pipelines
Routes UltraRAG pipeline build, run, and recipe workflows, including YAML orchestration and the Python API wrappers.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- pipelines
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Routes UltraRAG pipeline build, run, and recipe workflows, including YAML orchestration and the Python API wrappers.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/pipeline-dsl.md2.8 KB
- 参考文档references/troubleshooting.md2.5 KB
- 参考文档references/workflows.md3.4 KB
- 脚本scripts/smoke_sayhello_pipeline.py4.0 KB
03 SKILL.md 正文
Pipelines
Use this sub-skill when the task is about running, building, or authoring an UltraRAG pipeline.
Typical triggers
ultrarag build,ultrarag run, orultrarag showToolCall/PipelineCallusage from Pythonpipeline.yaml,servers:,pipeline:,loop:,branch:- Demo or experiment files under
examples/demos/orexamples/experiments/ - Questions about how a pipeline step consumes or produces variables
- Requests to reproduce or adapt a named example such as
RAG,LLM,LightResearch,AgentCPM-Report,webnote,search_o1,search_r1,rankcot,ircot,visrag, orsayhello
What this sub-skill covers
- CLI build/run/show entry points.
ToolCallandPipelineCallfrom the Python API.- Pipeline structure, variable resolution, step remapping, branch/loop logic,
and generated
parameter/andserver/artifacts. - Example workflows that combine benchmark data, retriever, generation, prompt, custom, memory, router, evaluation, and corpus steps.
- Safe smoke checks that prove the orchestration layer still works.
What stays elsewhere
- Individual MCP server signatures and backend-specific configuration belong in
sub-skills/servers/. - Flask UI, chat sessions, knowledge-base storage, and case study belong in
sub-skills/ui-and-storage/.
Start here
- Read
references/pipeline-dsl.mdfor the verified YAML syntax, data-flow rules, and build/run behavior. - Read
references/workflows.mdfor a map of example families and the best starting point for each one. - Read
references/troubleshooting.mdwhen a pipeline build, parameter merge, or execution step fails. - Run
scripts/smoke_sayhello_pipeline.pyfor a minimal end-to-end smoke check.
Common user questions this sub-skill should answer
- How do I build a pipeline from YAML?
- How do I run a pipeline with a custom parameter file?
- How do I express a loop or branch?
- Why did a step remap or variable lookup fail?
- Which example pipeline is closest to my task?
- How do I get a small smoke test working before a larger RAG workflow?
Practical workflow
- Identify the example family that matches the request.
- Check whether the task is only orchestration, or whether it also needs a specific server backend or UI route.
- Use the DSL reference to confirm the step shape and variable names.
- Use the smoke script or a minimal example to validate the plan.
Helpful commands
Use the bundled smoke helper instead of relying on a source example file:
python sub-skills/pipelines/scripts/smoke_sayhello_pipeline.py --repo-root <checkout>
ultrarag show ui
ultrarag show case --config_path <memory-case.json>
If the task is about a named experiment or demo, use the distilled family map in
references/workflows.md before guessing the pipeline structure.
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).