mirror of
https://github.com/magnus919/agent-skills.git
synced 2026-09-14 13:06:30 +03:00
40 lines
1.5 KiB
Markdown
40 lines
1.5 KiB
Markdown
# Haystack — Production Search & NLP Pipelines (deepset)
|
|
|
|
An expert-level skill for building **production search and NLP pipelines** with Haystack. Pipelines are validated DAGs with typed components and explicit connections.
|
|
|
|
## Why Install This Skill
|
|
|
|
When your agent loads this skill, it becomes a Haystack expert who can:
|
|
|
|
- **Design pipeline DAGs** — add_component + connect with typed input/output slots
|
|
- **Build RAG pipelines** — document indexing + query pipelines with embedding retrieval
|
|
- **Create agentic systems** — tool-using agents with ReAct pattern
|
|
- **Integrate generative AI** — PromptBuilder (Jinja2) + LLM generators
|
|
- **Evaluate pipeline quality** — faithfulness, relevancy, and custom metrics
|
|
- **Deploy with Hayhooks** — REST API deployment for production
|
|
|
|
## What You Get
|
|
|
|
| Directory | Purpose |
|
|
|-----------|---------|
|
|
| `SKILL.md` | Core paradigm, where-to-start table, framework comparison |
|
|
| `references/` | Deep dives into pipeline design, RAG, agents, evaluation, Hayhooks, and framework comparisons |
|
|
|
|
## Framework Comparison
|
|
|
|
Haystack uses explicit Pipeline DAGs (add_component + connect) — different from LangChain's LCEL pipe operator and LlamaIndex's query engines. Pipelines are validated at declaration time.
|
|
|
|
## Requirements
|
|
|
|
Python 3.8+ with `haystack-ai` package.
|
|
|
|
|
|
## Quick Start
|
|
|
|
Start with the setup and first workflow in SKILL.md, then use the linked resources for the specific task you need to complete.
|
|
|
|
|
|
## Triggers
|
|
|
|
Use this skill for the task types and keywords described in its SKILL.md description.
|