storage-backends
Choose, configure, validate, and troubleshoot nano-graphrag storage backends for KV data, vector indexes, graph persistence, Neo4j, and safe GraphML visualization.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- storage-backends
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Choose, configure, validate, and troubleshoot nano-graphrag storage backends for KV data, vector indexes, graph persistence, Neo4j, and safe GraphML visualization.
- disable-model-invocation
- true
02 资源文件 5
- 参考文档references/graphml-visualization.md3.0 KB
- 参考文档references/storage-backends.md13.2 KB
- 参考文档references/troubleshooting.md7.8 KB
- 脚本scripts/storage_smoke.py11.5 KB
- 脚本scripts/visualize_graphml.py10.4 KB
03 SKILL.md 正文
Storage Backends
Use this sub-skill when a user needs to choose a nano-graphrag storage class, configure GraphRAG storage arguments, validate local storage persistence, adapt a third-party vector database, set up Neo4j graph storage, or inspect a GraphML knowledge graph.
Do not use this sub-skill for provider-specific LLM or embedding implementation details; storage docs only state the vector embedding_func contract and should hand off actual provider code to the sibling provider-and-model-integrations skill. Do not use it for the high-level insert/query workflow; route that to the sibling core-graphrag-workflows skill.
Runtime references and scripts
- Read references/storage-backends.md when choosing among
JsonKVStorage,NanoVectorDBStorage,HNSWVectorStorage,NetworkXStorage,Neo4jStorage, or when implementing a FAISS/Milvus/Qdrant-style vector adapter. - Read references/graphml-visualization.md when the user has a GraphML file and wants JSON export or an optional local browser preview.
- Read references/troubleshooting.md when storage initialization, persistence, HNSW capacity/search, Neo4j/GDS, empty graph clustering, or vector adapter behavior fails.
- Run scripts/storage_smoke.py to instantiate NetworkX plus local vector storage with a fake embedding function and no provider API calls.
- Run scripts/visualize_graphml.py to convert a user-provided GraphML file to node-link JSON and, only if requested, generate/serve a local HTML preview.
Fast routing
- Default local file-backed setup: keep
JsonKVStorage,NanoVectorDBStorage, andNetworkXStorage; use a freshworking_dirfor an independent index. - Larger/faster local vector search: set
vector_db_storage_cls=HNSWVectorStorageand tunevector_db_storage_cls_kwargsbefore inserting data. - External graph service: set
graph_storage_cls=Neo4jStorageand provideaddon_params={"neo4j_url": ..., "neo4j_auth": (...)}after confirming Neo4j 5.x plus the GDS plugin are running. - Third-party vector database: subclass
BaseVectorStorage, preserveupsert,query, andindex_done_callback, return ranked results containing source ids plus any required metadata. - If a storage symptom is caused by LLM/entity extraction producing no graph, route to the sibling customization-and-troubleshooting skill after confirming the graph backend itself is healthy.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).