Knowledge Graph Endpoints

Entity CRUD, graph queries, context assembly, and impact analysis. The knowledge-graph-service runs on port 8003 and is proxied through the API gateway at http://localhost:8000.

Endpoints

Entity CRUD

POST /entities

Create a new entity vertex in the knowledge graph.

curl -X POST http://localhost:8000/entities \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "label": "Well",
    "properties": {
      "entity_id": "well-new-001",
      "well_name": "Smith Ranch #4",
      "api_number": "42-383-67890",
      "operator": "Permian Energy Inc.",
      "status": "drilling",
      "latitude": 31.9523,
      "longitude": -102.1245
    }
  }'

Valid labels: Well, Lease, Field, Formation, Operator, Permit, FlaringAuthorization, FlaringEvent, InfrastructureProject, Regulation, Wellpad, Facility, PipelineRoute, WellEvent, ComplianceDeadline, FilingPackage, WellConversation, Project, Equipment, Pipeline.

GET /entities/{label}

List all entities of a given type.

curl -H "Authorization: Bearer $TOKEN" \
  "http://localhost:8000/entities/Well"

GET /entities/{label}/{entity_id}

Get a specific entity by label and ID.

curl -H "Authorization: Bearer $TOKEN" \
  http://localhost:8000/entities/Well/well-001

Relationships

POST /edges

Create a relationship between two entities.

curl -X POST http://localhost:8000/edges \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "from_label": "Well",
    "from_id": "well-001",
    "to_label": "Lease",
    "to_id": "lease-001",
    "edge_label": "LOCATED_IN",
    "properties": {}
  }'

GET /edges/{label}/{entity_id}

Get all relationships for an entity.

curl -H "Authorization: Bearer $TOKEN" \
  "http://localhost:8000/edges/Well/well-001?direction=out"

Context Assembly

R34 Phase 4 unified both endpoints onto a single tenant-scoped graph walk (the context_assemble capability). The neighborhood traversal is one TenantScopedConnection round trip that returns each neighbor’s properties inline — no per-neighbor re-fetch, no default-graph dependency, no hardcoded per-section Cypher.

GET /context/assemble/managed/{entity_id}

The context_assemble capability. Returns two views from one walk:

• structured — an EntityContext: complete (every field; relationships by edge label, up to a 1000-per-relationship sanity ceiling). Consumed by the rules/governance layer.
• context_text — a truncated view for the LLM context window: a per-relationship cap (default 25) with explicit "…and N more … not shown" overflow markers, an 8K-token backstop (tiktoken cl100k_base proxy), and the domain-filtered significant/plain field split.

{entity_id} may be the AGE vertex uuid (what entity_resolve returns) or the business entity_id — the walk anchors on either. context_text and the R27 back-compat counts (fields_highlighted, detection_results_included, token_estimate, entities_referenced) stay top-level alongside the new structured / truncation_occurred / relationships_truncated fields.

Query params: entity_type_key (required), domain_tags (comma-separated), depth (default 1 — R34 groups first-hop), max_per_relationship (default 25), max_tokens (default 8000), tenant_id.

curl -H "Authorization: Bearer $TOKEN" \
  "http://localhost:8000/context/assemble/managed/well-001?entity_type_key=well&domain_tags=compliance"

GET /context/assemble/{well_api}

Legacy well-centric context by API number. R34 Phase 4 rewrote this as a thin adapter over the context_assemble capability: it resolves well_api → entity through the same tenant-scoped walk, then maps the result back into the legacy sections shape the hardcoded Rule 37/32 tools still read (R35 deletes the adapter and those tools together). No more api_number f-string on the default graph.

curl -H "Authorization: Bearer $TOKEN" \
  http://localhost:8000/context/assemble/42-383-12345

Graph Query

POST /query

Run a whitelisted graph query. R34 Phase 1 replaced the injection-prone raw-Cypher contract with a named-template registry: callers send a template_id plus param values (never Cypher). The request is resolved against the server-side whitelist (graph_query_templates) and executed through a TenantScopedConnection, so every query is tenant-scoped and injection-safe.

curl -X POST "http://localhost:8000/query?tenant_id=00000000-0000-0000-0000-000000000001" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "template_id": "operator_wells",
    "params": {"operator_id": "OP-0001"}
  }'

tenant_id is an optional query parameter (default: the dev tenant 00000000-0000-0000-0000-000000000001); it selects the per-tenant AGE graph namespace.

Registered templates:

Param kinds determine validation:

• value params (e.g. operator_id) are escaped via escape_cypher_string and interpolated as quoted literals.
• label params (e.g. edge_label) are whitelist-validated against EDGE_LABELS / VERTEX_LABELS and substituted raw — AGE cannot bind or escape a relationship type, so a value outside the whitelist is rejected (400), never substituted.

Errors return 400: unknown template_id, a param set that doesn’t exactly match the template, or a non-whitelisted label.

The response shape is unchanged — { "results": [ ... ], "count": N } keyed by the template’s columns.

Impact Analysis

GET /impact/{label}/{entity_id}

Bidirectional impact analysis — find all entities affected by changes to a given entity.

curl -H "Authorization: Bearer $TOKEN" \
  http://localhost:8000/impact/Well/well-001

GET /impact/tree

Get the infrastructure tree showing hierarchical relationships.

curl -H "Authorization: Bearer $TOKEN" \
  "http://localhost:8000/impact/tree?root_label=Wellpad&root_id=pad-001"

POST /seed

Seed the knowledge graph with demo data.

curl -X POST http://localhost:8000/seed \
  -H "Authorization: Bearer $TOKEN"