AI Automation Mainline Work Items

Updated: 2026-07-09 Scope: EwoooC / MOMO Pro System AI automation, PixelRAG, MCP/RAG, production UX.

Source Of Truth

Production readback stays authoritative after deploy: /health, container health, AI automation API readback, and smoke.
Gitea is the only source-control truth. GitHub remains frozen and must not be used.
Runtime truth must stay separate from source truth: a green test or registry does not mean MCP/RAG runtime is enabled.
PixelRAG visual receipts are evidence, not formal price truth. They cannot write competitor_prices, competitor_price_history, or ai_insights without replay and promotion gates.

Status	Work item	Evidence / next machine action
Completed	Multi-commerce PixelRAG visual evidence lane for momo, pchome, shopee_tw, coupang_tw, yahoo_shopping_tw, etmall_tw, friday_tw, rakuten_tw	API and CLI can emit expansion plans and marketplace manifests; PChome parser failures can queue manifests.
Completed	External MCP/RAG capability inventory absorbed into internal governance readback	`/api/ai-automation/external-mcp-rag-integration` and `scripts/ops/report_external_mcp_rag_integration.py` expose 9 capabilities, absorbed/unresolved counts, and runtime flags.
Completed	PixelRAG receipts to internal RAG candidate replay	`/api/ai-automation/pixelrag-rag-candidate-replay` and `scripts/ops/report_pixelrag_rag_candidate_replay.py` read receipts, split eligible vs blocked, and require OCR/VLM replay plus PromotionGate before knowledge writes.
In progress	MCP/RAG runtime health in AI automation smoke	`/api/ai-automation/smoke` and `/api/ai-automation/scheduled-health-summary` include external MCP/RAG integration, PixelRAG RAG candidate replay, and PixelRAG OCR/VLM replay contract families.
In progress	Formal production deploy/readback discipline	Every mainline change must update version, push Gitea main/dev, deploy to 188 without touching `momo-db`, and read back `/health` plus new endpoints.

Status	Work item	Evidence / next machine action
Completed	OCR/VLM replay contract for visual fields	`/api/ai-automation/pixelrag-ocr-vlm-replay` and `scripts/ops/report_pixelrag_ocr_vlm_replay.py` turn saved tiles into no-write field contracts, output schemas, validation rules, and Ollama-first worker actions.
Not started	Ollama-first OCR/VLM extraction worker	Execute the replay contract against approved local multimodal/OCR models and emit confidence/evidence receipts without writing formal price truth.
Not started	Ollama-first multimodal embedding benchmark	Verify local Qwen3-VL or equivalent visual embedding on GCP-A -> GCP-B -> 111 before any visual vector retrieval.
Not started	pgvector-compatible visual evidence metadata	Design metadata-first retrieval without FAISS in production unless ADR approves a different store.
Not started	Coupang platform probe / structured API strategy	Treat 403 as platform barrier; prefer public structured data or approved probe, never count blocked pages as product data.
Not started	Marketplace source contracts beyond MOMO/PChome	Promote stable Shopee/Coupang/Yahoo/ETMall/friday/Rakuten receipts into source contracts only after provenance, rate-limit, public boundary, and replay evidence.
In progress	Professional product website UI/UX text reduction	Continue compact AI surfaces, visual QA, and user-facing copy guardrails; avoid engineering logs or work-session prose in the UI.

Manual review is exception-only for low-risk, machine-verifiable controlled apply.
Break-glass still applies to secrets, destructive DB actions, force pushes, production provider switchovers, raw runtime volumes, and unverified core model replacement.
AI automation smoke must expose health honestly: MCP_ROUTER_ENABLED=false or RAG_ENABLED=false is a runtime gap, not a source-code failure.