlar · 2026-06-12T07:28:05Z

i fixed the formatting erros. @rohityan

ecanlar force-pushed the fix/agent-tool-input-schema-tool-choice-litellm branch from ded18ca to 6168c49 Compare June 1, 2026 09:51

adk-bot added models [Component] Issues related to model support tools [Component] This issue is related to tools labels Jun 1, 2026

surajksharma07 mentioned this pull request Jun 1, 2026

fix: AgentTool with input_schema + LiteLLM/Claude skips inner agent tool calls #5926

Open

rohityan self-assigned this Jun 1, 2026

rohityan and others added 2 commits June 1, 2026 19:11

Merge branch 'main' into fix/agent-tool-input-schema-tool-choice-litellm

fd0073a

ecanlar force-pushed the fix/agent-tool-input-schema-tool-choice-litellm branch from 23d582a to 67d087a Compare June 2, 2026 09:44

ecanlar and others added 2 commits June 3, 2026 08:26

Merge branch 'main' into fix/agent-tool-input-schema-tool-choice-litellm

8bfe2d0

fix(adk_issue_monitoring_agent): defer GITHUB_TOKEN validation to fir…

7c23d83

…st HTTP call Move the ValueError from module import time to the lazy _get_session() factory so that importing the package (e.g. pytest collection) does not fail in environments where GITHUB_TOKEN is not set.

ecanlar force-pushed the fix/agent-tool-input-schema-tool-choice-litellm branch from f8c61e3 to 7c23d83 Compare June 5, 2026 11:05

Merge branch 'main' into fix/agent-tool-input-schema-tool-choice-litellm

d2a85fa

rohityan added request clarification [Status] The maintainer need clarification or more information from the author and removed request clarification [Status] The maintainer need clarification or more information from the author labels Jun 5, 2026

rohityan added 2 commits June 9, 2026 10:58

Merge branch 'main' into fix/agent-tool-input-schema-tool-choice-litellm

04be4bb

Merge branch 'main' into fix/agent-tool-input-schema-tool-choice-litellm

3f004cf

rohityan added the needs review [Status] The PR/issue is awaiting review from the maintainer label Jun 9, 2026

rohityan requested a review from GWeale June 9, 2026 18:23

Merge branch 'main' into fix/agent-tool-input-schema-tool-choice-litellm

1edc39e

rohityan added request clarification [Status] The maintainer need clarification or more information from the author and removed needs review [Status] The PR/issue is awaiting review from the maintainer labels Jun 11, 2026

rohityan removed the request for review from GWeale June 11, 2026 23:47

Merge branch 'main' into fix/agent-tool-input-schema-tool-choice-litellm

f7d0b35

rohityan requested a review from GWeale June 11, 2026 23:57

rohityan added needs review [Status] The PR/issue is awaiting review from the maintainer and removed request clarification [Status] The maintainer need clarification or more information from the author labels Jun 11, 2026

ecanlar and others added 2 commits June 12, 2026 07:40

Merge branch 'main' into fix/agent-tool-input-schema-tool-choice-litellm

44c745b

ecanlar force-pushed the fix/agent-tool-input-schema-tool-choice-litellm branch from b680ca5 to 9cd8f31 Compare June 12, 2026 07:26

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(agent_tool): wrap input_schema payload in ReAct prompt and propagate tool_choice to LiteLLM#5924

fix(agent_tool): wrap input_schema payload in ReAct prompt and propagate tool_choice to LiteLLM#5924
ecanlar wants to merge 13 commits into
google:mainfrom
ecanlar:fix/agent-tool-input-schema-tool-choice-litellm

ecanlar commented Jun 1, 2026 •

edited

Loading

Uh oh!

adk-bot commented Jun 1, 2026

Uh oh!

ecanlar commented Jun 5, 2026

Uh oh!

rohityan commented Jun 5, 2026 •

edited

Loading

Uh oh!

rohityan commented Jun 5, 2026

Uh oh!

rohityan commented Jun 9, 2026

Uh oh!

rohityan commented Jun 9, 2026

Uh oh!

rohityan commented Jun 11, 2026 •

edited

Loading

Uh oh!

ecanlar commented Jun 12, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

ecanlar commented Jun 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Changes

src/google/adk/tools/agent_tool.py

src/google/adk/models/lite_llm.py

Unit Tests Added

tests/unittests/tools/test_agent_tool.py

tests/unittests/models/test_litellm.py

Pytest Results

Related

Uh oh!

adk-bot commented Jun 1, 2026

Uh oh!

ecanlar commented Jun 5, 2026

Uh oh!

rohityan commented Jun 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rohityan commented Jun 5, 2026

Uh oh!

rohityan commented Jun 9, 2026

Uh oh!

rohityan commented Jun 9, 2026

Uh oh!

rohityan commented Jun 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ecanlar commented Jun 12, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

ecanlar commented Jun 1, 2026 •

edited

Loading

`src/google/adk/tools/agent_tool.py`

`src/google/adk/models/lite_llm.py`

`tests/unittests/tools/test_agent_tool.py`

`tests/unittests/models/test_litellm.py`

rohityan commented Jun 5, 2026 •

edited

Loading

rohityan commented Jun 11, 2026 •

edited

Loading