AGENT SKILL

document-pipeline

Work with LightRAG document ingestion internals: parser routing, filename and process options, chunker selection, sidecars, multimodal analysis flow, doc-status metadata, and pipeline concurrency/refusal semantics.

License 许可
MIT
最近更新
2026-09-03
Stars
★ 351
Forks
27

01 前言 Front Matter

name
document-pipeline
license
MIT
metadata
{ "disco-role": "operating" }
description
Work with LightRAG document ingestion internals: parser routing, filename and process options, chunker selection, sidecars, multimodal analysis flow, doc-status metadata, and pipeline concurrency/refusal semantics.
disable-model-invocation
true

02 资源文件 5

03 SKILL.md 正文

LightRAG Document Pipeline

Use this sub-skill when a task involves file/text ingestion behavior, parser routing, process-option strings, chunker strategy selection, sidecar artifacts, multimodal analysis flow, doc-status state metadata, or upload/scan/delete concurrency reasoning.

Route By Task

  • Choose parser engines, filename hints, LIGHTRAG_PARSER rules, process options, engine parameters, or chunk parameters with parser-and-process-options.
  • Compare F / R / V / P chunking, understand sidecar artifacts, or reason about fallback/provenance behavior with chunking-and-sidecars.
  • Explain why enqueue, scan, delete, or processing was refused or deferred using pipeline-concurrency.
  • Diagnose unavailable parser services, invalid hints/options, missing sidecars, doc-status failures, cancellation, or stale parse/chunk metadata with troubleshooting.
  • Run scripts/check_pipeline_symbols.py for a safe installed-package import and symbol/signature check that does not call parser services, models, credentials, storages, or repository tests.

Boundaries

  • This sub-skill owns apipeline_enqueue_documents, apipeline_process_enqueue_documents, parse/analyze/process staging, parser routing functions, process_options, chunk_options, sidecar contracts, multimodal sidecar chunks, doc-status metadata carry-over, and pipeline_status fields.
  • For direct LightRAG construction, ainsert, aquery, lifecycle, custom KG insertion, embeddings, cache, rerank, and generic library examples, route to ../core-rag/SKILL.md.
  • For API endpoint request/response details, upload route models, WebUI operation, auth, server startup, or deployment, route to ../api-server/SKILL.md.
  • For storage backend selection, database services, migrations, workspace isolation, or destructive storage operations, route to ../storage-backends/SKILL.md.
  • For LLM, VLM, embedding, reranker provider setup, credentials, and provider-specific parameters, route to ../llm-providers/SKILL.md.

Non-Negotiables

  • Treat process_options as the exact per-document selector string: i, t, e, !, plus at most one chunk selector from F / R / V / P.
  • Do not make reusable guidance depend on source-checkout paths, repository scripts, examples, docs, tests, or local environment locations.
  • Be explicit when a parser or modality path needs optional dependencies, an external parser service, embeddings, or VLM/LLM roles; do not present those paths as always-available.
  • Preserve the concurrency contract: busy alone does not block enqueue, but destructive_busy and scanning_exclusive do.

04 同仓库更多 Skills

VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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).

★ 351