Fix(cli): detect resolve intent and report intent_source

[AdityaPainuli]

Closes #996

Problem

potpie resolve "<task>" without --intent silently hardcodes intent=feature and prints it as if it was detected. No detection ever runs, so debugging and recent-change tasks get the feature retrieval shape (no timeline family), and the output misleads users into thinking the intent was classified.

What changed

• Added detect_context_intent(task) in domain/agent_context_port.py: a deterministic, dependency-free keyword/phrase matcher (case-insensitive, word-boundary) seeded from the trigger language in each recipe's "when" field. Multi-match ties break via a documented priority order. Returns None when nothing matches, favoring no guess over a wrong guess, mirroring the ontology_classifier discipline.
• CLI resolve: --intent default changed from "feature" to None. Effective intent now resolves explicit → detected → default (unknown), following the (value, resolved_via) pattern from resolve_pot_id/resolve_pot_scope.
• The selection provenance rides the existing ResolveRequest.metadata → AgentEnvelope.metadata passthrough as intent_source (explicit | detected | default). No downstream signature changes.
• Both outputs surface it: JSON gets an intent_source key, the human header shows intent=<x> (source=<...>).

Vocabulary, recipes, normalize_context_intent, DEFAULT_INTENT_INCLUDES, and the MCP/HTTP surfaces are untouched. Correct detection routes recent-change/debugging tasks to timeline-bearing includes automatically via the existing intent→includes mapping.

Behavior

Command	Before	After
resolve "webhook throwing a 500 error"	intent=feature	intent=debugging (source=detected)
resolve "what changed recently in auth?"	intent=feature, zero timeline items	intent=review (source=detected), timeline included
resolve "xyz" (no match)	intent=feature	intent=unknown (source=default)
resolve "..." --intent docs	intent=docs	intent=docs (source=explicit)

Testing

• 18 new unit tests: representative task→intent mappings, recent-change phrasing selects a timeline-bearing intent, empty/ambiguous input returns None, word-boundary anti-false-match ("latest" must not trigger test), CLI precedence, and intent_source in both output formats.
• Full unit suite: 2063 passed, 1 skipped, 0 failures.
• ruff check and ruff format clean (repo pre-commit config, v0.14.3).

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository UI (base), Organization UI (inherited)

Review profile: ASSERTIVE

Plan: Pro

Run ID: bb4beb13-2482-4744-a4c7-b0bf810c920d

📥 Commits

Reviewing files that changed from the base of the PR and between f4140de and bf1ccd5.

📒 Files selected for processing (1)

• potpie/context-engine/tests/unit/test_agent_context_port.py

---

📝 Walkthrough

Summary by CodeRabbit

• New Features
  • The resolve command can infer intent from the task text when no intent is provided.
  • Responses now include an intent_source indicator (explicit, detected, or default), shown in both structured and human-readable output.
• Bug Fixes
  • Improved intent detection using deterministic, word-boundary matching with priority-based tie-breaking to reduce false positives.
  • Added safe fallback behavior for empty or unclear task descriptions.
• Tests
  • Expanded unit coverage for intent inference, intent_source surfacing, edge cases, and regressions.

Walkthrough

The resolve CLI now detects intent from task text when --intent is omitted, records whether intent was explicit, detected, or defaulted, and exposes that source in JSON and human-readable responses. Deterministic tests cover canonical matches, priority, blank input, and substring safety.

Changes

Resolve intent selection

Layer / File(s)	Summary
Deterministic task intent detection potpie/context-engine/domain/agent_context_port.py, potpie/context-engine/tests/unit/test_agent_context_port.py	Adds prioritized, case-insensitive, word-boundary intent detection with neutral results for blank or unmatched tasks, plus unit coverage.
Resolve intent selection and propagation potpie/context-engine/adapters/inbound/cli/commands/query.py, potpie/context-engine/tests/unit/test_cli_resolve_intent.py	Combines explicit intent with task detection, defaults to unknown, and stores the selected intent source in ResolveRequest metadata.
Intent source response visibility potpie/context-engine/adapters/inbound/cli/commands/query.py, potpie/context-engine/tests/unit/test_cli_resolve_intent.py	Adds intent_source to JSON envelopes and displays it in human-readable resolve summaries, including default fallback behavior.

Sequence Diagram(s)

sequenceDiagram
  participant User
  participant ResolveCommand
  participant IntentDetector
  participant ResolveRequest
  participant ResponseEnvelope
  User->>ResolveCommand: run resolve with task
  ResolveCommand->>IntentDetector: detect intent when --intent is omitted
  IntentDetector-->>ResolveCommand: intent or no match
  ResolveCommand->>ResolveRequest: set intent and intent_source metadata
  ResolveRequest-->>ResponseEnvelope: provide resolve metadata
  ResponseEnvelope-->>User: display intent and source

Loading

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title concisely reflects the main CLI intent-detection and intent_source change.
Description check	✅ Passed	The description clearly matches the CLI intent detection and intent_source updates.
Linked Issues check	✅ Passed	The changes satisfy #996 by detecting intent, using an unknown fallback, surfacing intent_source, and improving timeline routing.
Out of Scope Changes check	✅ Passed	The changes appear focused on the requested CLI intent behavior and related tests, with no clear unrelated additions.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

📋 Issue Planner

Built with CodeRabbit's Coding Plans for faster development and fewer bugs.

View plan used: #996

---

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.}

[coderabbitai]

Actionable comments posted: 1

🤖 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 `@potpie/context-engine/domain/agent_context_port.py`:
- Around line 187-281: Add a consistency assertion adjacent to _INTENT_KEYWORDS
and _INTENT_DETECTION_PRIORITY that compares their intent-key sets and fails
clearly when either structure contains an intent missing from the other. Keep
_INTENT_MATCHERS construction unchanged, and ensure the guard runs before it
accesses _INTENT_MATCHERS[intent].

🪄 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: Repository UI (base), Organization UI (inherited)

Review profile: ASSERTIVE

Plan: Pro

Run ID: 4eeee886-7b71-45a9-8b96-35e036a63d0e

📥 Commits

Reviewing files that changed from the base of the PR and between a6ae3da and 88aa962.

📒 Files selected for processing (4)

• potpie/context-engine/adapters/inbound/cli/commands/query.py
• potpie/context-engine/domain/agent_context_port.py
• potpie/context-engine/tests/unit/test_agent_context_port.py
• potpie/context-engine/tests/unit/test_cli_resolve_intent.py

[FarbodDaneshjoo]

Nice work! I noticed there's good coverage for representative mappings and the fallback behavior, but I couldn't find a regression test for a task matching multiple intents.

Since detect_context_intent() relies on _INTENT_DETECTION_PRIORITY to resolve ties, would it make sense to add a regression test for a multi-match phrase? That would help ensure the documented priority order remains stable over time.

[AdityaPainuli]

Good catch, added. It pins four multi-match phrases ( security > debugging > operations > review > feature), and each case also guards-asserts the lower-priority intents genuinely match the phrase, so a future keyword change can't silently reduce the test to a single-match.

[FarbodDaneshjoo]

Thanks for adding it! I like the extra guard assertions—they make the priority behavior much clearer and help ensure the test stays meaningful over time.