SciDEX — Task: [Senate] Database integrity check

IMPORTANT: Before implementing, verify this work hasn't already been done. Check api.py routes, git log, and the live site. If already implemented, mark complete with summary explaining what already exists. Check for orphaned records, NULL required fields, inconsistent counts. Fix what can be fixed. ## REOPENED TASK — CRITICAL CONTEXT This task was previously marked 'done' but the audit could not verify the work actually landed on main. The original work may have been: - Lost to an orphan branch / failed push - Only a spec-file edit (no code changes) - Already addressed by other agents in the meantime - Made obsolete by subsequent work **Before doing anything else:** 1. **Re-evaluate the task in light of CURRENT main state.** Read the spec and the relevant files on origin/main NOW. The original task may have been written against a state of the code that no longer exists. 2. **Verify the task still advances SciDEX's aims.** If the system has evolved past the need for this work (different architecture, different priorities), close the task with reason "obsolete: " instead of doing it. 3. **Check if it's already done.** Run `git log --grep=''` and read the related commits. If real work landed, complete the task with `--no-sha-check --summary 'Already done in '`. 4. **Make sure your changes don't regress recent functionality.** Many agents have been working on this codebase. Before committing, run `git log --since='24 hours ago' -- ` to see what changed in your area, and verify you don't undo any of it. 5. **Stay scoped.** Only do what this specific task asks for. Do not refactor, do not "fix" unrelated issues, do not add features that weren't requested. Scope creep at this point is regression risk. If you cannot do this task safely (because it would regress, conflict with current direction, or the requirements no longer apply), escalate via `orchestra escalate` with a clear explanation instead of committing.

Git Commits (8)

[Verify] Senate DB integrity check — PASS (PostgreSQL era) [task:61a03e15-65fc-4992-a261-bc17e81d12a3]2026-04-23

[Verify] DB integrity check — FAIL: 100+ page errors, 65 FK violations, no backups [task:2324c574-262e-4cf6-8846-89ff1d59ca1e]2026-04-17

[Senate] Add DB integrity repair script; fix 52 self-loops, 5 invalid gap_ids, 114 orphaned artifact_links [task:2324c574-262e-4cf6-8846-89ff1d59ca1e]2026-04-10

[Senate] Log DB integrity check findings: backup missing, 192 orphaned edges, 5 self-loops, 5 invalid gap_ids [task:2324c574-262e-4cf6-8846-89ff1d59ca1e]2026-04-10

[Senate] Add FK consistency verification results [task:2324c574-262e-4cf6-8846-89ff1d59ca1e]2026-04-10

[Senate] Fix NameError: restore get_db() at wiki-quality endpoints [task:2324c574-262e-4cf6-8846-89ff1d59ca1e]2026-04-10

[Senate] Add DB integrity repair script; fix 52 self-loops, 5 invalid gap_ids, 114 orphaned artifact_links [task:2324c574-262e-4cf6-8846-89ff1d59ca1e]2026-04-10

[Senate] Log DB integrity check findings: backup missing, 192 orphaned edges, 5 self-loops, 5 invalid gap_ids [task:2324c574-262e-4cf6-8846-89ff1d59ca1e]2026-04-10

Spec File

[Senate] CI: Database integrity check and backup verification

Task ID: aa1c8ad8-f157-4e03-941d-9fdb76dbb12a Type: Recurring (daily) Priority: 58 Layer: Senate (governance, quality gates)

Objective

Run automated database health checks to ensure data integrity and backup availability.

Success Criteria

☐ PRAGMA integrity_check passes on PostgreSQL

☐ Latest backup file exists and is less than 24 hours old

☐ Report any issues found

Implementation Notes

Run PRAGMA integrity_check on main database

Check for backup files in standard backup location

Verify most recent backup timestamp

Log results

Work Log

2026-04-04 05:30 PDT - Starting integrity check

Starting daily database integrity and backup verification.

2026-04-04 05:33 PDT - Completed ✅

Database Integrity: PASSED

PRAGMA integrity_check returned "ok"
Main database: 59MB at postgresql://scidex

Backup Verification: PASSED

Latest backup: scidex-20260404T123001Z.db.gz (20MB compressed)
Created: 2026-04-04 05:30 PDT (< 15 minutes ago)
Backup compression: valid (gunzip -t passed)
Total backups maintained: 108 files with automated rotation
Backup frequency: every 15 minutes via cron

Notes:

Empty .backup files in main directory are legacy/stale - real backups are in /data/backups/sqlite/
All health checks passed successfully

2026-04-06 04:05 PDT - Completed ✅

Database Integrity: PASSED

PRAGMA integrity_check returned "ok"
Main database: 1.9GB at postgresql://scidex

Backup Verification: PASSED

Latest backup: scidex-20260406T110001Z.db.gz (356MB compressed)
Age: 4 minutes old (< 24 hours threshold)
Backup compression: valid (gunzip -t passed)
Backup frequency: every 5 minutes via cron

Notes:

All health checks passed successfully

2026-04-06 17:45 PDT - Completed ✅

Database Integrity: PASSED

PRAGMA integrity_check returned "ok"
Main database: 1.9GB at postgresql://scidex

Backup Verification: PASSED

Latest backup: scidex-20260407T004001Z.db.gz (360MB compressed)
Age: ~5 minutes old (< 24 hours threshold)
Backup compression: valid (gunzip -t passed)
Backup frequency: every 5 minutes via cron

Notes:

All health checks passed successfully

2026-04-08 11:30 PDT - Completed ✅

Database Integrity: PASSED

PRAGMA integrity_check returned "ok"
Main database: 2.3GB at postgresql://scidex

Backup Verification: PASSED

Latest backup: scidex-20260409T012501Z.db.gz (514MB compressed)
Age: ~4 minutes old (< 24 hours threshold)
Backup compression: valid (gunzip -t passed)
Total backups maintained: 1,229 files
Backup frequency: every 5 minutes via cron

Notes:

All health checks passed successfully

2026-04-09 06:02 UTC - Completed ✅

Database Integrity: PASSED

PRAGMA integrity_check returned "ok" for both PostgreSQL (2.2GB) and orchestra.db (16MB)

Backup Verification: PASSED

Latest scidex backup: scidex-20260409T060001Z.db.gz (494MB compressed, 1 min old)
Latest orchestra backup: orchestra-20260409T060001Z.db.gz (2.6MB, 0 min old)
Total backups maintained: 2,581 files
Backup frequency: every 5 minutes via cron

S3 Offsite Sync: PASSED (resolved from prior warning)

Previous run flagged S3 as 112h stale — now resolved
Last sync completed 2026-04-08 04:52 PDT (66.8 GiB snapshot)

Row Counts: Stable (184 analyses, 333 hypotheses, 688K edges, 15.8K papers)

Notes:

NeuroWiki DB remains a 0-byte file — no data to backup (wiki content lives in PostgreSQL)
All critical health checks passed, 0 warnings

2026-04-09 06:06 UTC - Completed ✅

Database Integrity: PASSED

PRAGMA integrity_check returned "ok" for both PostgreSQL (2.2GB) and orchestra.db (16MB)

Backup Verification: PASSED

Latest scidex backup: scidex-20260409T060501Z.db.gz (517MB compressed, 1 min old)
Latest orchestra backup: orchestra-20260409T060001Z.db.gz (2.6MB, 6 min old)
Backup compression: valid (gunzip -t passed)
Total backups maintained: 2,583 files
Backup frequency: every 5 minutes via cron

S3 Offsite Sync: PASSED

Daily sync cron running at 04:00 UTC
Last completed sync: home-20260408T101701Z (66.8 GiB snapshot)
Syncs running daily and completing successfully

Row Counts: Stable (184 analyses, 333 hypotheses, 688K edges, 15.9K papers, 123 gaps)

Notes:

NeuroWiki DB not present on disk (wiki content lives in PostgreSQL wiki_pages table)
All critical health checks passed, 0 warnings

2026-04-09 06:08 UTC - Completed ✅

Database Integrity: PASSED

PRAGMA integrity_check returned "ok" for both PostgreSQL (2.2GB) and orchestra.db (16MB)

Backup Verification: PASSED

Latest scidex backup: scidex-20260409T060501Z.db.gz (517MB compressed, ~3 min old)
Latest orchestra backup: orchestra-20260409T060501Z.db.gz (2.6MB, ~2 min old)
Backup compression: valid (gunzip -t passed)
Total backups maintained: 2,565 files
Backup frequency: every 5 minutes via cron

S3 Offsite Sync: PASSED

Daily sync cron running at 04:00 UTC
Last completed sync: home-20260408T101701Z (66.8 GiB snapshot)
Syncs running daily and completing successfully

Row Counts: Stable (184 analyses, 333 hypotheses, 688K edges, 15.9K papers)

Notes:

All critical health checks passed, 0 warnings

2026-04-09 06:11 UTC - Completed ✅

Database Integrity: PASSED

PRAGMA integrity_check returned "ok" for PostgreSQL (2.2GB)
PRAGMA integrity_check returned "ok" for orchestra.db (16MB at /home/ubuntu/Orchestra/orchestra.db)
Note: /home/ubuntu/scidex/orchestra.db is a 0-byte stub; real Orchestra DB is at /home/ubuntu/Orchestra/

Backup Verification: PASSED

Latest scidex backup: scidex-20260409T061001Z.db.gz (419MB compressed, ~0 min old)
Latest orchestra backup: orchestra-20260409T060501Z.db.gz (2.5MB, ~5 min old)
Backup compression: valid (gunzip -t passed on scidex-20260409T060501Z.db.gz)
Config backups: env snapshots current (env-20260409T061001Z.tar.gz)
Total backups maintained: 2,585 files
Backup frequency: every 5 minutes via cron

S3 Offsite Sync: PASSED

Daily sync cron running at 04:00 UTC
Last completed sync: home-20260408T101701Z (66.8 GiB snapshot)
5 consecutive daily syncs completed successfully (Apr 4–8)

Row Counts: Stable

184 analyses, 333 hypotheses, 15,873 papers, 123 knowledge gaps, 30 kg_edges

Notes:

All critical health checks passed, 0 warnings

2026-04-10 20:45 UTC - FK Consistency Verification ✅ (task:2324c574-262e-4cf6-8846-89ff1d59ca1e)

Database Integrity: PASSED

PRAGMA integrity_check returned "ok"
Main database: PostgreSQL at postgresql://scidex

FK Consistency Checks: PASSED (minor notes)

debate_rounds → debate_sessions: 0 orphaned
debate_sessions → analyses: 0 orphaned
market_transactions → hypotheses: 0 orphaned
price_history → hypotheses: 0 orphaned
knowledge_edges → analyses: 192 orphaned (edges from deleted/missing analyses - not recoverable via FK repair)

Data Quality Issues Found: 5 minor issues (non-critical, all historical/test data)

1 hypothesis with null target_gene — hyp_test_0a572efb (test record, archived)

1 analysis with null triggered_by — SDA-2026-04-02-gap-senescent-clearance-neuro (archived)

71 agent_performance records orphaned — agent_id="falsifier" not in actor_reputation (historical)

192 knowledge_edges with orphaned analysis_id — from debates where source analysis was removed (not recoverable)

4 duplicate analysis titles — from merged analyses (not data corruption)

Score Validation: PASSED

0 hypotheses with composite_score out of [0,1] range
0 knowledge_gaps with priority_score out of [0,1] range

Row Counts: Stable

188 analyses, 333 hypotheses, 127 knowledge_gaps
688,359 knowledge_edges, 586 debate_rounds, 116 debate_sessions
15,929 papers, 13,640 wiki_entities, 17,435 wiki_pages
930 agent_performance records (71 orphaned), 8,327 market_transactions

Notes:

All issues are historical data (test records, archived items, merged duplicates)
No active data corruption detected
The 192 orphaned knowledge_edges reference analyses that were removed/renamed — these edges came from debate transcripts and cannot be recovered via FK repair
System integrity is sound — no fixes required for active data

2026-04-11 19:45 UTC — Database Integrity Check ✅

Database Integrity: PASSED

PRAGMA integrity_check returned "ok"
Main database: 3.6GB at postgresql://scidex

Row Counts: Stable

690,276 knowledge_edges, 335 hypotheses, 246 analyses

Backup Status: WARNING

/data/backups/sqlite/ directory is empty
Backup cron not found in crontab
This is a backup infrastructure issue, not a data integrity issue

Notes:

Database integrity is sound
Backup infrastructure needs attention (separate task recommended)

2026-04-12 11:05 UTC — Database Integrity Check ✅

Database Integrity: PASSED

PRAGMA quick_check returned "ok" for PostgreSQL (3.3GB — full integrity_check skipped due to size/timeout)
PRAGMA integrity_check returned "ok" for orchestra.db (36MB at /home/ubuntu/Orchestra/orchestra.db)

Backup Verification: PASSED

Latest scidex backup: scidex-20260412T180002Z.db.gz (665MB compressed, 3 min old)
Latest orchestra backup: orchestra-20260412T180002Z.db.gz (7.0MB, 2 min old)
Backup compression: valid (gunzip -t passed on both)
Total backups maintained: 110 files in /data/backups/sqlite/
Backup frequency: every 5 minutes via cron

Row Counts: Growing normally

264 analyses, 364 hypotheses, 16,039 papers

Notes:

Backup infrastructure restored since 2026-04-11 warning — backups now running every 5 min
All critical health checks passed, 0 warnings

2026-04-12 19:27 UTC — Database Integrity Check ⚠️

Database Integrity: PASSED

PRAGMA quick_check returned "ok" for PostgreSQL (3.3GB main file + 1.97GB WAL)
PRAGMA integrity_check returned "ok" for orchestra.db (38MB copy via /tmp)

WAL Note: PostgreSQL WAL has 481,222 uncheckpointed frames (~1.97GB). PASSIVE checkpoint returned 0/481222 — active readers/writers are holding the WAL open. This is normal during heavy write activity but the WAL should be monitored; if it keeps growing it may indicate the API is not periodically checkpointing.

Backup Verification: WARNING

/data/backups/sqlite/ directory does not exist — /data mount point is absent
No backup cron job found in crontab -l
Last known backup was from the 11:05 UTC run (scidex-20260412T180002Z.db.gz)
Backup infrastructure has disappeared again (same issue as 2026-04-11)

Row Counts: Growing normally

264 analyses, 364 hypotheses, 3,324 knowledge_gaps (3,322 open)
700,954 knowledge_edges, 16,115 papers, 17,539 wiki_pages
150 debate_sessions, 807 debate_rounds

Action Required:

Backup cron and /data mount need to be restored
Consider filing a separate infrastructure task for persistent backup monitoring

2026-04-12 21:05 UTC — Database Integrity Check ⚠️

Database Integrity: PASSED

PRAGMA quick_check returned "ok" for PostgreSQL (3.21 GB main file + 1.99 GB WAL = 5.16 GB effective)
PRAGMA integrity_check returned "ok" for orchestra.db (37 MB, via /tmp copy)

WAL Status: GROWING — monitor required

WAL: 509,271 frames (~1.99 GB), up from 481,222 frames at 19:27 UTC (~1.5h ago, +28K frames)
PASSIVE checkpoint: (0 blocked, 506,293 wal_pages, 0 checkpointed) — active connections holding WAL open
WAL growth rate: ~18K frames/hour; if unchecked, WAL will consume significant disk space
Recommendation: API should call PRAGMA wal_checkpoint(RESTART) periodically when write load is low

Backup Verification: CRITICAL — no backups

/data mount point does not exist
No backup cron accessible (system cron and user crontab inaccessible in sandbox)
No .db.gz backup files found anywhere on disk
Backup infrastructure has been absent since 2026-04-11 19:45 UTC (two consecutive integrity check cycles)
This is the third consecutive run without backups — escalation warranted

Row Counts: Growing normally

265 analyses, 364 hypotheses, 700,954 knowledge_edges, 3,324 knowledge_gaps (3,322 open)
150 debate_sessions, 822 debate_rounds, 16,115 papers, 17,539 wiki_pages

Action Required:

Backup infrastructure still absent — urgent restoration needed
WAL checkpoint blocked by active connections — API should checkpoint periodically

2026-04-13 00:50 UTC — Slot 50 Fixes Applied ⚠️

Problem: Slot 50 agent identified two critical infrastructure issues:

/data mount missing → backup-all.sh failed silently (no backups for 2 days)

WAL growing unbounded (~18K frames/hour, now ~2 GB)

Fix 1: backup-all.sh fallback directory

Modified backup-all.sh to detect /data availability and fall back to

/home/ubuntu/scidex/backups/ when /data is not mounted

Backup script now creates backup directories with mkdir -p before use
Verified: script now creates /home/ubuntu/scidex/backups/sqlite/ when /data absent
Status: backups should resume immediately; cron job needs to be re-established

Fix 2: WAL checkpointing in API (api.py)

Added background thread at API startup that runs a passive WAL checkpoint

every 1 hour (PRAGMA wal_checkpoint(PASSIVE))

Uses PASSIVE mode — if there are concurrent readers/writers, checkpoint

returns 0 checkpointed pages without blocking

Thread is daemonized, starts when API process starts
WAL growth should now be bounded as checkpoint frames get merged into main DB

DB Status: Healthy

267 analyses, 373 hypotheses, 701,112 knowledge_edges
WAL ~2 GB (passive checkpoint will reduce on next idle window)
No data corruption or integrity issues

2026-04-12 22:51 UTC — Slot 42 ✅

Database Integrity: PASSED

PRAGMA quick_check returned "ok" for PostgreSQL (3.71 GB main + 0.00 GB WAL)
WAL checkpoint (PASSIVE): 1,384/1,384 frames checkpointed — WAL fully merged, no growth

Row Counts: Growing normally

273 analyses, 382 hypotheses, 701,115 knowledge_edges
3,324 knowledge_gaps (3,314 open), 159 debate_sessions, 860 debate_rounds
16,118 papers, 17,539 wiki_pages

Backup Verification: WARNING (persistent)

/data mount absent — no /data/backups/sqlite directory
/home/ubuntu/scidex/backups/sqlite/ does not exist (backup-all.sh fallback not yet triggered)
Crontab inaccessible in sandbox environment
backup-all.sh fix was committed by Slot 50 (2026-04-13 00:50 UTC) — backups should resume once cron is re-established

WAL Status: HEALTHY — fully checkpointed (was growing at ~18K frames/hour previously; now stable)

Notes:

Database integrity is sound, no data corruption detected
WAL issue resolved by API background checkpoint thread added by Slot 50
Backup infrastructure requires external verification (cron re-establishment outside agent scope)

---

Verification — 2026-04-17T10:50:00Z

Result: FAIL Verified by: MiniMax-M2 via task 2324c574-262e-4cf6-8846-89ff1d59ca1e

Tests run

Target	Command	Expected	Actual	Pass?
PRAGMA integrity_check	`sqlite3 PostgreSQL "PRAGMA integrity_check"`	ok	100+ btree page errors	✗
Analyses with invalid gap_id	`SELECT COUNT(*) FROM analyses WHERE gap_id NOT IN (SELECT id FROM knowledge_gaps)`	0	20	✗
Hypotheses with invalid analysis_id	`SELECT COUNT(*) FROM hypotheses WHERE analysis_id NOT IN (SELECT id FROM analyses)`	0	25	✗
Self-loop edges (partial)	`SELECT source_id,target_id FROM knowledge_edges WHERE source_id=target_id`	0	20+ (HIF1A\	HIF1A, partial — query returns error 11)	✗
NULL titles (analyses)	SQL count	0	0	✓
NULL titles (hypotheses)	SQL count	0	0	✓
NULL titles (knowledge_gaps)	SQL count	0	0	✓
composite_score out of [0,1]	SQL count	0	0	✓
priority_score out of [0,1]	SQL count	0	0	✓
API status endpoint	`curl -s -o /dev/null -w '%{http_code}' http://localhost:8000/api/status`	200	200	✓
Backup files	`find /home/ubuntu/scidex/backups /data/backups -name "*.db.gz" -mtime -1`	exists	0 files found	✗

Attribution

The current failing state reflects active DB corruption:

Commit ea69d990a ([Senate] Add DB integrity repair script) applied fixes but those changes were on an orphan branch and never reached origin/main (verified by e0b57349a)
Subsequent writes have re-introduced FK violations (20 analyses with invalid gap_ids, 25 hypotheses with invalid analysis_ids, 20+ self-loop edges)
Btree page corruption (100+ errors in PRAGMA integrity_check) has accumulated — likely from the 2026-04-17 corruption incident documented in AGENTS.md
API still responds (200) despite corruption — errors are in FTS/index pages, not primary data

Notes

DB size: 4.3 GB (4.1 GB main + 341 MB WAL) at postgresql://scidex
DB is actively in use (WAL mode, page_count=1,068,805)
Corruption type: B-tree page-level (out-of-order Rowids, 2nd references, invalid page numbers) concentrated in high page numbers (993xxx-103xxxx) — likely FTS index pages
FK violations are fixable by re-running scripts/repair_db_integrity.py logic (if the script existed on main — it doesn't; it was on an orphan branch)
Backup infrastructure is down (no /data mount, no /home/ubuntu/scidex/backups dir)
Escalation recommended: DB corruption needs repair + backup restoration

2026-04-20 17:35 UTC — Database Integrity Check ⚠️

Database Integrity: PASSED

PostgreSQL healthy: 3542 MB, 6 active connections, 0 lock waits, 0 long-running queries
Deadlocks: 15 (historical, not current)
Version: PostgreSQL 16.13
In recovery: false (primary, writable)

Row Counts (PostgreSQL scidex):

hypotheses: 747 | analyses: 395 | papers: 17,447
knowledge_edges: 711,721 | wiki_pages: 17,575

Backup Verification: CRITICAL — PostgreSQL backup infrastructure missing

Last scidex backup logged: 2026-04-13 (7 days ago, SQLite format, pre-migration)
backup-all.sh line 193: backup_sqlite postgresql://scidex scidex — FAILS: backup_sqlite uses sqlite3 CLI which cannot connect to PostgreSQL
backup_postgres.sh referenced in comments at backup-all.sh:256 but does not exist
No /data/backups/postgres/ directory and no pg_dump cron
backup_log confirms zero PostgreSQL backup entries since migration (2026-04-20)
Neo4j backup: last success 2026-04-19 02:00 UTC (1 day ago — PASS)
Orchestra DB: no recent backup entries in backup_log

Root Cause:
SQLite → PostgreSQL migration (2026-04-20) left backup infrastructure broken. The backup_sqlite function in backup-all.sh cannot back up PostgreSQL; backup_postgres.sh (pg_dump-based) was never created despite being referenced in comments.

Action Required (P1):

Create backup_postgres.sh using pg_dump to /data/backups/postgres/
Add pg_dump cron job (every 5 min or hourly)
Update verify_backup_integrity.sh to use pg_dump format for scidex
Note: backup-all.sh line 193 is dead code for PostgreSQL

Escalation: This is an infrastructure gap requiring a dedicated task. Recommend filing new backup_dr task for PostgreSQL backup implementation.

Payload JSON

{
  "requirements": {
    "coding": 8,
    "reasoning": 7,
    "safety": 9
  },
  "_watchdog_repair_task_id": "37fad167-17e0-4de0-a961-8f3c950d0411",
  "_watchdog_repair_created_at": "2026-04-17T09:07:09.239908+00:00",
  "completion_shas": [
    "bd579f91e"
  ],
  "completion_shas_checked_at": "2026-04-17T10:48:39.058065+00:00",
  "_reset_note": "This task was reset after a database incident on 2026-04-17.\n\n**Context:** SciDEX migrated from SQLite to PostgreSQL after recurring DB\ncorruption. Some work done during Apr 16-17 may have been lost.\n\n**Before starting work:**\n1. Check if the task's goal is ALREADY satisfied (run the relevant checks)\n2. Check `git log --all --grep=task:YOUR_TASK_ID` for prior commits\n3. If complete, verify and mark done. If partial, continue. If not done, proceed.\n\n**DB change:** SciDEX now uses PostgreSQL. `get_db()` auto-detects via\nSCIDEX_DB_BACKEND=postgres env var.",
  "_reset_at": "2026-04-18T06:29:22.046013+00:00",
  "_reset_from_status": "done"
}