Cost notes: gemini-3.5-flash output is 6x input; run #543 re-priced (~$7.50, not $2.02)

Cost "f": price the agent at the model it runs; thinking/vision token breakdown; model switch; near-duplicate gate

Cost: bounded Gemini thinking on every call (default low), figure mining off by default; one-shot config migration

Never store a credential: redact at every DB write and at boot; keep PDFs when Gemini refuses for budget reasons

Gemini token accounting: usageMetadata -> Extraction, VisionStats, PdfResult

pipeline sync: figure citation linking (figure_links.py, --link-figures) from the pipeline repo

Figure & graph mining in the autonomous cycle: figures.py + Postgres port, crops stored in DB, review UI

Anti-hang hardening: TCP keepalives + statement timeout on PG connections, cap Retry-After sleeps, sweep stale 'running' runs at boot and before each cycle

Initial deploy: autonomous ingestion agent (plan→discover→dedupe→download→ingest→report)