[Feat] RAG Ingestion 5/11: Vector Store Management API

[yossiovadia]

Summary

Implement OpenAI-compatible REST endpoints for vector store CRUD operations.

Part of #1262

Scope

Files to create:

• pkg/apiserver/route_vectorstore.go - REST endpoint handlers
• pkg/apiserver/route_vectorstore_test.go

Files to modify:

• pkg/apiserver/server.go - Register new routes

Endpoints

Method	Path	Description
POST	/v1/vector_stores	Create vector store
GET	/v1/vector_stores	List vector stores (pagination)
GET	/v1/vector_stores/{id}	Get vector store details
POST	/v1/vector_stores/{id}	Update vector store
DELETE	/v1/vector_stores/{id}	Delete vector store + all data

Key Design Decisions

• VectorStoreManager orchestrates API calls + backend operations
• In-memory registry of vector stores (map[string]*VectorStore)
• Response format matches OpenAI API (object, id, created_at, etc.)
• Follows existing API server patterns (parseJSONRequest, writeJSONResponse)

Acceptance Criteria

• All 5 CRUD endpoints work correctly
• Pagination with limit/order/after/before on list
• Delete cascades to backend collection
• Error handling for missing/invalid IDs
• Tests for all endpoints

Dependencies

Depends on PR 1 (types), PR 3 (backends).

Branch: feat/rag-05-vectorstore-api → feat/rag-ingestion