NitinSukthe-G

Ragagentsv1v2 — Testing skill for Claude Code

Testing community

RAG chat agent for your own documents, built in two stages.

How to install Ragagentsv1v2

This entry records only its repository, not the path inside it, so there is no exact command to give. Open NitinSukthe-G/ragagentsv1v2 and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Ragagentsv1v2 does

RAG chat agent for your own documents, built in two stages. V1 answers with local models only (hybrid search, reranking, extractive QA, table/date/number logic), no LLM. V2 adds a tool-calling LLM agent (Sarvam/Gemini/OpenAI/Claude) with streamed, cited answers. FastAPI · MongoDB · Chroma/Pinecone

Alternatives in Testing

Fix Issue — by metabase - Addresses GitHub issues by taking issue number as parameter, analyzing context, implementing sol 46.5k ★
Create Test Repo — Create a new ArcKit test repo with full scaffolding, auto-detecting the next version number 2.2k ★
Mda — MDA Open Spec — a Markdown superset for agent-facing documents 615 ★

README

RAG Chat Agent — V1 → V2

A document question-answering chat agent, built in two stages. Both versions use the same ingestion and retrieval pipeline: FastAPI, MongoDB, Chroma and `all-MiniLM-L6-v2` embeddings. They differ in how the answer is produced.

	V1 — no LLM	V2 — LLM agent
Answering	Extractive QA plus plain-Python reasoners (numeric, date, table, comparison)	A tool-calling LLM agent that writes cited answers
Retrieval	Hybrid (Chroma + BM25), cross-encoder reranking, 16 rule-based intents	Semantic search, which the agent calls as a tool
Summaries and follow-ups	❌	✅
Streaming	❌	✅ SSE
Pluggable parts	—	LLM (Sarvam / Gemini / OpenAI / Claude), vector DB (Chroma / Pinecone)
Extras	Confidence score, evidence snippets	Redis cache and rate limit, `/stats` cost and latency tracking

V1 was built without an LLM on purpose. Retrieval had to be correct first, because a wrong passage produces a visibly wrong answer instead of a fluent made-up one. V2 then adds an LLM on top of that same retrieval.

Quick start

cd v1        # or: cd v2
python -m venv .venv && source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env                                 # fill in MONGODB_URI (+ an LLM key for v2)
python -m uvicorn app.main:app --reload              # http://localhost:8000

Setup, architecture, API and design notes for each version are in its own README: