docs: finalize feat_overnight_final_solution (PR #440)

docs/00_overview/DASHBOARD.md

@@ -7,7 +7,7 @@ _Top-level index across MVP1 → GA v1+ as of **2026-06-04**. Click a release na
 | Release | Theme | Progress | Status |
 |---|---|---|---|
 | [MVP1 / v0.1](MVP1_DASHBOARD.md) | The Loop | 94 / 94 scoped done | **Complete** |
-| [MVP2 / v0.2](MVP2_DASHBOARD.md) | Three-Engine + Real Signals | 14 / 25 scoped done · 26 remaining | **In progress** |
+| [MVP2 / v0.2](MVP2_DASHBOARD.md) | Three-Engine + Real Signals | 15 / 25 scoped done · 25 remaining | **In progress** |
 | MVP3 / v0.3 | Observable | — | **Not yet scoped** |
 | GA v1 / v1.0 | Production-ready | — | **Not yet scoped** |
 

docs/00_overview/dashboard.html

@@ -392,7 +392,7 @@ <h2>Releases</h2>
 <div class="roadmap-row">
   <div class="release-name"><a href="mvp2_dashboard.html">MVP2 / v0.2</a></div>
   <div class="theme">Three-Engine + Real Signals</div>
-  <div class="progress">14 / 25 scoped done · 26 remaining</div>
+  <div class="progress">15 / 25 scoped done · 25 remaining</div>
   <span class="state-pill in_progress">In progress</span>
 </div>
 

docs/00_overview/planned_features/02_mvp2/feat_overnight_final_solution/feature_spec.md → docs/00_overview/implemented_features/2026_06_04_feat_overnight_final_solution/feature_spec.md

@@ -34,7 +34,7 @@
 | Digest worker | [`backend/workers/digest.py`](../../../../backend/workers/digest.py) | Line 1289 reads `auto_followup_depth = study.config.get("auto_followup_depth")` and, if not None, enqueues `enqueue_followup_study` via Arq with deterministic `_job_id=f"enqueue_followup_study:{study_id}"`. The digest also persists `suggested_followups` as JSONB on the `digests` row before this dispatch. |
 | Digest model | [`backend/app/db/models/digest.py`](../../../../backend/app/db/models/digest.py) | `Digest.suggested_followups: Mapped[list[dict[str, Any]]]` — NOT NULL, JSONB, server_default `'[]'::jsonb`. 1:1 with `studies` via UNIQUE FK on `study_id`. Consumers read via `parse_followup_list()` per spec D-defensive-ingest. |
 | Study model | [`backend/app/db/models/study.py`](../../../../backend/app/db/models/study.py) | `studies.config: JSONB` carries `auto_followup_depth`. Self-FK `parent_study_id`. `parent_proposal_id` + `parent_proposal_followup_index` ([lines 86-97](../../../../backend/app/db/models/study.py#L86-L97)) are the lineage columns the manual "Run this followup" path uses — DB CHECK `studies_parent_proposal_pair_check` requires both-set-or-both-NULL. |
-| Proposal model | [`backend/app/db/models/proposal.py`](../../../../backend/app/db/models/proposal.py) | `Proposal.status` CHECK constraint — `status IN ('pending', 'pr_opened', 'pr_merged', 'rejected')` ([line 42](../../../../backend/app/db/models/proposal.py#L42)). **No `superseded` value today** — adding one requires a migration (deferred to Phase 3 `phase3_idea.md`). |
+| Proposal model | [`backend/app/db/models/proposal.py`](../../../../backend/app/db/models/proposal.py) | `Proposal.status` CHECK constraint — `status IN ('pending', 'pr_opened', 'pr_merged', 'rejected')` ([line 42](../../../../backend/app/db/models/proposal.py#L42)). **No `superseded` value today** — adding one requires a migration (deferred to Phase 3 `feat_overnight_final_solution_phase3/idea.md`). |
 | Chain endpoint | [`backend/app/api/v1/studies.py:856-867`](../../../../backend/app/api/v1/studies.py#L856-L867) + Pydantic `StudyChainLink` at [`schemas.py:867-885`](../../../../backend/app/api/v1/schemas.py#L867-L885) | `GET /api/v1/studies/{id}/chain` returns `links: list[StudyChainLink]` with the rolled-up `best_link_id` + `cumulative_lift` + `stop_reason` + `proposal_id_for_best_link`. The `StudyChainLink` shape is explicitly extensible (the convergence-indicator spec FR-7 added `convergence_verdict` as a soft-contract additive field — see [`convergence.py:77-89`](../../../../backend/app/domain/study/convergence.py#L77-L89)). |
 | Chain panel | [`ui/src/components/studies/auto-followup-chain-panel.tsx`](../../../../../ui/src/components/studies/auto-followup-chain-panel.tsx) | Calls `useStudyChain(studyId)` and renders the ordered link list + cumulative-lift + stop-reason + best-config CTA per feat_overnight_autopilot FR-4. Reusing this panel — no replacement. |
 | Wizard depth selector | [`ui/src/components/studies/create-study-modal.tsx:1460-1468`](../../../../../ui/src/components/studies/create-study-modal.tsx#L1460-L1468) | The `🌙 Run overnight (compound automatically)` label + `cs-auto-followup` testid + `InfoTooltip glossaryKey="overnight_autopilot"` + `Select` writing `auto_followup_depth: 0..5` into `config`. This feature ADDS a strategy toggle immediately below it. |
@@ -87,8 +87,8 @@ No URL changes. The chain panel and the wizard mount at their existing positions
 ### Out of scope
 
 - Any change to `evaluate_chain_gate`, the budget peek, the depth decrement, the cancel cascade, or the layer-1/layer-2 idempotency contract. The strategy dispatch happens AFTER all of these.
-- A `superseded` value on `proposals.status` (Phase 3 → `phase3_idea.md`). MVP2 leans on the existing `/chain` endpoint's `best_link_id` + `proposal_id_for_best_link` to give the operator a single morning artifact; marking non-winning links' proposals `superseded` is a separate UX decision + migration that's not required for the core "explore + roll up" capability.
-- A standalone morning summary card on the `/studies` list (Phase 2 → `phase2_idea.md`, coordinates with the existing `feat_overnight_studies_summary_card` sibling idea).
+- A `superseded` value on `proposals.status` (Phase 3 → `feat_overnight_final_solution_phase3/idea.md`). MVP2 leans on the existing `/chain` endpoint's `best_link_id` + `proposal_id_for_best_link` to give the operator a single morning artifact; marking non-winning links' proposals `superseded` is a separate UX decision + migration that's not required for the core "explore + roll up" capability.
+- A standalone morning summary card on the `/studies` list (Phase 2 → `feat_overnight_final_solution_phase2/idea.md`, coordinates with the existing `feat_overnight_studies_summary_card` sibling idea).
 - A new follow-up kind, a change to the digest LLM prompt, or a change to the digest's structured-output schema.
 - Multi-child fan-out per parent. The shipped engine's linear-chain invariant (D-7 of `feat_overnight_autopilot`) holds — strategy selection picks ONE follow-up per link.
 - Operator-pickable mid-chain strategy switching. Strategy is set at study create and inherited verbatim by descendants.
@@ -106,8 +106,8 @@ No URL changes. The chain panel and the wizard mount at their existing positions
 ### Phase boundaries
 
 - **Phase 1 (this spec, MVP2):** FR-1 through FR-9 — the strategy wire contract, the wizard toggle, the worker dispatch, the cycle guard, the chain endpoint additive field, the panel badge, telemetry, tutorial, glossary key. Ships the autonomous cross-knob/cross-template exploration capability behind an opt-in toggle.
-- **Phase 2 (deferred to [`phase2_idea.md`](phase2_idea.md)):** Dedicated morning summary card surfacing the rolled-up winner + the explored path + total lift, separate from the chain panel. Coordinates with [`feat_overnight_studies_summary_card`](../feat_overnight_studies_summary_card/idea.md). Rationale for deferral: the existing `/chain` endpoint already exposes the data needed; a polished morning card is a UX add-on that should follow rather than block the capability.
-- **Phase 3 (deferred to [`phase3_idea.md`](phase3_idea.md)):** Proposal `superseded` status value + state-transition logic that marks non-winning chain links' proposals `superseded` so the morning artifact is unambiguously *one* answer. Rationale for deferral: requires a migration that reopens shipped schema (CHECK constraint on `proposals.status`) and a UX decision on whether superseded proposals appear in the `/proposals` index at all. Phase 1 delivers cross-knob exploration; Phase 3 polishes the rollup. Build it when an incident or design partner asks for the cleaner index.
+- **Phase 2 (deferred to [`feat_overnight_final_solution_phase2/idea.md`](../../planned_features/02_mvp2/feat_overnight_final_solution_phase2/idea.md)):** Dedicated morning summary card surfacing the rolled-up winner + the explored path + total lift, separate from the chain panel. Coordinates with [`feat_overnight_studies_summary_card`](../feat_overnight_studies_summary_card/idea.md). Rationale for deferral: the existing `/chain` endpoint already exposes the data needed; a polished morning card is a UX add-on that should follow rather than block the capability.
+- **Phase 3 (deferred to [`feat_overnight_final_solution_phase3/idea.md`](../../planned_features/02_mvp2/feat_overnight_final_solution_phase3/idea.md)):** Proposal `superseded` status value + state-transition logic that marks non-winning chain links' proposals `superseded` so the morning artifact is unambiguously *one* answer. Rationale for deferral: requires a migration that reopens shipped schema (CHECK constraint on `proposals.status`) and a UX decision on whether superseded proposals appear in the `/proposals` index at all. Phase 1 delivers cross-knob exploration; Phase 3 polishes the rollup. Build it when an incident or design partner asks for the cleaner index.
 
 ---
 
@@ -678,15 +678,15 @@ If the chain produces an unexpected swap_template result the operator wants to a
 - [ ] Coverage gate ≥ 80% holds.
 - [ ] Rollout gates from §16 satisfied (no schema change, no migration, no flag).
 - [ ] `docs/01_architecture/api-conventions.md` + `data-model.md` + `ui-architecture.md` + `tutorial-first-study.md` updated.
-- [ ] Phase 2 + Phase 3 deferred-work tracking files (`phase2_idea.md`, `phase3_idea.md`) exist alongside this spec.
+- [ ] Phase 2 + Phase 3 deferred-work tracking files exist as their own planned_features folders (`feat_overnight_final_solution_phase2/`, `feat_overnight_final_solution_phase3/`).
 - [ ] No open questions remain in §19.
 
 ## 19) Open questions and decision log
 
 ### Open questions
 
 - **OQ-1 (resolved at GPT-5.5 cycle 1, finding C1-B1)** — How does the chain-panel badge resolve the "short template name" for a `swap_template` link's display? **Resolved as D-11**: per-link `GET /api/v1/query-templates/{id}` fetch from the frontend (FR-7 updated). Rationale: at most 0–5 extra small fetches per chain, already TanStack-Query-cached client-side, keeps `/chain`'s response shape stable.
-- **OQ-2 (resolved at GPT-5.5 cycle 2, finding C2-B3)** — Should the strategy toggle ALSO show as a read-only line on the study detail page? **Resolved as D-15**: deferred to Phase 2 (`phase2_idea.md`). The chain-panel badges per link (FR-7) already surface the strategy a chain link followed; an extra detail-page line would be a redundant secondary surface. If operator feedback during MVP2 says the chain panel is too far down the page to spot quickly, Phase 2 picks it up as part of the morning summary card scope.
+- **OQ-2 (resolved at GPT-5.5 cycle 2, finding C2-B3)** — Should the strategy toggle ALSO show as a read-only line on the study detail page? **Resolved as D-15**: deferred to Phase 2 (`feat_overnight_final_solution_phase2/idea.md`). The chain-panel badges per link (FR-7) already surface the strategy a chain link followed; an extra detail-page line would be a redundant secondary surface. If operator feedback during MVP2 says the chain panel is too far down the page to spot quickly, Phase 2 picks it up as part of the morning summary card scope.
 
 _No open questions remain — §18's "no open questions" gate is satisfied._
 
@@ -709,4 +709,4 @@ _No open questions remain — §18's "no open questions" gate is satisfied._
   - Rationale: a single, unambiguous contract everywhere (FR-3, FR-5, FR-6, AC-3, AC-6, AC-9, AC-12, AC-18 all reconcile). The earlier draft had FR-3 and FR-5/AC-12 contradicting each other on the legacy-path persistence — D-12 resolves in favor of the clean-legacy contract.
 - **D-13 (2026-06-03, GPT-5.5 cycle 1 finding C1-A3 accept)** — `auto_followup_strategy` field type is `str | None` (NOT `Literal[...]`). The pair-and-value check happens in the `_validate_auto_followup_strategy` model_validator with the message-prefix path so the canonical `AUTO_FOLLOWUP_STRATEGY_INVALID` error code reaches the response envelope. Mirrors the existing `_validate_auto_followup_depth` pattern — a Pydantic `Literal[...]` at field-level would surface generic `VALIDATION_ERROR` for unknown values, violating §8.6's error-code contract.
 - **D-14 (2026-06-03, GPT-5.5 cycle 1 finding C1-A4 accept)** — The wizard does NOT write `auto_followup_visited_template_ids`. The worker is the sole writer. The anchor's missing key is treated as `[anchor.template_id]` by the worker. The create-study contract test asserts a wizard-submitted `auto_followup_visited_template_ids` is 422-rejected. Rationale: single-writer rule eliminates the "two writers must agree on the seed value" coordination surface.
-- **D-15 (2026-06-03, GPT-5.5 cycle 2 finding C2-B3 accept)** — Strategy read-only line on the study detail page (OQ-2) is deferred to Phase 2 (`phase2_idea.md`). The FR-7 per-link chain-panel badges are sufficient for MVP2; an extra detail-page line is redundant and would crowd the existing detail-page layout. Phase 2 picks it up if operator feedback says the chain panel is too far down to spot quickly during morning review.
+- **D-15 (2026-06-03, GPT-5.5 cycle 2 finding C2-B3 accept)** — Strategy read-only line on the study detail page (OQ-2) is deferred to Phase 2 (`feat_overnight_final_solution_phase2/idea.md`). The FR-7 per-link chain-panel badges are sufficient for MVP2; an extra detail-page line is redundant and would crowd the existing detail-page layout. Phase 2 picks it up if operator feedback says the chain panel is too far down to spot quickly during morning review.

docs/00_overview/planned_features/02_mvp2/feat_overnight_final_solution/implementation_plan.md → docs/00_overview/implemented_features/2026_06_04_feat_overnight_final_solution/implementation_plan.md

@@ -1,7 +1,7 @@
 # Implementation Plan — Overnight → final solution (autonomous cross-knob tuning)
 
 **Date:** 2026-06-03
-**Status:** Ready for Execution
+**Status:** Complete (PR #440, squash-merged `1e9522a0` 2026-06-04)
 **Primary spec:** [`feature_spec.md`](feature_spec.md)
 **Policy source(s):** [`CLAUDE.md`](../../../../CLAUDE.md) (Absolute Rules), [`docs/01_architecture/api-conventions.md`](../../../../01_architecture/api-conventions.md)
 
@@ -30,7 +30,7 @@
 | FR-9 (glossary key) | Epic 1 / Story 1.2 | `overnight_strategy` glossary key ships with the wizard toggle |
 | FR-9 (tutorial + runbook) | Epic 4 / Story 4.1 | Tutorial Step 12 sub-section + autopilot runbook event section |
 
-All spec FRs covered. No deferred FRs in Phase 1 (Phase 2 + Phase 3 tracked in `phase2_idea.md` + `phase3_idea.md`).
+All spec FRs covered. No deferred FRs in Phase 1 (Phase 2 + Phase 3 tracked in `feat_overnight_final_solution_phase2/idea.md` + `feat_overnight_final_solution_phase3/idea.md`).
 
 ## 2) Delivery structure
 

docs/00_overview/implemented_features/2026_06_04_feat_overnight_final_solution/pipeline_status.md

@@ -0,0 +1,34 @@
+# Pipeline Status — feat_overnight_final_solution
+
+**Release:** mvp2
+
+## Idea
+- Status: Complete
+- File: idea.md
+
+## Spec
+- Status: Approved
+- Date: 2026-06-03
+- File: feature_spec.md
+- Cross-model review: GPT-5.5 passed (2 cycles to convergence; 0 High-severity findings at cycle 2)
+- Cycle 1: 11 findings (6 High, 5 Medium, 0 Low) — all 11 accepted and applied
+- Cycle 2: 6 findings (0 High, 5 Medium, 1 Low) — all 6 accepted and applied (internal-consistency cleanups from cycle 1 edits)
+- Phases: 3 total (Phase 1 covered by this spec; Phase 2 + Phase 3 deferred with `feat_overnight_final_solution_phase2/idea.md` + `feat_overnight_final_solution_phase3/idea.md`)
+
+## Plan
+- Status: Approved
+- Date: 2026-06-03
+- File: implementation_plan.md
+- Cross-model review: GPT-5.5 passed (2 cycles; cycle 1: 10 findings (5 High, 5 Medium) all accepted+applied; cycle 2: 0 findings — converged)
+- Stories: 7 across 4 epics (Epic 1 schema+wizard, Epic 2 worker dispatch, Epic 3 chain surface, Epic 4 docs)
+- Phases covered: Phase 1 (Phase 2 + 3 split out to their own planned_features folders `feat_overnight_final_solution_phase2/` + `feat_overnight_final_solution_phase3/` at finalization)
+
+## Implementation
+- Status: Complete
+- Date: 2026-06-04
+- PR: #440 (squash-merged `1e9522a0`)
+- CI: green (all 17 `pr.yml` checks)
+- Stories: 7/7 complete across 4 epics
+- Cross-model review: Gemini 1 finding (rejected — hunk-isolated `child_id` false positive); GPT-5.5 final review 3 findings (0 High; 2 Medium + 1 Low all accepted + applied in `ac2fdc8a`)
+- Tests: 17 domain unit + 10 worker integration + 11 contract + 4 schema unit + 6 wizard vitest + 2 chain-panel vitest + 4 enum source-of-truth + 1 glossary value-lock
+- Deferred: Phase 2 (`feat_overnight_final_solution_phase2/idea.md` — morning summary card) + Phase 3 (`feat_overnight_final_solution_phase3/idea.md` — proposal `superseded` status) remain tracked; tangential `chore_e2e_overnight_strategy_radix_select_timing` + adjacent `feat_proposal_full_param_space_view` ideas filed