P2
Provider-adapter boundary normalizes OpenAI, Gemini, Claude, and local VLMs into one canonical result via strict structured output.
in progress CODE TEST HITL QA 💲 API COST
Satisfied by WI: WI-011 WI-056 WI-058 WI-077
P3
Output pipeline (syntax -> schema -> enum -> coordinate -> business-rule validation) with retry/fallback policy is defined and tested.
in progress CODE TEST
Satisfied by WI: WI-012 WI-059 WI-080
P4
Deterministic candidate ranking: the model returns evidence/component scores; TroveSnap computes the final ranking.
todo CODE TEST
Satisfied by WI: WI-013 WI-060
P7
MCP tool surface exposes read/search/candidate tools first, with approval gates on publish/bulk/delete tools, wrapping existing ingestion functions.
todo CODE TEST HITL QA
Satisfied by WI: WI-016 WI-065

Work items (5)

Code	Item	Status	Priority / risk	Owner	Gates
WI-011	Provider gateway + adapter boundary — core (registry, capability, schema+prompt compiler, image prep, ProviderAttempt) provider-adapter-boundary	done	P0 high risk	Paul	P2
WI-012	Output validation core — trust pipeline (stages 0–9, issue/receipt model, rule registry, deterministic repair, branded TrustedScanResult) validation-pipeline	done	P0 medium risk	Paul	P3
WI-013	Candidate ranking engine — deterministic features, fixed-point scoring, priority bands, tie-break, explainability, canonical hash deterministic-ranking	queued	P1 medium risk	Paul	P4
WI-015	Data authority & RLS foundation (tenants/memberships/roles/permissions/assignments, auth helpers, deny-by-default membership RLS, schema+migration contract, demo parity) data-model-layers	queued	P0 high risk	Paul	P6
WI-016	Governed MCP read/candidate surface (client registration, tool registry+schemas, permission-aware reads, P5 candidate tools, idempotency, audit, rate limits, injection safety) mcp-surface	queued	P1 medium risk	Paul	P7

Tracker v1.44.1 · generated 2026-06-23 17:52:26 UTC from docs/tracking/*.json. What's new