fix(agents): show reasoning effort button for agent session reasoning models

[DeJeune]

What this PR does

Before this PR:

When selecting a reasoning model (e.g., Claude with extended thinking, o3) in an agent session, the reasoning effort / thinking button was invisible — users could not configure thinking intensity.

After this PR:

The reasoning effort button correctly appears for reasoning models in agent sessions, matching the behavior in normal chat.

Why we need it and why it was done in this way

Root cause: InputbarTools internally called useAssistant(assistantId) to resolve the assistant and model. For agent sessions, the assistantId is an agent ID that doesn't exist in the Redux state.assistants.assistants array. This caused:

1. assistant resolved to undefined (cast as Assistant)
2. model fell back to the global defaultModel (e.g., gpt-4o)
3. The thinking tool's condition: isReasoningModel(model) returned false
4. The button was filtered out of availableTools entirely

The fix: Pass assistant and model as props from the caller (which already has the correct data) instead of having InputbarTools fetch them internally via useAssistant.

The following tradeoffs were made:

• Removes useAssistant from InputbarTools, making it a controlled component that relies on the caller to provide assistant/model data. This is a minor interface change but makes the component more predictable and reusable.

The following alternatives were considered:

1. Add modelOverride prop as a fallback — Would keep useAssistant as the primary source and only override when needed. Rejected because it's a band-aid that doesn't address the root cause: useAssistant is being called with an ID that fundamentally cannot exist in Redux for agent sessions.

2. Dispatch agent assistant stub into Redux store — Would make useAssistant(agentId) find the assistant naturally. Rejected because:

   • The stub would appear in the assistants sidebar UI (no type-based filtering exists)
   • It would be persisted by redux-persist, polluting storage
   • useAssistant's useEffect would sync reasoning effort settings back to Redux, conflicting with agent session's local state management
   • Requires cleanup on unmount with potential race conditions
   • Redux data model changes are currently blocked until v2.0.0

3. Pass props from caller (chosen approach) — The simplest fix. Both callers (Inputbar and AgentSessionInputbar) already have the correct assistant and model. Three files changed, ~9 lines added / ~7 removed.

Breaking changes

None. Internal component interface change only — InputbarTools is not exported as a public API.

Special notes for your reviewer

The assistantId prop was removed from InputbarToolsNewProps since assistant.id provides the same information. No external consumers exist.

Checklist

• PR: The PR description is expressive enough and will help future contributors
• Code: Write code that humans can understand and Keep it simple
• Refactor: You have left the code cleaner than you found it (Boy Scout Rule)
• Upgrade: Impact of this change on upgrade flows was considered and addressed if required
• Documentation: A user-guide update was considered and is present (link) or not required. Check this only when the PR introduces or changes a user-facing feature or behavior.
• Self-review: I have reviewed my own code (e.g., via /gh-pr-review, gh pr diff, or GitHub UI) before requesting review from others

Release note

fix: reasoning effort button now appears correctly when using reasoning models in agent sessions

[cherry-ai-bot]

The latest update addresses the non-null assertion concern cleanly. AgentSessionInputbar now passes assistant.model without forcing !, and InputbarTools accepts an optional model override while falling back to assistant.model internally. That keeps the original fix intact for agent-session reasoning models without introducing a new runtime hazard. I do not see any remaining blocking issues in the current patch.