configuration-data
Initialize, inspect, and troubleshoot GraphRAG settings, environment interpolation, auth, input readers, storage/cache/table providers, vector stores, and offline data readiness before indexing or querying.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- configuration-data
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Initialize, inspect, and troubleshoot GraphRAG settings, environment interpolation, auth, input readers, storage/cache/table providers, vector stores, and offline data readiness before indexing or querying.
- disable-model-invocation
- true
02 资源文件 5
- 参考文档references/config-schema-and-env.md4.1 KB
- 参考文档references/input-storage-cache-vector.md4.7 KB
- 参考文档references/troubleshooting.md5.7 KB
- 脚本scripts/inspect_config.py5.4 KB
- 脚本scripts/smoke_input_readers.py5.9 KB
03 SKILL.md 正文
GraphRAG Configuration and Data
Use this sub-skill when a task mentions settings.yaml, .env, GraphRagConfig, Azure/OpenAI auth, LiteLLM model settings, input file schemas, input_storage, output_storage, cache, table_provider, vector_store, LanceDB, Azure AI Search, Cosmos DB, or safe checks before running an index.
Route Boundaries
- Stay here for initialization, config loading, schema validation, provider selection, secret-safe summaries, and offline input reader checks.
- Route full index execution, update runs, workflow selection, and interpreting
PipelineRunResultoutputs to../indexing/. - Route completed-index DataFrame loading and search calls to
../querying/. - Route prompt generation and
graphrag prompt-tuneusage to../prompt-tuning/. - Route custom reader/provider implementation, factory registration, and package-extension design to
../package-extensions/.
Fast Workflow
- Create or locate a project root containing
settings.yaml,settings.yml, orsettings.json;graphrag init --root <project>generates the defaultsettings.yamland.envskeleton. - Fill
completion_modelsandembedding_modelswith LiteLLM-compatiblemodel_provider,model,auth_method, and eitherapi_keyor Azure managed identity fields. - Set
input.typeand reader columns, then matchinput_storage.typeandinput_storage.base_diror remote storage settings to the data source. - Choose local or remote
output_storage,cache.storage,table_provider, andvector_storeproviders before indexing. - Validate without network calls:
python scripts/inspect_config.py --root <project> python scripts/smoke_input_readers.py - If config and readers pass, continue with indexing in
../indexing/; do not run indexing just to validate settings.
Key APIs and Commands
from graphrag.config.load_config import load_configloadssettings.[yaml|yml|json]from a root and returnsGraphRagConfig.load_config(root_dir, cli_overrides={...})recursively overlays CLI-style nested dictionaries before Pydantic validation.GraphRagConfigvalidates file base directories, resolves file paths, expands LanceDBdb_uri, and overlays default vector schemas for GraphRAG embeddings.graphrag init --root <project>writes default config content with${GRAPHRAG_API_KEY}interpolation and common prompt file paths.graphrag index --root <project>belongs to../indexing/after this sub-skill has checked prerequisites.
Provider Matrix
- Inputs:
text,csv,json,jsonl,markitdown, andparquetthroughgraphrag_input.create_input_reader. - Storage:
file,memory,blob, andcosmosdbviaStorageConfig, depending on the section using it. - Cache:
json,memory, andnone; JSON cache uses a nestedstorageblock. - Tables:
parquet,csv, andcosmosdbthroughtable_provider.type. - Vectors:
lancedb,azure_ai_search, andcosmosdbthroughvector_store.type.
References
- See
references/config-schema-and-env.mdforGraphRagConfig, model auth,.env, and validation behavior. - See
references/input-storage-cache-vector.mdfor reader schemas and storage/cache/table/vector provider settings. - See
references/troubleshooting.mdfor common validation failures and doc/code drift traps.
Bundled Offline Checks
scripts/inspect_config.pyloads a project config and prints a JSON summary with secret-like values redacted.scripts/smoke_input_readers.pycreates tiny temporarytext,csv,json,jsonl, and optionalparquetinputs and reports local reader readiness without credentials or network access.
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).