langchain-deepinquiry

LangChain integration for DeepInquiry: a library of short factual claims, each returned with its sources, a verification tier (gold, silver, bronze), a source-difference score and a signed CiteSig attestation.

Install

pip install langchain-deepinquiry

Get a free API key (500 requests a month, no card) at https://www.deepinquiry.ai/signup and set it:

export DEEPINQUIRY_API_KEY="di_live_..."

Retriever

from langchain_deepinquiry import DeepInquiryRetriever

retriever = DeepInquiryRetriever(k=4)
docs = retriever.invoke("how many hearts does an octopus have")
for d in docs:
    print(d.page_content)
    print(d.metadata["verification_tier"], d.metadata["source"])

Each Document carries the fact id, the page to cite (source), the verification tier, the score, the source URLs and the signed attestation (citesig).

Agent tools

from langchain_deepinquiry import DeepInquirySearchTool, DeepInquiryGetFactTool, DeepInquiryRandomFactTool

tools = [DeepInquirySearchTool(), DeepInquiryGetFactTool(), DeepInquiryRandomFactTool()]

Pass them to any LangChain or LangGraph agent. Results come back as plain text with the sources and a link to cite.

Try the API first

There is a live, keyless demo on https://www.deepinquiry.ai/api that shows the exact response.

Terms

The code in this package is MIT licensed. Facts returned by the API have their own terms: the free tier requires attribution to DeepInquiry (the links in each result serve for that) and paid plans allow commercial use. See https://www.deepinquiry.ai/terms and https://www.deepinquiry.ai/data-license.

Published by Machine Accountability Corporation.