✨ Vector candidate recall reports its completeness truthfully, and the tie-closure loop becomes shared adapter logic

[ebigunso]

Summary

v0.1.6 Wave 1, Task_2 (port contract, ADR-I-0024), targeting the planning branch per the phase's merge shape.

• The vector candidate port returns a result envelope: canonical candidates plus a completeness verdict in the public retrieval telemetry vocabulary (not requested, exhaustive, boundary tie closed, boundary tie open). The pipeline records it beside the returned count and never repairs, retries, or fails on it.
• The service adapter's private tie-closure loop is extracted into a shared crate-visible module with an engine-neutral fetch callback and its unit tests, so the embedded adapter reuses it; the backend fetch cap propagates into the bound so hitting it reports open, never a false exhaustive.
• Scope-only query: an empty object-type scope selects zero without issuing a search, and an empty configured scope is rejected at the boundary. A zero-norm query scores every candidate zero with a truthful verdict.
• Every test fake reports exhaustive; the verdict type is re-exported from both public barrels with an external doctest.

Validation

• Worker (cm-worker): fmt; Clippy -D warnings on Rust 1.95 and 1.97; full service-up cargo test; ignored Qdrant library tests 4/4; zero-hit census for match-or-unknown conditions and filters beyond object-type scope.
• Reviewer (cm-reviewer, Tier D): approved at 5b30856 after one revision round (checked backend-limit conversion with a BoundaryTieOpen regression at the u32 cap; public re-exports); service-up suite 411 passed / 0 failed.

🤖 Generated with Claude Code

[Copilot]

🟡 Changes recommended

The zero-norm shortcut accepts wrong-dimensional embeddings by bypassing Qdrant’s dimension validation.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

Adds truthful vector recall completeness telemetry and centralizes bounded tie-cohort closure.

Changes:

• Introduces completeness verdicts and propagates them through retrieval telemetry.
• Extracts shared tie-closure logic and adds zero-norm Qdrant handling.
• Enforces explicit object-type scopes and updates test stores.

File summaries

File	Description
src/usecases/retrieve.rs	Validates scope and records completeness.
src/usecases/remember.rs	Updates test vector store.
src/usecases/correct_forget.rs	Adapts tests to recall envelopes.
src/test_support.rs	Updates fake behavior and fixtures.
src/ports/vector_candidate.rs	Defines the recall envelope.
src/models/vector/candidate_record.rs	Requires scope and detects zero norm.
src/memory.rs	Tests empty-scope rejection.
src/lib.rs	Re-exports the verdict.
src/api/types/retrieval.rs	Adds validation and telemetry vocabulary.
src/api/types.rs	Exports the verdict through API types.
src/adapters/qdrant/tie_closure.rs	Implements shared bounded tie closure.
src/adapters/qdrant/store.rs	Produces truthful Qdrant recall verdicts.
src/adapters/qdrant.rs	Exposes shared adapter logic.
src/adapters/oxigraph/tests.rs	Updates test vector-store implementation.

Review details

• Files reviewed: 14/14 changed files
• Comments generated: 1
• Review effort level: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[ebigunso]

Fixed in 789d34a: a wrong-sized query embedding now fails with the existing collection-compatibility error (vector size mismatch) before either the zero-norm scroll or the vector search is chosen; the guard sits after the intentional empty-scope and zero-limit no-op and before tie closure. Regression test added. Reviewed internally (cm-reviewer, no findings); service-up suite and ignored Qdrant tests pass.

[Copilot]

🟢 Approval recommended

The implementation matches ADR-I-0024, propagates verdicts without control-flow coupling, and covers the relevant boundary cases.

Review details

• Files reviewed: 14/14 changed files
• Comments generated: 0 new
• Review effort level: Balanced

[Copilot]

🟢 Approval recommended

The port, pipeline, adapter, exports, and tests consistently implement the stated completeness contract without unresolved defects.

Review details

• Files reviewed: 14/14 changed files
• Comments generated: 0 new
• Review effort level: Balanced

[Copilot]

🟡 Changes recommended

The core public and port-level completeness semantics need explicit documentation to prevent incorrect adapter implementations and consumer interpretation.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Review details

• Files reviewed: 14/14 changed files
• Comments generated: 2
• Review effort level: Balanced