Implement first knowledge engine runtime slice

docs/service-api-boundary.md

@@ -0,0 +1,42 @@
+# Service API Boundary
+
+Date: 2026-05-05
+
+## Decision
+
+The first service boundary will be an optional FastAPI layer over the
+programmatic contracts implemented in `kontextual_engine`. The service must not
+define separate business behavior; it should expose the same artifact,
+collection, relationship, ingestion, query, workflow, and context operations as
+HTTP resources.
+
+## First Resource Shape
+
+Planned endpoint groups:
+
+- `GET /health`
+- `POST /collections`, `GET /collections`, `GET /collections/{id}`
+- `POST /artifacts`, `GET /artifacts/{id}`, `GET /artifacts`
+- `POST /relationships`, `GET /relationships`
+- `POST /ingest`
+- `POST /query/artifacts`, `POST /query/relationships`
+- `POST /runs`, `GET /runs/{id}`, `GET /runs/{id}/manifest`
+- `POST /context/artifact/{id}`
+
+## Rules
+
+- The Python API is canonical until contracts stabilize.
+- HTTP handlers must translate request/response envelopes only.
+- Markdown parsing, document transforms, selectors, contracts, and local
+  indexes must remain adapter calls to `markitect-tool`.
+- Provider calls must remain adapter calls to `llm-connect`.
+- Structured diagnostics and errors must be preserved in responses.
+
+## Deferred
+
+- Authentication and authorization.
+- Durable database backend.
+- OpenAPI model polishing.
+- Streaming run execution.
+- Provider-backed assisted steps.
+