DEDUPED Hot AI repos. Real use cases. What people build. Get the Monday 5

How to answer multi-step questions with agentic retrieval with Ragflow

Turn on Agentic Retrieval so the model breaks down a hard question, retrieves several times and checks evidence.

Short answer: Deploy RAGFlow and load the documents into a dataset; then enable Agentic Retrieval for complex questions; pick a thinking mode: Low, Medium, High or Ultra, depending on how complex the question is. All steps.

  • Research
  • Intermediate
  • 30 minutes
  • From the README
  • Any agent

cost. Higher thinking modes make more model calls per question, which costs more with a paid LLM provider.

Who it is forAn analyst asking questions that span several documents, such as comparing terms across contracts.

The problemSingle-shot retrieval returns a few chunks and misses context spread across many files.

The outcomeBetter grounded answers built from multiple rounds of retrieval, at a thinking depth you choose.

What you need first

  • Docker 24.0.0 or later and Docker Compose v2.26.1 or later
  • A starting machine of 4 CPU cores, 16 GB RAM and 50 GB free disk
  • An LLM, an embedding model and a reranker, added with their service address and API key after sign-in
  • If you use Elasticsearch, vm.max_map_count set to at least 262144

Steps

  1. Deploy RAGFlow and load the documents into a dataset
  2. Enable Agentic Retrieval for complex questions
  3. Pick a thinking mode: Low, Medium, High or Ultra, depending on how complex the question is
  4. Optionally use Knowledge Compilation to generate a Wiki, Graph or Timeline of the dataset first
  5. Check the citations in the answer before relying on it

Prompt for your agent

In infiniflow/ragflow, enable Agentic Retrieval on dataset <name> and ask <question> at Medium and then High thinking mode. Compare the two answers and their citations and tell me which is better supported.

Caveat. RAGFlow 1.0 is at release candidate stage, so expect changes between versions.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for infiniflow/ragflow and "Answer multi-step questions with agentic retrieval"

Add Deduped to Claude Code once

claude mcp add --transport http deduped https://dedupedai.com/mcp

Cursor, ChatGPT, others

Share: X LinkedIn Hacker News Email Copy link

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer