fix: CoordiNode Cypher compatibility bugs in langchain and llama-index adapters

[polaz]

Problem

E2E testing against v0.4.3 revealed four compatibility bugs discovered by running the README examples against a live CoordiNode instance:

langchain-coordinode

1. add_graph_documents() not implemented — inherited pass from GraphStore base, all graph data silently discarded
2. refresh_schema() uses unsupported Cypher functions — labels(n) and type(r) both return null in CoordiNode; must use n.__label__ and r.__type__ internal properties

llama-index-graph-stores-coordinode

3. upsert_relations() uses unsupported MERGE syntax — MERGE (src)-[r:TYPE]->(dst) fails with "MERGE create from non-NodeScan pattern"; replaced with MATCH+MATCH+CREATE
4. get_triplets() / get_rel_map() use type(r) — returns null; must use r.__type__

Fix

• Implement add_graph_documents() in CoordinodeGraph with MERGE nodes + CREATE edges (with fallback for servers that don't support WHERE NOT pattern guards)
• Replace all labels(n), type(r) with n.__label__, r.__type__
• Replace MERGE edge pattern with MATCH+CREATE in upsert_relations()

Verification

All 15 e2e tests pass after the fix (core connect, langchain schema/query/add_graph_documents, llama-index upsert/get_triplets/structured_query/from_existing).

[coderabbitai]

🔗 Related PRs

#2 - fix: SDK API correctness — constructor, cypher result, VectorResult fields [merged]
#12 - fix(ci): install workspace packages as editable + fix lint [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!