fix: M4 Phase 1 bugs - dev.chunks refs, search_path, uuid column

Bug fixes from M4 report:
- 4 remaining dev.chunks → dev.chunk in SQL queries
- search_path includes public for pgvector extension
- get_chunk_by_chunk_id_and_uuid: uuid → file_uuid
- New endpoint: GET /api/v1/file/:uuid/chunk/:chunk_id

docs_v1.0/M4_workspace/2026-05-11_Phase1_bug_report.md

@@ -0,0 +1,97 @@
+# Bug Report: Schema Migration Bugs Found During M4 Phase 1 Testing
+
+**Date**: 2026-05-11
+**From**: M4 (Integration & Testing)
+**To**: M5 (Development)
+**Priority**: High (3 bugs, all cause 500 errors)
+
+---
+
+## Bug 1: Stale `dev.chunks` References After Table Rename
+
+M5 renamed `dev.chunks` → `dev.chunk` but 4 SQL queries still hardcoded `dev.chunks`, causing "relation does not exist" errors.
+
+### Affected Files
+
+| File | Line | Old | New |
+|------|------|-----|-----|
+| `src/core/db/postgres_db.rs` | 4626 | `FROM dev.chunks` | `FROM dev.chunk` |
+| `src/api/five_w1h_agent_api.rs` | 779 | `UPDATE dev.chunks SET embedding` | `UPDATE dev.chunk SET embedding` |
+| `src/worker/processor.rs` | 1083 | `UPDATE dev.chunks SET metadata` | `UPDATE dev.chunk SET metadata` |
+| `src/worker/job_worker.rs` | 1005 | `FROM dev.chunks WHERE` | `FROM dev.chunk WHERE` |
+
+### Why M5 Tests Passed
+M5 likely still had the OLD table (ALTER RENAME keeps source, M5 may not have dropped it). M4 followed the migration exactly (RENAME), so the old table is gone.
+
+### Trigger
+- `POST /api/v1/search/smart` → 500 error "relation dev.chunks does not exist"
+
+---
+
+## Bug 2: `search_path` Missing `public` — pgvector `vector` Type Not Found
+
+`after_connect` sets `SET search_path TO dev` which REPLACES the default `"$user", public` — losing access to `public` schema where pgvector extension lives.
+
+### Affected Code
+```rust
+// src/core/db/postgres_db.rs:745
+sqlx::query(&format!("SET search_path TO {}", schema))
+```
+
+### Fix
+```rust
+sqlx::query(&format!("SET search_path TO {}, public", schema))
+```
+
+### Trigger
+- Any query using `::vector` cast (e.g., `search_parent_chunks_semantic`) after `SET search_path TO dev`
+- Error: "type vector does not exist"
+
+---
+
+## Bug 3: `get_chunk_by_chunk_id_and_uuid` Uses Wrong Column Name
+
+The function queries `SELECT ... uuid, ... WHERE uuid = $2` but the `dev.chunk` table column is `file_uuid`, not `uuid`.
+
+### Affected Code
+```rust
+// src/core/db/postgres_db.rs:2776
+"SELECT ... uuid, ... FROM {} WHERE chunk_id = $1 AND uuid = $2"
+```
+
+### Fix
+```sql
+SELECT ... file_uuid as uuid, ... FROM dev.chunk WHERE chunk_id = $1 AND file_uuid = $2
+```
+
+### Trigger
+- `GET /api/v1/file/:uuid/chunk/:chunk_id` → 500 "column uuid does not exist"
+- `POST /api/v1/search/universal` → same error (internally calls this function)
+
+---
+
+## Bug 4 (Observation): PG Has No Embeddings for M5 Charade
+
+M5's Charade data (`aeed71342a899fe4b4c57b7d41bcb692`) has 0 embeddings in PostgreSQL (all 4,188 sentence chunks have `embedding IS NULL`). Vectors are only in Qdrant.
+
+### Impact
+- `POST /api/v1/search/smart` returns empty results (requires `embedding IS NOT NULL`)
+- `POST /api/v1/search/universal` works (queries Qdrant)
+
+### M4 Workaround
+Demo script step 20 changed from `search/smart` to `search/universal` to show meaningful results.
+
+---
+
+## M5 Response
+
+All 4 bugs fixed and verified (37/37 API tests ✅):
+
+| # | Bug | Fix | Status |
+|:-:|-----|-----|:------:|
+| 1 | `dev.chunks` refs | 4 files updated: `postgres_db.rs`, `five_w1h_agent_api.rs`, `processor.rs`, `job_worker.rs` | ✅ |
+| 2 | `search_path` missing `public` | `postgres_db.rs:745` — `SET search_path TO dev, public` | ✅ |
+| 3 | `uuid` vs `file_uuid` | `postgres_db.rs:2777` — `/* ... */ AND file_uuid = $2` with `file_uuid as uuid` alias | ✅ |
+| 4 | PG embeddings empty | Confirmed design choice — vectors in Qdrant only. `search/universal` works. | ⚠️ By design |
+
+New source bundle updated at `docs_v1.0/M4_HANDOVER/momentry_core_v1.0.2_source.tar.gz`.

docs_v1.0/M4_workspace/2026-05-11_chunk_detail_endpoint.md

@@ -0,0 +1,70 @@
+# API Report: Missing Chunk Detail Endpoint
+
+**Date**: 2026-05-11
+**From**: M4 (Integration & Testing)
+**To**: M5 (Development)
+**Priority**: Medium
+
+---
+
+## Issue
+
+Portal's `ChunkDetailView` needs to fetch a single chunk by `file_uuid` + `chunk_id`. Currently no dedicated endpoint exists for this.
+
+## Proposed Endpoint
+
+```
+GET /api/v1/file/:file_uuid/chunk/:chunk_id
+```
+
+- Method: `GET`
+- Path params: `file_uuid` (String), `chunk_id` (String)
+- Response: Single `Chunk` object (same structure as chunk items in search results)
+
+## Existing Building Block
+
+The DB-layer method already exists:
+
+```rust
+// src/core/db/postgres_db.rs:2770
+pub async fn get_chunk_by_chunk_id_and_uuid(
+    &self,
+    chunk_id: &str,
+    uuid: &str,
+) -> Result<Option<Chunk>>
+```
+
+Currently only used internally by the search handler (`server.rs:1494, 1534`). Only a route + handler wrapper is needed.
+
+## Why Not the Old `/file/:uuid/chunks` Endpoint
+
+The old endpoint returned ALL chunks for a file, and the portal filtered client-side. This is inefficient. The new single-chunk endpoint replaces it cleanly.
+
+## Portal Impact
+
+`portal/src/views/ChunkDetailView.vue:245` currently calls:
+```
+GET /api/v1/file/{uuid}/chunks  → filters client-side by chunk_id
+```
+Will switch to:
+```
+GET /api/v1/file/{uuid}/chunk/{chunk_id}  → direct single result
+```
+
+## Temporary Workaround (M4 Only)
+
+M4 has temporarily re-added `GET /api/v1/file/:file_uuid/chunks` for portal compatibility. This will be removed once the new endpoint is available.
+
+## M5 Response
+
+**Status**: ✅ Implemented
+
+**Endpoint**: `GET /api/v1/file/:file_uuid/chunk/:chunk_id`
+
+**Verified**:
+- `0-01` → 200 (corrected chunk)
+- `1446-01` → 200 (corrected chunk)  
+- `story_240` → 200 (story chunk)
+- `nonexistent` → 404
+
+M4 can remove the temporary workaround (`/file/:uuid/chunks`) and switch Portal to use this endpoint.