fix: [#652] warn when plan agent model refuses to tool-call

[anandgupta42]

What does this PR do?

Warns the user when the plan agent finishes a step with finish_reason=stop and zero tool calls in the session — catching models that refuse to tool-call in plan mode (observed on ollama-cloud/qwen3-coder-next, occasionally openai/gpt-5.4).

Signals emitted (one-shot per session):

• log.warn with session/model/provider context
• plan_no_tool_generation telemetry event (new type) — per-model abandonment tracking
• Synthetic text part in the TUI with a /model recommendation

Guards:

• Fires at most once per session (planNoToolWarningEmitted flag)
• Only for agent="plan", finish_reason="stop", sessionToolCallsMade===0
• Other finish reasons (tool-calls, length, error) don't trigger
• Other agents (builder, analyst, explore, general) don't trigger

The text part is synthetic: true — shown to the user in the TUI but excluded from the transcript replayed to the LLM next turn (existing filter at prompt.ts:648, 795).

Type of change

• Bug fix
• New feature
• Test coverage
• Documentation
• Refactoring
• Infrastructure

Issue for this PR

Closes #652

How did you verify your code works?

• Typecheck: clean on changed files (existing config.ts errors unchanged)
• Marker guard: passes (altimate_change markers around all additions)
• 6 new unit tests in processor.test.ts covering:
  • happy path (fires on plan + stop + no-tools)
  • negative: already tool-called this session
  • negative: finish_reason=tool-calls
  • negative: non-plan agents (builder/analyst/general/explore)
  • negative: finish_reason=length/error/content-filter/unknown
  • one-shot flag (doesn't re-fire)
• All 46 processor tests pass

Checklist

• Typecheck passes
• Marker guard passes
• No breaking changes (additive telemetry event + synthetic warning)
• Telemetry event typed and documented
• Unit tests for all detection branches

---

Summary by cubic

Warns when the plan agent stops without any tool calls so users aren’t stuck and we can track model refusals. Addresses #652.

• Bug Fixes
  • Trigger: agent="plan", finish_reason=stop, and no session tool calls.
  • Actions: emit log.warn; send plan_no_tool_generation telemetry (with model/provider and tokens_output); add a synthetic TUI note suggesting /model.
  • Guards: one-shot per session; ignored for other agents or finish reasons.
  • Tests: 6 new cases cover positive/negative paths; all processor tests pass.

^{Written for commit 31db2ed. Summary will update on new commits.}

Summary by CodeRabbit

New Features

• Plan agent now displays a warning when it completes without generating any tool calls, providing visibility into agent behavior during sessions.

[claude]

Claude Code Review

This repository is configured for manual code reviews. Comment @claude review to trigger a review and subscribe this PR to future pushes, or @claude review once for a one-time review.

_{Tip: disable this comment in your organization's Code Review settings.}

[coderabbitai]

Note

Currently processing new changes in this PR. This may take a few minutes, please wait...

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: d18e5bbf-4057-49fd-89e8-ee1f7113f5b8

📥 Commits

Reviewing files that changed from the base of the PR and between ba9882f and 31db2ed.

📒 Files selected for processing (3)

• packages/opencode/src/altimate/telemetry/index.ts
• packages/opencode/src/session/processor.ts
• packages/opencode/test/session/processor.test.ts

 __________________________________________________
< Your stack overflowed; I brought a bigger stack. >
 --------------------------------------------------
  \
   \   \
        \ /\
        ( )
      .( o ).

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

✅ Unit Test PR creation complete.

• Create PR with unit tests
• Commit unit tests in branch fix/plan-agent-no-tool-warning

---

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.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Note

Unit test generation is a public access feature. Expect some limitations and changes as we gather feedback and continue to improve it.

---

Generating unit tests... This may take up to 20 minutes.

[cubic-dev-ai]

No issues found across 3 files

[coderabbitai]

✅ Created PR with unit tests: #654