fix: vector_search() always returns empty — E2E tests + LangChain similarity_search()

[polaz]

Problem

CoordinodeClient.vector_search() always returns an empty list, even when nodes with embedding properties exist in the graph. This silently breaks:

• CoordinodePropertyGraphStore.vector_query() in llama-index adapter
• Any LangChain workflow relying on vector similarity

client = CoordinodeClient("localhost:7080")
results = client.vector_search(label="Chunk", property="embedding", vector=[...], top_k=5)
# → [] (always empty, even with data)

Root cause

This is a DB-side gap: VectorService gRPC endpoint exists and accepts requests, but the wiring from VectorService to the HNSW index executor is incomplete in CoordiNode v0.3. The RPC returns without querying the index.

Tracked as G073 in CoordiNode DB repository.

SDK work (after G073 is fixed in DB)

Once G073 is resolved and vector_search returns actual results:

1. Add E2E integration test — test_vector_search_returns_results with a real embedding vector
2. Add similarity_search() to LangChain adapter (CoordinodeGraph) — currently missing; should wrap client.vector_search() with label/property defaults
3. Verify vector_query() in LlamaIndex adapter — test against live DB, confirm scores are correct (currently returns r.distance — verify this matches expected similarity semantics)

Acceptance criteria

• client.vector_search(label="Chunk", property="embedding", vector=[...], top_k=5) returns results when data exists
• Integration test passes against live CoordiNode instance
• CoordinodeGraph.similarity_search(query_vector, k=5) implemented in LangChain adapter
• LlamaIndex vector_query() tested E2E

Gate

Blocked by G073 — requires CoordiNode DB fix: VectorService → HNSW executor wiring.
Do not start SDK implementation until a CoordiNode release with G073 fixed is available.

[coderabbitai]

🔗 Related PRs

#15 - fix: CoordiNode Cypher compatibility — add_graph_documents, type/label, MATCH+CREATE [closed]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[polaz]

Implemented in 062c235:

• similarity_search(query_vector, k, label, property) added to CoordinodeGraph
• Guards against empty vector (server returns INVALID_ARGUMENT for empty vectors)
• Two integration tests: seeded node found in top-k results + empty vector returns []

SDK E2E test (test_vector_search_returns_results) was already added in the previous test_vector_search_returns_results commit. LlamaIndex vector_query() E2E coverage added in test_vector_query_returns_results.