Production report
Single-page view of what this site has actually probed, what has been verified, and what is still missing. Mirrors TRANSPARENCY.md and the live repo state. Refreshed on every build.
Executive summary
Tables probed
418
118 pass · 293 fail · 7 partial
Functions probed
10
7 pass · 0 fail
Coverage
28.2%
Across all sources
Rows captured
24,182
3,075 trials
Findings (F-IDs)
9
Stable named issues
History snapshots
443
Per-table append-only
Dated snapshots
8
Frozen states
Sources covered
10/46
of KNOWN_SOURCES
Per-source coverage
| Source | Catalog | T | F | pass | part | fail | coverage | status |
|---|---|---|---|---|---|---|---|---|
| opencode | present | 20 | — | 20 | — | — | 100.0% | probed |
| google_calendar | present | 5 | — | 5 | — | — | 100.0% | probed |
| twilio | present | 3 | — | 3 | — | — | 100.0% | probed |
| voyage_ai | present | 1 | — | 1 | — | — | 100.0% | probed |
| claude | present | 2 | — | 2 | — | — | 100.0% | probed |
| claude_code_sessions | present | 2 | — | 2 | — | — | 100.0% | probed |
| gmail | present | 14 | 3 | 12 | — | 2 | 70.6% | probed |
| render | present | 3 | — | 2 | — | 1 | 66.7% | probed |
| github | present | 364 | 7 | 71 | 7 | 286 | 21.0% | probed |
| chess | present | 4 | — | — | — | 4 | 0.0% | probed |
| notion | absent | — | — | — | — | — | — | queued |
| tavily | absent | — | — | — | — | — | — | queued |
| serper | absent | — | — | — | — | — | — | queued |
| exa | absent | — | — | — | — | — | — | queued |
| firecrawl | absent | — | — | — | — | — | — | queued |
| web_scraper | absent | — | — | — | — | — | — | queued |
| brightdata | absent | — | — | — | — | — | — | queued |
| zerops | absent | — | — | — | — | — | — | queued |
| assemblyai | absent | — | — | — | — | — | — | queued |
| deepgram | absent | — | — | — | — | — | — | queued |
| elevenlabs | absent | — | — | — | — | — | — | queued |
| groq_ai | absent | — | — | — | — | — | — | queued |
| ollama | absent | — | — | — | — | — | — | queued |
| cloudinary | absent | — | — | — | — | — | — | queued |
| absent | — | — | — | — | — | — | queued | |
| google_cloud_vision | absent | — | — | — | — | — | — | queued |
| google_sheets | absent | — | — | — | — | — | — | queued |
| cloudsync | absent | — | — | — | — | — | — | queued |
| rclone | absent | — | — | — | — | — | — | queued |
| azure | absent | — | — | — | — | — | — | queued |
| microsoft_graph_v4 | absent | — | — | — | — | — | — | queued |
| stream_video | absent | — | — | — | — | — | — | queued |
| kairon_live | absent | — | — | — | — | — | — | queued |
| acmecorp | absent | — | — | — | — | — | — | queued |
| acmetask | absent | — | — | — | — | — | — | queued |
| agentmail | absent | — | — | — | — | — | — | queued |
| blogger | absent | — | — | — | — | — | — | queued |
| careops_appointments | absent | — | — | — | — | — | — | queued |
| careops_doctor_chats | absent | — | — | — | — | — | — | queued |
| careops_family_notes | absent | — | — | — | — | — | — | queued |
| careops_lab_reports | absent | — | — | — | — | — | — | queued |
| careops_medications | absent | — | — | — | — | — | — | queued |
| careops_patients | absent | — | — | — | — | — | — | queued |
| careops_pharmacy_receipts | absent | — | — | — | — | — | — | queued |
| careops_prescription_ocr | absent | — | — | — | — | — | — | queued |
| careops_symptom_logs | absent | — | — | — | — | — | — | queued |
Findings (9)
critical
1
high
1
medium
4
low
3
critical
1 tablesF026Coral catalog mentions a column the schema doesn't havehigh
5 tablesF022This surface requires a GitHub App JWTmedium
5 tablesF023GitHub returns 404 for these surfacesmedium
3 tablesF024GitHub returns 404 for these surfacesmedium
3 tablesF025GitHub rejects these queries as 422medium
3 tablesF028GitHub rejects these queries as 422low
5 tablesF020GitHub sunset the underlying endpointlow
5 tablesF021GitHub rate-limits this surfacelow
5 tablesF027Other failure modeDated snapshots (8)
twilio·2026-09-02— v0.5.0: twilio first probe — all 3 tables (account, calls, messages) pass cleanly. Trial account, all empty required_filters. Live data: 1 account row, 1 call (failed inbound), 17 messages (most outbound-api from +12674406256). 17/17 trials on account + calls + messages, 0 errors.3 tables · 0 fns · 2026-09-02T04:30:04voyage_ai·2026-09-02— v0.5.0: voyage_ai first probe — embeddings pass cleanly (4 variants × 4 trials = 16 calls, 0 errors after rate-limit backoff). Live Voyage AI API, per-call endpoint; tight free-tier rate limit, so 15s backoff between trials is mandatory.1 tables · 0 fns · 2026-09-02T04:58:10gmail·2026-09-01.1— v0.5.1: gmail message_attachments fix — re-probed with real DigitalOcean invoice PDF attachment (message_id=1a05c918ce469e58, attachment_id=ANGjdJ-Kl_..., size=82541 bytes) harvested live 2026-09-01. 12/14 pass; only message_headers (coral_bug.column_missing) and settings_filters (Gmail upstream EOF) remain documented Coral/Gmail bugs.14 tables · 0 fns · 2026-09-02T03:38:22opencode·2026-09-01— v0.5.0: opencode re-probe — all 20 tables (10 original + 10 v2: account_states, accounts, control_accounts, credentials, data_migrations, event_sequences, events, migrations, permissions, session_context_epochs). 20/20 pass with median latency populated via summary fallback. retry on transient registration; verdict fixed for empty tables.20 tables · 0 fns · 2026-09-01T12:28:26google_calendar·2026-09-01— v0.5.0: google_calendar re-probe — all 5 tables (calendars, events, calendar_colors, event_colors, settings). 5/5 pass. local Google Calendar API source, no token scope needed; empty required_filters.5 tables · 0 fns · 2026-09-01T13:12:52gmail·2026-09-01— v0.5.0: gmail re-probe — 14 tables: 11 pass (6 unfiltered + 5 id-filtered), 3 fail documented as known issues (settings_filters EOF upstream, message_headers coral_bug.column_missing, message_attachments needs real attachment_id from payload scan). Probe runner captures these honestly instead of hiding them.14 tables · 0 fns · 2026-09-01T14:14:32github·2026-08-31— v0.5.0: full re-probe with rows[], metadata.coral_version, history.* on every table364 tables · 7 fns · 2026-08-31T09:39:13github·2026-08-30— v0.4.0: findings + coverage + snapshots + token_runs + unlock_matrix + /findings + /timeline + /security/credentials + KNOWN_GAPS + attribution tests + SKILL.md364 tables · 7 fns · 2026-08-30T14:44:42
What this report is missing (production audit)
A reviewer arriving at this site should be able to answer any question from the data alone. These are the gaps that remain.
lowgithub coverage is 21.0% (71/364)
Most failures are upstream 404/rate-limit, fixable with better test data.
low1 critical finding(s) — review /open
Findings are stable F-IDs you can link to from PRs and issues.
Always-on improvements (regardless of source)
- Cross-source comparison page. Side-by-side: which sources have the best coverage, which are subtly broken, which have private-only resources. Pulls from
coverage.json. - Per-trial row diff. For a probe with 200 rows, render a paginated table view (sortable, filterable, column-show/hide). The data is already in
rows[]. - Coral version timeline.
metadata.coral_versionis already captured per probe. Plot pass-rate by coral version to catch regressions. - Per-row “type inference + stats” panel. For each column returned, show dominant type, null count, sample values, distinct count. Already implemented in
ColumnProfile; needs wider placement. - Auto-classify retry outcomes. When a retry flips a verdict from fail → pass, log a “fix” event with the filter value that worked. Already started in
retry_batch.py. - Lighthouse / web-vitals page. Render performance of the static site. Currently measured ad-hoc.
- Per-source “how to read this” guide. On the source page, one paragraph explaining what tables mean + which `xxx_*.history` versions to compare. Especially helpful for sources without probe data yet.
- Rate-limit dashboard. Token-by-token rate-limit usage over time. Helps diagnose the 77 rate-limited tables on github.
What you can do next
If you want to extend coverage
- Pick a source from /sources (40 queued)
- File
[claim]issue - Run
catalog_export.py & probe_batch.py - Push PR — site auto-updates
If you want to fix the github report
- Open coral bug for F026 (column_missing)
- Delete 5 dead Classroom tables from coral
- Mark 10 App-JWT-required tables in coral catalog
- Use known-good filter values per table
- Switch PAT → App JWT to clear the 60 req/hr wall
This page is generated at build time from
reports/ JSON.