Compare commits
1 Commits
main
...
claude/r14
| Author | SHA1 | Date | |
|---|---|---|---|
| 3888db926f |
5
.gitignore
vendored
5
.gitignore
vendored
@@ -14,3 +14,8 @@ venv/
|
||||
.claude/*
|
||||
!.claude/commands/
|
||||
!.claude/commands/**
|
||||
|
||||
# Editor / IDE state — user-specific, not project config
|
||||
.obsidian/
|
||||
.vscode/
|
||||
.idea/
|
||||
|
||||
@@ -23,8 +23,6 @@
|
||||
- **capture_clients**: claude-code (Stop hook + cwd project inference), openclaw (before_agent_start + llm_output plugin, verified live)
|
||||
- **wiki**: http://dalidou:8100/wiki (browse), /wiki/projects/{id}, /wiki/entities/{id}, /wiki/search
|
||||
- **dashboard**: http://dalidou:8100/admin/dashboard (now shows pipeline health, interaction totals by client, all registered projects)
|
||||
- **active_track**: Engineering V1 Completion (started 2026-04-22). V1-0 landed (`2712c5d`), V1-A gated on soak (~2026-04-26) + density (100+ active memories, currently 84). Plan: `docs/plans/engineering-v1-completion-plan.md`. Resume map: `docs/plans/v1-resume-state.md`.
|
||||
- **open_branches**: `claude/r14-promote-400` at `3888db9` — P2 route fix (ValueError → 400), pending Codex review
|
||||
|
||||
## Active Plan
|
||||
|
||||
@@ -145,7 +143,7 @@ One branch `codex/extractor-eval-loop` for Day 1-5, a second `codex/retrieval-ha
|
||||
| R11 | Codex | P2 | src/atocore/api/routes.py:773-845 | `POST /admin/extract-batch` still accepts `mode="llm"` inside the container and returns a successful 0-candidate result instead of surfacing that host-only LLM extraction is unavailable from this runtime. That is a misleading API contract for operators. | fixed | Claude | 2026-04-12 | (pending) |
|
||||
| R12 | Codex | P2 | scripts/batch_llm_extract_live.py:39-190 | The host-side extractor duplicates the LLM system prompt and JSON parsing logic from `src/atocore/memory/extractor_llm.py`. It works today, but this is now a prompt/parser drift risk across the container and host implementations. | fixed | Claude | 2026-04-12 | (pending) |
|
||||
| R13 | Codex | P2 | DEV-LEDGER.md:12 | The new `286 passing` test-count claim is not reproducibly auditable from the current audit environments: neither Dalidou nor the clean worktree has `pytest` available. The claim may be true in Claude's dev shell, but it remains unverified in this audit. | fixed | Claude | 2026-04-12 | (pending) |
|
||||
| R14 | Codex | P2 | src/atocore/api/routes.py (POST /entities/{id}/promote) | The HTTP `POST /entities/{id}/promote` route does not translate the new service-layer `ValueError("source_refs required: cannot promote a candidate with no provenance...")` into a 400. A legacy no-provenance candidate promoted through the API currently surfaces as a 500. Does not block V1-0 acceptance; tidy in a follow-up. | open | Claude | 2026-04-22 | |
|
||||
| R14 | Codex | P2 | src/atocore/api/routes.py (POST /entities/{id}/promote) | The HTTP `POST /entities/{id}/promote` route does not translate the new service-layer `ValueError("source_refs required: cannot promote a candidate with no provenance...")` into a 400. A legacy no-provenance candidate promoted through the API currently surfaces as a 500. Does not block V1-0 acceptance; tidy in a follow-up. | fixed | Claude | 2026-04-22 | (pending) |
|
||||
|
||||
## Recent Decisions
|
||||
|
||||
|
||||
@@ -1,31 +1,6 @@
|
||||
# AtoCore — Current State (2026-04-22)
|
||||
# AtoCore — Current State (2026-04-19)
|
||||
|
||||
Live deploy: `2712c5d` · Dalidou health: ok · Harness: 17/18 · Tests: 547 passing.
|
||||
|
||||
## V1-0 landed 2026-04-22
|
||||
|
||||
Engineering V1 completion track has started. **V1-0 write-time invariants**
|
||||
merged and deployed: F-1 shared-header fields (`extractor_version`,
|
||||
`canonical_home`, `hand_authored`) added to `entities`, F-8 provenance
|
||||
enforcement at both `create_entity` and `promote_entity`, F-5 synchronous
|
||||
conflict-detection hook on every active-entity write path (create, promote,
|
||||
supersede) with Q-3 fail-open. Prod backfill ran cleanly — 31 legacy
|
||||
active/superseded entities flagged `hand_authored=1`, follow-up dry-run
|
||||
returned 0 remaining rows. Test count 533 → 547 (+14).
|
||||
|
||||
R14 (P2, non-blocking): `POST /entities/{id}/promote` route fix translates
|
||||
the new `ValueError` into 400. Branch `claude/r14-promote-400` pending
|
||||
Codex review + squash-merge.
|
||||
|
||||
**Next in the V1 track:** V1-A (minimal query slice + Q-6 killer-correctness
|
||||
integration). Gated on pipeline soak (~2026-04-26) + 100+ active memory
|
||||
density target. See `docs/plans/engineering-v1-completion-plan.md` for
|
||||
the full 7-phase roadmap and `docs/plans/v1-resume-state.md` for the
|
||||
"you are here" map.
|
||||
|
||||
---
|
||||
|
||||
## Snapshot from previous update (2026-04-19)
|
||||
Live deploy: `877b97e` · Dalidou health: ok · Harness: 17/18.
|
||||
|
||||
## The numbers
|
||||
|
||||
|
||||
@@ -168,40 +168,16 @@ These are the current practical priorities.
|
||||
"Zerodur" for p04 constraint queries. Investigate if it's a missing
|
||||
memory or retrieval ranking issue.
|
||||
|
||||
## Active — Engineering V1 Completion Track (started 2026-04-22)
|
||||
|
||||
The Engineering V1 sprint moved from **Next** to **Active** on 2026-04-22.
|
||||
The discovery from the gbrain review was that V1 entity infrastructure
|
||||
had been built incrementally already; the sprint is a **completion** plan
|
||||
against `engineering-v1-acceptance.md`, not a greenfield build. Full plan:
|
||||
`docs/plans/engineering-v1-completion-plan.md`. "You are here" single-page
|
||||
map: `docs/plans/v1-resume-state.md`.
|
||||
|
||||
Seven phases, ~17.5–19.5 focused days, runs in parallel with the Now list
|
||||
where surfaces are disjoint, pauses when they collide.
|
||||
|
||||
| Phase | Scope | Status |
|
||||
|---|---|---|
|
||||
| V1-0 | Write-time invariants: F-1 header fields + F-8 provenance enforcement + F-5 hook on every active-entity write + Q-3 flag-never-block | ✅ done 2026-04-22 (`2712c5d`) |
|
||||
| V1-A | Minimum query slice: Q-001 subsystem-scoped variant + Q-6 killer-correctness integration test on p05-interferometer | 🟡 gated — starts when soak (~2026-04-26) + density (100+ active memories) gates clear |
|
||||
| V1-B | KB-CAD + KB-FEM ingest (`POST /ingest/kb-cad/export`, `POST /ingest/kb-fem/export`) + D-2 schema docs | pending V1-A |
|
||||
| V1-C | Close the remaining 8 queries (Q-002/003/007/010/012/014/018/019; Q-020 to V1-D) | pending V1-B |
|
||||
| V1-D | Full mirror surface (3 spec routes + regenerate + determinism + disputed + curated markers) + Q-5 golden file | pending V1-C |
|
||||
| V1-E | Memory→entity graduation end-to-end + remaining Q-4 trust tests | pending V1-D (note: collides with memory extractor; pauses for multi-model triage work) |
|
||||
| V1-F | F-5 detector generalization + route alias + O-1/O-2/O-3 operational + D-1/D-3/D-4 docs | finish line |
|
||||
|
||||
R14 (P2, non-blocking): `POST /entities/{id}/promote` route returns 500
|
||||
on the new V1-0 `ValueError` instead of 400. Fix on branch
|
||||
`claude/r14-promote-400`, pending Codex review.
|
||||
|
||||
## Next
|
||||
|
||||
These are the next major layers after V1 and the current stabilization pass.
|
||||
These are the next major layers after the current stabilization pass.
|
||||
|
||||
1. Phase 6 AtoDrive — clarify Google Drive as a trusted operational
|
||||
source and ingest from it
|
||||
2. Phase 13 Hardening — Chroma backup policy, monitoring, alerting,
|
||||
failure visibility beyond log files
|
||||
3. Engineering V1 implementation sprint — once knowledge density is
|
||||
sufficient and the pipeline feels boring and dependable
|
||||
|
||||
## Later
|
||||
|
||||
|
||||
@@ -1,161 +0,0 @@
|
||||
# V1 Completion — Resume State
|
||||
|
||||
**Last updated:** 2026-04-22 (after V1-0 landed + R14 branch pushed)
|
||||
**Purpose:** single-page "you are here" so any future session can pick up
|
||||
the V1 completion sprint without re-reading the full plan history.
|
||||
|
||||
## State of play
|
||||
|
||||
- **V1-0 is DONE.** Merged to main as `2712c5d`, deployed to Dalidou,
|
||||
prod backfill ran cleanly (31 legacy entities flagged
|
||||
`hand_authored=1`, zero violations remaining).
|
||||
- **R14 is on a branch.** `claude/r14-promote-400` at `3888db9` —
|
||||
HTTP promote route returns 400 instead of 500 on V1-0 `ValueError`.
|
||||
Pending Codex review + squash-merge. Non-blocking for V1-A.
|
||||
- **V1-A is next but GATED.** Doesn't start until both gates clear.
|
||||
|
||||
## Start-gates for V1-A
|
||||
|
||||
| Gate | Condition | Status as of 2026-04-22 |
|
||||
|---|---|---|
|
||||
| Soak | Four clean nightly cycles since F4 confidence-decay first real run 2026-04-19 | Day 3 of 4 — expected clear around **2026-04-26** |
|
||||
| Density | 100+ active memories | 84 active as of last ledger update — need +16. Lever: `scripts/batch_llm_extract_live.py` against 234-interaction backlog |
|
||||
|
||||
**When both are green, start V1-A.** If only one is green, hold.
|
||||
|
||||
## Pre-flight checklist when resuming
|
||||
|
||||
Before opening the V1-A branch, run through this in order:
|
||||
|
||||
1. `git checkout main && git pull` — make sure you're at the tip
|
||||
2. Check `DEV-LEDGER.md` **Orientation** for current `live_sha`, `test_count`, `active_memories`
|
||||
3. Check `/health` on Dalidou returns the same `build_sha` as Orientation
|
||||
4. Check the dashboard for pipeline health: http://dalidou:8100/admin/dashboard
|
||||
5. Confirm R14 branch status — either merged or explicitly deferred
|
||||
6. Re-read the two core plan docs:
|
||||
- `docs/plans/engineering-v1-completion-plan.md` — the full 7-phase plan
|
||||
- `docs/architecture/engineering-v1-acceptance.md` — the acceptance contract
|
||||
7. Skim the relevant spec docs for the phase you're about to start:
|
||||
- V1-A: `engineering-query-catalog.md` (Q-001 + Q-006/Q-009/Q-011 killer queries)
|
||||
- V1-B: `tool-handoff-boundaries.md` (KB-CAD/KB-FEM export shapes)
|
||||
- V1-C: `engineering-query-catalog.md` (all remaining v1-required queries)
|
||||
- V1-D: `human-mirror-rules.md` (mirror spec end-to-end)
|
||||
- V1-E: `memory-vs-entities.md` (graduation flow)
|
||||
- V1-F: `conflict-model.md` (generic slot-key detector)
|
||||
|
||||
## What V1-A looks like when started
|
||||
|
||||
**Branch:** `claude/v1-a-pillar-queries`
|
||||
|
||||
**Scope (~1.5 days):**
|
||||
- **Q-001 shape fix.** Add a subsystem-scoped variant of `system_map()`
|
||||
matching `GET /entities/Subsystem/<id>?expand=contains` per
|
||||
`engineering-query-catalog.md:71`. The project-wide version stays
|
||||
(it serves Q-004).
|
||||
- **Q-6 integration test.** Seed p05-interferometer with five cases:
|
||||
1 satisfying Component, 1 orphan Requirement, 1 Decision on flagged
|
||||
Assumption, 1 supported ValidationClaim, 1 unsupported ValidationClaim.
|
||||
One test asserting Q-006 / Q-009 / Q-011 return exactly the expected
|
||||
members.
|
||||
- The four "pillar" queries (Q-001, Q-005, Q-006, Q-017) already work
|
||||
per Codex's 2026-04-22 audit. V1-A does NOT re-implement them —
|
||||
V1-A verifies them on seeded data.
|
||||
|
||||
**Acceptance:** Q-001 subsystem-scoped variant + Q-6 integration test both
|
||||
green. F-2 moves from 🟡 partial to slightly-less-partial.
|
||||
|
||||
**Estimated tests added:** ~4 (not ~12 — V1-A scope shrank after Codex
|
||||
confirmed most queries already work).
|
||||
|
||||
## Map of the remaining phases
|
||||
|
||||
```
|
||||
V1-0 ✅ write-time invariants landed 2026-04-22 (2712c5d)
|
||||
↓
|
||||
V1-A 🟡 minimum query slice gated on soak + density (~1.5d when started)
|
||||
↓
|
||||
V1-B KB-CAD/KB-FEM ingest + D-2 ~2d
|
||||
↓
|
||||
V1-C close 8 remaining queries ~2d
|
||||
↓
|
||||
V1-D full mirror + determinism ~3-4d (biggest phase)
|
||||
↓
|
||||
V1-E graduation + trust tests ~3-4d (pauses for multi-model triage)
|
||||
↓
|
||||
V1-F F-5 generalization + ops + docs ~3d — V1 done
|
||||
```
|
||||
|
||||
## Parallel work that can run WITHOUT touching V1
|
||||
|
||||
These are genuinely disjoint surfaces; pick any of them during the gate
|
||||
pause or as scheduling allows:
|
||||
|
||||
- **Density batch-extract** — *required* to unblock V1-A. Not optional.
|
||||
- **p04-constraints harness fix** — retrieval-ranking change, fully
|
||||
disjoint from entities. Safe to do anywhere in the V1 track.
|
||||
- **Multi-model triage (Phase 11 entry)** — memory-side work, disjoint
|
||||
from V1-A/B/C/D. **Pause before V1-E starts** because V1-E touches
|
||||
memory module semantics.
|
||||
|
||||
## What NOT to do
|
||||
|
||||
- Don't start V1-A until both gates are green.
|
||||
- Don't touch the memory extractor write path while V1-E is open.
|
||||
- Don't name the rejected "Minions" plan in any doc — neutral wording
|
||||
only ("queued background processing / async workers") per Codex
|
||||
sign-off.
|
||||
- Don't rename the `project` field to `project_id` — Codex + Antoine
|
||||
agreed it stays as `project`, with a doc note in
|
||||
`engineering-ontology-v1.md` that this IS the project_id per spec.
|
||||
|
||||
## Open review findings
|
||||
|
||||
| id | severity | summary | status |
|
||||
|---|---|---|---|
|
||||
| R14 | P2 | `POST /entities/{id}/promote` returns 500 on V1-0 `ValueError` instead of 400 | fixed on branch `claude/r14-promote-400`, pending Codex review |
|
||||
|
||||
Closed V1-0 findings: P1 "promote path allows provenance-less legacy
|
||||
candidates" (service.py:365-379), P1 "supersede path missing F-5 hook"
|
||||
(service.py:581-591), P2 "`--invalidate-instead` backfill too broad"
|
||||
(v1_0_backfill_provenance.py:52-63). All three patched and approved in
|
||||
the squash-merge to `2712c5d`.
|
||||
|
||||
## How agreement between Claude + Codex has worked so far
|
||||
|
||||
Three review rounds before V1-0 started + three during implementation:
|
||||
|
||||
1. **Rejection round.** Claude drafted a gbrain-inspired "Phase 8
|
||||
Minions + typed edges" plan; Codex rejected as wrong-packaging.
|
||||
Record: `docs/decisions/2026-04-22-gbrain-plan-rejection.md`.
|
||||
2. **Completion-plan rewrite.** Claude rewrote against
|
||||
`engineering-v1-acceptance.md`. Codex first-round review fixed the
|
||||
phase order (provenance-first).
|
||||
3. **Per-file audit.** Codex's second-round audit found F-1 / F-2 /
|
||||
F-5 gaps, all folded in.
|
||||
4. **Sign-off round.** Codex's third-round review resolved the five
|
||||
remaining open questions inline and signed off: *"with those edits,
|
||||
I'd sign off on the five questions."*
|
||||
5. **V1-0 review.** Codex found two P1 gaps (promote re-check missing,
|
||||
supersede hook missing) + one P2 (backfill scope too broad). All
|
||||
three patched. Codex re-ran probes + regression suites, approved,
|
||||
squash-merged.
|
||||
6. **V1-0 deploy + prod backfill.** Codex deployed + ran backfill,
|
||||
logged R14 as P2 residual.
|
||||
|
||||
Protocol has been: Claude writes, Codex audits, human Antoine ratifies.
|
||||
Continue this for V1-A onward.
|
||||
|
||||
## References
|
||||
|
||||
- `docs/plans/engineering-v1-completion-plan.md` — full 7-phase plan
|
||||
- `docs/decisions/2026-04-22-gbrain-plan-rejection.md` — prior rejection
|
||||
- `docs/architecture/engineering-ontology-v1.md` — V1 ontology (18 predicates)
|
||||
- `docs/architecture/engineering-query-catalog.md` — Q-001 through Q-020 spec
|
||||
- `docs/architecture/engineering-v1-acceptance.md` — F/Q/O/D acceptance table
|
||||
- `docs/architecture/promotion-rules.md` — candidate → active flow
|
||||
- `docs/architecture/conflict-model.md` — F-5 spec
|
||||
- `docs/architecture/human-mirror-rules.md` — V1-D spec
|
||||
- `docs/architecture/memory-vs-entities.md` — V1-E spec
|
||||
- `docs/architecture/tool-handoff-boundaries.md` — V1-B KB-CAD/KB-FEM
|
||||
- `docs/master-plan-status.md` — Now / Active / Next / Later
|
||||
- `DEV-LEDGER.md` — Orientation + Open Review Findings + Session Log
|
||||
@@ -2187,12 +2187,17 @@ def api_promote_entity(
|
||||
from atocore.engineering.service import promote_entity
|
||||
target_project = req.target_project if req is not None else None
|
||||
note = req.note if req is not None else ""
|
||||
try:
|
||||
success = promote_entity(
|
||||
entity_id,
|
||||
actor="api-http",
|
||||
note=note,
|
||||
target_project=target_project,
|
||||
)
|
||||
except ValueError as e:
|
||||
# V1-0 F-8 re-check raises ValueError for no-provenance candidates
|
||||
# (see service.promote_entity). Surface as 400, not 500.
|
||||
raise HTTPException(status_code=400, detail=str(e))
|
||||
if not success:
|
||||
raise HTTPException(status_code=404, detail=f"Entity not found or not a candidate: {entity_id}")
|
||||
result = {"status": "promoted", "id": entity_id}
|
||||
|
||||
@@ -160,6 +160,42 @@ def test_promote_rejects_legacy_candidate_without_provenance(tmp_data_dir):
|
||||
assert got.status == "candidate"
|
||||
|
||||
|
||||
def test_api_promote_returns_400_on_legacy_no_provenance(tmp_data_dir):
|
||||
"""R14 (Codex, 2026-04-22): the HTTP promote route must translate
|
||||
the V1-0 ValueError for no-provenance candidates into 400, not 500.
|
||||
Previously the route didn't catch ValueError so legacy bad
|
||||
candidates surfaced as a server error."""
|
||||
init_db()
|
||||
init_engineering_schema()
|
||||
|
||||
import uuid as _uuid
|
||||
from fastapi.testclient import TestClient
|
||||
from atocore.main import app
|
||||
|
||||
entity_id = str(_uuid.uuid4())
|
||||
with get_connection() as conn:
|
||||
conn.execute(
|
||||
"INSERT INTO entities (id, entity_type, name, project, "
|
||||
"description, properties, status, confidence, source_refs, "
|
||||
"extractor_version, canonical_home, hand_authored, "
|
||||
"created_at, updated_at) "
|
||||
"VALUES (?, 'component', 'Legacy HTTP', 'p04-gigabit', "
|
||||
"'', '{}', 'candidate', 1.0, '[]', '', 'entity', 0, "
|
||||
"CURRENT_TIMESTAMP, CURRENT_TIMESTAMP)",
|
||||
(entity_id,),
|
||||
)
|
||||
|
||||
client = TestClient(app)
|
||||
r = client.post(f"/entities/{entity_id}/promote")
|
||||
assert r.status_code == 400
|
||||
assert "source_refs required" in r.json().get("detail", "")
|
||||
|
||||
# Row still candidate — the 400 didn't half-transition.
|
||||
got = get_entity(entity_id)
|
||||
assert got is not None
|
||||
assert got.status == "candidate"
|
||||
|
||||
|
||||
def test_promote_accepts_candidate_flagged_hand_authored(tmp_data_dir):
|
||||
"""The other side of the promote re-check: hand_authored=1 with
|
||||
empty source_refs still lets promote succeed, matching
|
||||
|
||||
Reference in New Issue
Block a user