AGENT SKILL

pipelines-and-methods

Plan and implement FlashRAG end-to-end RAG pipelines, quick-start demos, method reproductions, active/reasoning/branching/multimodal pipelines, multi-turn prompts, and experiment-runner dry-runs without executing large models by default.

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

01 前言 Front Matter

name
pipelines-and-methods
license
MIT
metadata
{ "disco-role": "operating" }
description
Plan and implement FlashRAG end-to-end RAG pipelines, quick-start demos, method reproductions, active/reasoning/branching/multimodal pipelines, multi-turn prompts, and experiment-runner dry-runs without executing large models by default.
disable-model-invocation
true

02 资源文件 6

03 SKILL.md 正文

FlashRAG Pipelines and Methods

Use this sub-skill when the task is about choosing or wiring FlashRAG pipelines, adapting quick-start demos, reproducing supported methods, planning multi-turn or multimodal workflows, or dry-checking an experiment configuration before any large model/index execution.

Route first

  • For standard text RAG, start with SequentialPipeline or naive_run; see pipeline API and quick-start/multi-turn.
  • For method reproduction, map the requested method name to the experiment-runner recipe and prerequisites; see method recipes.
  • For active, branching, adaptive, or reasoning RAG, select the pipeline by required control flow and model support; see pipeline API and multimodal/reasoning.
  • For multimodal RAG, check dataset modality, prompt template, multimodal retriever settings, generator capability, and hardware before running; see multimodal/reasoning.
  • Before execution, run the bundled safe checker: python skills/flashrag/sub-skills/pipelines-and-methods/scripts/validate_pipeline_config.py --config CONFIG.yaml --method METHOD --dry-run-plan.
  • If a run fails or looks wrong, triage with troubleshooting.

Safety defaults

  • Do not execute model downloads, index building, vLLM server/model loading, or full benchmark runs unless explicitly requested.
  • Prefer test_sample_num, small retrieval_topk, short generation_params, and do_eval=False for smoke tests.
  • Treat model paths, corpus paths, index paths, cache paths, and output paths as user-provided runtime values; do not hard-code machine-specific paths in public instructions.
  • Keep low-level config schema details, index construction, generator/refiner internals, and metric/UI work in their dedicated FlashRAG skill areas.

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