Operations
Status
Today's ingestion usage, limits and pending work.
Read from the same mechanisms the pipeline itself uses to enforce caps and plan ingestion. This is a server-rendered snapshot.
01 / Discovery quota
YouTube Data API
0% of today's recorded budget
Accounting note Backfill discovery runs are recorded here. discover_youtube_candidates.py is not yet wired into the ledger.
02 / Transcript collection
TranscriptAPI
0% of today's cap
Attempts are counted per UTC day from the collector's success and failure stores. Failed fetch attempts count toward the cap.
03 / Model usage
OpenRouter / AI Research
No API key is set, so requests cannot be sent.
Daily cap is 0: calls are disabled. No gauge is drawn.
Configured modelanthropic/claude-sonnet-5-20260630
04 / Pending work
Ingestion backlog
Not available
The current next-work selection could not be read. This does not mean zero.
Discovered candidates selected by the collector with no daily cap applied. The backlog has no independent ceiling, so there is nothing to gauge against.
Method
What is counted
- YouTube Data API
- Persisted estimated quota units from backfill runs recorded in the UTC-day ledger.
- TranscriptAPI
- Collector fetch attempts dated today in UTC, including recorded failures.
- OpenRouter
- Logged model-call attempts per UTC day, including failures; recorded spend includes estimated-cost rows where applicable.
- Backlog
- The collector's own uncapped next-work selection from the channel-upload backfill candidates and its existing success and failure stores.
Scope
Known accounting limits
- discover_youtube_candidates.py is not yet included in the YouTube quota ledger; the displayed figure may omit its use.
- Backlog has no independent ceiling and may exceed the daily transcript cap while the roster catches up. The collector works newest-first.
- “Not available” means the backlog selection cannot be read; it does not mean zero.
- Not a recommendation. A mention here is not a buy, sell, or hold signal, and is not endorsed, scored, or ranked by this tool in any way -- see every result for what a channel said, not what this product thinks about it.
- Automated speech-to-text and automated matching. Transcripts are machine-generated and mention detection is automated; both can misfire. Always confirm by watching the linked moment before relying on what a card says was said.
- Coverage is partial. Only the tracked channels and videos this product has actually processed are searchable. No result for a symbol in a window does not mean it was not discussed anywhere -- only that it was not found in what has been indexed so far.