Skip to content

🔎 The trace says how well each description matched, and the documents say what a description brings - #141

Open
ebigunso wants to merge 6 commits into
feature/2026-09-22/measured-floorsfrom
feature/2026-09-21/scene-trace-and-docs
Open

ebigunso wants to merge 6 commits into
feature/2026-09-22/measured-floorsfrom
feature/2026-09-21/scene-trace-and-docs

Conversation

@ebigunso

Copy link
Copy Markdown
Owner

Scene-words plan Task_3, stacked on #137. This completes the plan's three tasks.

  • With the trace on, each description search (the setting; all participants' words together) reports one best score, attributed to the scene references that share that search, or none when no experience carries scene words. Trigger: a description given at recall with the trace on. Reader: the companion repository's paraphrase measurement, so a later decision on whether a description can remind of nothing rests on numbers. With the trace off the result is byte-equal to the previous task.
  • The README now opens with the behavior: a key or a name means the character knows who or where this is and brings where things stand; a description only reminds, of the last few times like this, and never opens the history of who was there. The schema reference describes the content surface and the two scene surfaces.

Evidence: Codex Tier D approved at 4cf9763 (eight parent/final facade cases with literal trace-off equality; 457 tests); Claude Tier A approved at 0e3ab9f with wording changes, applied. Stacked PRs run no Rust CI; orchestrator gates pass.

🤖 Generated with Claude Code

@ebigunso
ebigunso changed the base branch from feature/2026-09-21/scene-reminders to feature/2026-09-22/measured-floors September 22, 2026 22:00
@ebigunso
ebigunso added this pull request to stack #149 September 22, 2026 22:00
…kept

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Copilot AI balanced review requested due to automatic review settings September 22, 2026 22:01
…21/scene-trace-and-docs

Co-Authored-By: Codex <noreply@openai.com>

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copilot review overview

🟡 Changes recommended

The serde compatibility behavior violates repository policy, and several completion records are future-dated.

Get a fresh assessment by requesting another Copilot review.

Review effort: Balanced
Findings: 1 Medium severity

Open (1)
What changed in this PR

Adds per-scene-search trace diagnostics and documents description-based recall behavior.

Changes:

  • Reports best scores, references, and omitted counts per scene search.
  • Refactors vector-root construction while preserving untraced retrieval behavior.
  • Updates retrieval documentation and closes the scene-words plan.
File Description
tests/​scene_reminder_tests.rs Adapts reminder assertions to search traces.
src/​usecases/​retrieve/​scene.rs Builds scene-search traces and candidate roots.
src/​usecases/​retrieve.rs Consumes refactored roots and trace data.
src/​models/​vector/​candidate_record.rs Limits truncation helper to tests.
src/​memory/​retrieval_turn_tests.rs Updates omission-count assertions.
src/​memory/​retrieval_scene_tests.rs Tests scores, references, and trace-off equality.
src/​lib.rs Exports the new trace type.
src/​api/​types/​retrieval.rs Defines the scene-search trace API.
src/​api/​types.rs Re-exports the trace type.
src/​adapters/​qdrant/​tie_closure.rs Rebuilds admitted canonical candidates.
README.md Explains description recall behavior.
docs/​design/​database/​vector_payload_design.md Documents search surfaces and diagnostics.
docs/​design/​database/​schema_cheat_sheet.md Summarizes scene-surface retrieval.
docs/​coding-agent/​plans/​completed/​v0-2-scene-words-plan.md Records plan completion.

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

/// Participant words share one search and therefore one count.
#[serde(default, skip_serializing_if = "BTreeMap::is_empty")]
pub scene_cue_omitted_counts: BTreeMap<CueKind, usize>,
#[serde(default, skip_serializing_if = "Vec::is_empty")]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants