Catalog & PIM
Headless product CRUD with administrative UI, variant management, and editorial workflows.
ContextCommerce
ContextCommerce provides the commerce domain for ContextUnity — a Django-based Product Information Management (PIM) platform with AI-powered enrichment, matching, and taxonomy.
Harvester Pipelines
Supplier data ingestion via pluggable fetchers and format transformers.
AI Enrichment
Product descriptions, taxonomy classification, and semantic matching orchestrated via ContextRouter.
MCP Server
FastMCP tools for AI-driven product search, matching, enrichment, and supplier operations.
Architecture Overview
ContextCommerce operates as a domain-specific extension that connects to the core ContextUnity service mesh. It delegates heavy AI processing to ContextRouter, stores taxonomy and product vectors in ContextBrain, and utilizes ContextWorker for durable background execution.
Manifest-Driven Integration
Commerce uses a multi-graph manifest in contextunity.project.yaml to declare 5 AI pipelines:
| Graph | Template | Purpose |
|---|---|---|
gardener | yaml:gardener | Taxonomy classification + field extraction |
enricher | yaml:enricher | AI-driven product enrichment (NER, descriptions) |
writer | yaml:enricher | Product description generation (reuses enricher template) |
rlm_bulk_matcher | yaml:rlm_bulk_matcher | RLM-based bulk product matching |
rag | yaml:retrieval_augmented | Universal RAG for product Q&A |
Each graph declares its own federated tools — domain handlers that run inside the Commerce extension and are called back by Router via BiDi gRPC. Commerce has zero dependency on contextunity-router — all AI logic is delegated through the manifest.
See Graph Compiler & Templates for template details.