Concept 07

Retrieval (RAG)

Retrieval augmented generation means the assistant looks things up in your documents before answering, and answers from what it found rather than from memory. Your interviews, policies, past specs and support tickets become something it can search. The quality of the answer is the quality of the search and the documents.

Standing: Established since 2023-11
Stars, week: +1,181 across 6 repositories
Stars, month: +8,083, through 2026-11-08
Listings: 6 chosen by hand

The standing

Where it stands

Among the 14 concepts. Bars: GitHub stars gained this week.

Established · 7

Rising · 7

How to use it today

In ChatGPT or Claude, as a person doing the work, not as a developer.

1Put the documents that matter in one place the assistant can reach: a Project in ChatGPT or Claude, a connected drive, or a folder in Claude Code.
2Ask questions that cite: "from the interview notes, what did customers say about pricing, with quotes"; insist on quotes or references in the answer.
3Keep sources clean and dated; delete superseded versions, since retrieval does not know which policy is current.
4When the answer is wrong, check what it retrieved before blaming the model: usually the right passage was never found.
5For a team, use a connected knowledge base (Notion, Confluence, a drive) through a connector rather than uploading copies.

What changes day to day

You stop summarizing documents for the assistant and start curating the document set. Answers come with references, and "where did that come from" becomes a question you can answer.

Where it goes wrong

Retrieval returns the most similar text, not the most relevant or the most recent.

Stale documents produce confident, outdated answers with real citations.

Listings that use it well

Chosen by hand after reading them, each with the reason. Install from the listing's own page.

Listing	Kind	Why it is here	Stars
ai-agent-book	Agent skills	rag-pipeline covers chunking, dense and sparse embeddings, hybrid retrieval with RRF and rerankers, recall and nDCG.	57k
llm-application-dev	Agent skills	rag-implementation, embedding-strategies, hybrid-search-implementation and vector-index-tuning skills.	42k
claude-skills	Agent skills	rag-architect designs chunking, embeddings, vector stores and hybrid pipelines.	12k
AI-Research-SKILLs	Agent skills	Skills for LlamaIndex, LangChain, Chroma and FAISS.	14k
agentset	Agent skills	An open source RAG platform with built in citations and an MCP server.	2.1k
mcp-server-qdrant	Agent skills	Qdrant's official MCP server, a vector store an assistant can write to and search.	1.6k

Elsewhere

Repositories that embody the idea and are not in the catalog.

run-llama/llama_index ↗Data framework for indexing and querying documents for LLMs.
infiniflow/ragflow ↗RAG engine built around document parsing and chunking quality.
onyx-dot-app/onyx ↗Self-hosted enterprise search and RAG assistant (formerly Danswer) with 50+ connectors.

Standing

Our judgment, with the reason and the sources it rests on.

Established since 2023-11. A settled technique since 2023; the naive version (chunk, embed, stuff) is giving way to agentic search and long context, but retrieval itself is in every product. Vendor-built retrieval has shipped since OpenAI's Assistants API in November 2023 (retired in August 2026; its file search moved to the Responses API); Anthropic's Contextual Retrieval (September 2024) refined chunk-based RAG, and its 2025 context engineering guidance recommends mixing a little retrieval up front with search by the agent as it works. Settled practice: the methods are known, the tools are mature, and the question is how well you use it.

New models and developer products announced at DevDay2023-11-06
Introducing Contextual Retrieval2024-09-19
Effective context engineering for AI agents2025-09-29