support graph store & graph_rag module

[xucailiang]

For Local Development (Optional - With Neo4j)

# Start Neo4j with GDS plugin
docker run -d --rm \
  --name neo4j \
  -p 7474:7474 -p 7687:7687 \
  -e NEO4J_AUTH=neo4j/password \
  -e "NEO4J_PLUGINS=[\"graph-data-science\"]" \
  neo4j:latest

Running Example(set DASHSCOPE_API_KEY first)

python examples/graph_rag/graph_knowledge_example.py

🚀 Running Tests

# Run tests with Neo4j
NEO4J_AVAILABLE=true pytest tests/graph_rag/ -v

# Run all tests (mock mode - works in GitHub Actions)
pytest tests/graph_rag/ -v

# Skip tests requiring Neo4j GDS plugin
pytest tests/graph_rag/ -v -m "not requires_gds"

📝 Note on GDS Community Detection Tests

Community detection tests require real API models and are not included in the test suite.

To test GDS features: Use examples/graph_rag/graph_knowledge_example.py with real DashScope API credentials.

[cla-assistant]

All committers have signed the CLA.

[xucailiang]

Do we have any progress on this? I can provide more documentation to assist with the review if needed. Our team has been using a temporary repository for some time and would like to switch back to the main Agentscope repository for feature updates.
Thank you for your efforts. @qbc2016