feat: add tool call release decisions#6165
Open
maxpetrusenkoagent wants to merge 6 commits into
Open
Conversation
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (4)
🚧 Files skipped from review as they are similar to previous changes (1)
📝 WalkthroughWalkthroughIntroduces a ChangesToolCallDecision release-control mediation
Sequence Diagram(s)sequenceDiagram
participant Agent as AgentExecutor
participant Hook as BeforeToolCallHook
participant Resolver as resolve_tool_call_decision
participant Decision as ToolCallDecision
participant Tool as ToolFunction
Agent->>Hook: call(ToolCallHookContext)
Hook-->>Agent: ToolCallDecision | bool | None
Agent->>Resolver: resolve_tool_call_decision(hook_result, tool_name)
alt result is ToolCallDecision
Resolver->>Decision: should_execute?
alt PROCEED
Decision-->>Resolver: True
Resolver-->>Agent: None
Agent->>Tool: execute()
else NEEDS_REVIEW or SILENCE
Decision-->>Resolver: False
Resolver->>Decision: block_message(tool_name)
Decision-->>Resolver: formatted block message
Resolver-->>Agent: block message (str)
Agent-->>Agent: result = block message (skip tool)
end
else result is False
Resolver-->>Agent: generic block message
Agent-->>Agent: result = block message (skip tool)
else result is None or True
Resolver-->>Agent: None
Agent->>Tool: execute()
end
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Suggested reviewers
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Author
|
@coderabbitai review |
✅ Action performedReview finished.
|
Author
|
Review-fix update:
Remaining external blocker: repository GitHub Actions are still |
|
Tip For best results, initiate chat on the files or code changes.
For the GitHub Actions blocker, you may want to ping a maintainer directly (e.g., via a comment tagging them) to request approval of the workflow runs. Once CI passes, the PR should be in a good state for merge consideration. Is there anything else you'd like me to look into or review on this PR? |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (1)
lib/crewai/src/crewai/hooks/tool_call_decision.py (1)
41-42: ⚡ Quick win
frozen=Trueis undermined by aliasing mutablereview_context.On Line 41 / Line 54 / Line 67, non-empty dicts are stored by reference. External mutation after creation can still change decision payloads.
Suggested patch
`@dataclass`(frozen=True) class ToolCallDecision: @@ + `@staticmethod` + def _copy_context( + review_context: dict[str, Any] | None, + ) -> dict[str, Any]: + return dict(review_context) if review_context is not None else {} + `@classmethod` def proceed( @@ return cls( decision=ToolCallDecisionType.PROCEED, reason=reason, - review_context=review_context or {}, + review_context=cls._copy_context(review_context), ) @@ return cls( decision=ToolCallDecisionType.NEEDS_REVIEW, reason=reason, - review_context=review_context or {}, + review_context=cls._copy_context(review_context), ) @@ return cls( decision=ToolCallDecisionType.SILENCE, reason=reason, - review_context=review_context or {}, + review_context=cls._copy_context(review_context), )Also applies to: 54-55, 67-68
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@lib/crewai/src/crewai/hooks/tool_call_decision.py` around lines 41 - 42, The frozen=True dataclass is storing mutable dict objects by reference, allowing external code to mutate the review_context after object creation despite the frozen constraint. To fix this, create defensive copies of the review_context dict in all three locations where it appears (at lines 41-42, 54-55, and 67-68) by replacing review_context or {} with dict(review_context or {}) to ensure a copy is stored rather than a reference to the original mutable object.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@docs/v1.13.0/en/learn/tool-hooks.mdx`:
- Around line 28-55: The file docs/v1.13.0/en/learn/tool-hooks.mdx is part of a
frozen version snapshot documentation and should not be modified. Move all the
changes you made to this file (the before_hook function documentation and the
release_control example with ToolCallDecision) to the editable source
documentation location instead, typically found in the main docs folder without
the version number prefix (docs/en/learn/tool-hooks.mdx or similar). Keep the
docs/v1.13.0/ snapshot unchanged unless this is explicitly a [docs-freeze]
release-cut PR.
In `@lib/crewai/src/crewai/hooks/tool_hooks.py`:
- Around line 120-133: The function resolve_tool_call_decision currently fails
open by returning None at the end if the hook_result does not match any expected
types (ToolCallDecision instance or boolean False), which silently allows tool
execution for unexpected return values. Replace the final return None statement
with a blocking response that raises an exception or returns a block message
indicating an invalid hook return type, ensuring that any unexpected hook return
values fail closed by preventing tool execution rather than allowing it.
---
Nitpick comments:
In `@lib/crewai/src/crewai/hooks/tool_call_decision.py`:
- Around line 41-42: The frozen=True dataclass is storing mutable dict objects
by reference, allowing external code to mutate the review_context after object
creation despite the frozen constraint. To fix this, create defensive copies of
the review_context dict in all three locations where it appears (at lines 41-42,
54-55, and 67-68) by replacing review_context or {} with dict(review_context or
{}) to ensure a copy is stored rather than a reference to the original mutable
object.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro Plus
Run ID: 950aee1b-78de-4cdd-8efd-26750a092098
📒 Files selected for processing (12)
docs/v1.13.0/en/learn/tool-hooks.mdxlib/crewai/src/crewai/agents/crew_agent_executor.pylib/crewai/src/crewai/experimental/agent_executor.pylib/crewai/src/crewai/hooks/__init__.pylib/crewai/src/crewai/hooks/decorators.pylib/crewai/src/crewai/hooks/tool_call_decision.pylib/crewai/src/crewai/hooks/tool_hooks.pylib/crewai/src/crewai/hooks/types.pylib/crewai/src/crewai/hooks/wrappers.pylib/crewai/src/crewai/utilities/agent_utils.pylib/crewai/src/crewai/utilities/tool_utils.pylib/crewai/tests/hooks/test_tool_hooks.py
Comment on lines
+28
to
+55
| def before_hook(context: ToolCallHookContext) -> bool | ToolCallDecision | None: | ||
| # Return False to block execution with a generic message | ||
| # Return ToolCallDecision.needs_review(...) to pause for review | ||
| # Return ToolCallDecision.silence(...) to suppress execution | ||
| # Return True, None, or ToolCallDecision.proceed(...) to allow execution | ||
| ... | ||
| ``` | ||
|
|
||
| For runtime release-control, return a `ToolCallDecision` from a before-tool hook. | ||
| This lets a hook distinguish normal execution from calls that need review or should | ||
| be silenced before the tool runs: | ||
|
|
||
| ```python | ||
| from crewai.hooks import ToolCallDecision, ToolCallHookContext, before_tool_call | ||
|
|
||
| @before_tool_call | ||
| def release_control(context: ToolCallHookContext) -> ToolCallDecision | None: | ||
| if context.tool_name == "send_email": | ||
| return ToolCallDecision.needs_review( | ||
| reason="Outbound email requires human approval", | ||
| review_context={"input": context.tool_input}, | ||
| ) | ||
|
|
||
| if context.tool_name == "post_to_social_media": | ||
| return ToolCallDecision.silence(reason="Public posting is disabled") | ||
|
|
||
| return ToolCallDecision.proceed() | ||
| ``` |
There was a problem hiding this comment.
Do not modify frozen snapshot docs under docs/v*.
This file is under a versioned frozen-docs path. Move this update to the editable docs source and keep this snapshot unchanged (unless this is explicitly a [docs-freeze] release-cut PR).
As per coding guidelines, docs/v[0-9]*.[0-9]*.[0-9]*/**: Never modify files under docs/v*/ as they are frozen release snapshots, except in [docs-freeze] release-cut PRs.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@docs/v1.13.0/en/learn/tool-hooks.mdx` around lines 28 - 55, The file
docs/v1.13.0/en/learn/tool-hooks.mdx is part of a frozen version snapshot
documentation and should not be modified. Move all the changes you made to this
file (the before_hook function documentation and the release_control example
with ToolCallDecision) to the editable source documentation location instead,
typically found in the main docs folder without the version number prefix
(docs/en/learn/tool-hooks.mdx or similar). Keep the docs/v1.13.0/ snapshot
unchanged unless this is explicitly a [docs-freeze] release-cut PR.
Source: Coding guidelines
added 4 commits
June 18, 2026 14:00
….com/maxpetrusenkoagent/crewAI into hermes/oss-pr-2026-06-15-crewai-6025
…2026-06-18-crewai-6165
Author
|
Review-fix sweep update:
Verification run locally:
Confidence: high for the review-fix scope. Remaining external state may still depend on repository Actions approval for forked PR workflows. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
ToolCallDecisiontri-state release-control results for before-tool hooksCloses #6025
Test Plan
uv run pytest lib/crewai/tests/hooks/test_tool_hooks.py::TestToolHooksIntegration::test_needs_review_decision_blocks_react_tool_execution lib/crewai/tests/hooks/test_tool_hooks.py::TestToolHooksIntegration::test_tool_call_decision_factories -quv run pytest lib/crewai/tests/hooks/test_tool_hooks.py -quv run ruff check --diff lib/crewai/src/crewai/hooks/tool_call_decision.py lib/crewai/src/crewai/hooks/tool_hooks.py lib/crewai/src/crewai/hooks/types.py lib/crewai/src/crewai/hooks/decorators.py lib/crewai/src/crewai/hooks/wrappers.py lib/crewai/src/crewai/agents/crew_agent_executor.py lib/crewai/src/crewai/experimental/agent_executor.py lib/crewai/src/crewai/utilities/agent_utils.py lib/crewai/src/crewai/utilities/tool_utils.pyuv run ruff format --check lib/crewai/src/crewai/hooks/tool_call_decision.py lib/crewai/src/crewai/hooks/tool_hooks.py lib/crewai/src/crewai/hooks/types.py lib/crewai/src/crewai/hooks/decorators.py lib/crewai/src/crewai/hooks/wrappers.py lib/crewai/src/crewai/agents/crew_agent_executor.py lib/crewai/src/crewai/experimental/agent_executor.py lib/crewai/src/crewai/utilities/agent_utils.py lib/crewai/src/crewai/utilities/tool_utils.py lib/crewai/tests/hooks/test_tool_hooks.pygit diff --checkSummary by CodeRabbit
New Features
ToolCallDecisionfor release control, enabling PROCEED, NEEDS_REVIEW, or SILENCE decisions with optional reasons and review context.Refactor
Tests
Documentation