Loading...
The NextBid nightly audit surface lands here once the Spark-only orchestrator clears its bake on the existing 5 audit systems. Until then, this page is a placeholder so the route reserves real estate and the tile on the Dev Tools hub is honest about what is in flight.
Local Ollama on Spark via LiteLLM. External APIs may enrich, never load-bear.
Standalone audit operator on Spark — no House dependency, no OpenClaw.
Runs ahead of the 2-3am House lane. Phase -1 warms models before real audit work begins.
First 2 weeks read-only — establish baseline, then promote out of shadow.
| Tab | Purpose |
|---|---|
| Today | Latest run rollup, system status table, PUNCHLIST, staleness indicator |
| Scoreboard | 6 frozen SQL-view metrics — same numbers every day, no drift |
| History | Day-over-day ledger browser, diff view, last 14 nights |
| Phases | Phase-gate state per run (scan / deep-scan / diff / synth / WO-drop / post) |
| Shadow WOs | Improvement WOs queue — SHADOW vs ACTIONABLE, promote on review |
| Config | Per-phase model routing via LiteLLM, SHADOW_MODE toggle, OpenAI budget cap |
nextbid_corenb_audit_operator container — LiteLLM-first, no OpenClaw, no Claude