ai-rag

Featured

Designs retrieval-augmented generation and search systems. Use when choosing retrieval, chunking, hybrid search, grounding, or RAG evaluation patterns.

AI & Automation 80 stars 17 forks Updated 1 weeks ago MIT

Install

View on GitHub

Quality Score: 89/100

Stars 20%
64
Recency 20%
90
Frontmatter 20%
70
Documentation 15%
100
Issue Health 10%
80
License 10%
100
Description 5%
100

Skill Content

# RAG & Retrieval Engineering Build retrieval systems that are grounded, observable, and explicit about tradeoffs. This skill covers: - Retrieval architecture choice: long-context vs hosted file search vs tool-first/MCP vs SQL/graph vs classic vector RAG - Corpus preparation: parsing, metadata, chunking, ACLs, freshness, invalidation - Retrieval quality: sparse, dense, hybrid, late interaction, reranking, multimodal retrieval - Answer quality: grounding, citation coverage, refusal on missing evidence, regression testing **July 2026 posture** - Choose the retrieval mode before tuning chunk size. A vector index is not the default answer to every knowledge problem. - Separate retrieval quality from answer quality and evaluate both. - Treat retrieved text, tool responses, and MCP resources as untrusted input. - Prefer primary sources for vendor or framework recommendations; volatile facts must be verified live. - Treat OpenTelemetry GenAI semantic conventions as useful but still evolving. - **Context-budget note (Opus 4.7 tokenizer):** The Claude Opus 4.7 tokenizer encodes ~1.0–1.35× more tokens than the pre-2026 tokenizer for the same text. Chunk-size and token-budget heuristics from earlier than 2026 are invalid — re-measure on your own corpus with the current tokenizer before setting chunk sizes or context-window budgets. - **Managed retrieval is a real option:** Anthropic's `web_search_20260209` (and `web_search_20250305`) server tools and OpenAI's file-search are API-na...

Details

Author
vasilyu1983
Repository
vasilyu1983/AI-Agents-public
Created
9 months ago
Last Updated
1 weeks ago
Language
Python
License
MIT

Integrates with

Similar Skills

Semantically similar based on skill content — not just same category