firecrawl

mirror of https://github.com/firecrawl/firecrawl.git synced 2026-08-28 18:51:39 +08:00

Files

T

mogery c90a794514 perf(api): overlap requests-row insert with v2 search work, expose DB pool gauges

Two changes from the /v2/search longtail investigation:

1. v2/search: logRequest was awaited at the top of the controller, so every
   search paid the Postgres insert (queued behind the saturated pool —
   p90 ~6.6s) before any search work started. It's now kicked off
   un-awaited and awaited just before the logSearch child insert, keeping
   the request_id FK ordering (same pattern as the scrape controllers).
   robustInsert never rejects, so the await cannot throw.

2. Admin /metrics: expose db_pool_waiting_count / db_pool_idle_count /
   db_pool_total_count / db_pool_max_count gauges per application_name
   (firecrawl-api, firecrawl-api-rr, firecrawl-index) so pool saturation
   is visible in Prometheus instead of only via the throttled warn log.

2026-08-27 20:59:04 +02:00

api

perf(api): overlap requests-row insert with v2 search work, expose DB pool gauges

2026-08-27 20:59:04 +02:00

dot-net-sdk

feat(api): add opt-in prompt-injection check to JSON-mode extraction (#4392 )

2026-08-26 08:46:22 -03:00

elixir-sdk

feat(sdks): add agent effort param and trace/snapshot endpoints (#4415 )

2026-08-26 20:45:17 +02:00

go-html-to-md-service

fix(go-html-to-md-service): emit nanosecond-precision log timestamps (#3474 )