Live Demo

Verified Document Search

documents indexed · chunks · collections · ZK proof verified by zkVerify

An AI search system that can prove — cryptographically — that its results come from your actual documents, not fabricated citations. Every answer generates a ZK proof verified by zkVerify, anchored on the Horizen blockchain — provably, without exposing the source documents. The collection below is a live demo using publicly available military doctrine as the example corpus.

Corpus Setup (one-time indexing)
📄
Your PDFs
Pipeline
(extract, chunk, embed)
🔓
Merkle Tree
+ ZK Proof
🖸
Horizen
Blockchain
Query (at search time)
🔍
Search
API
🔓
Generate
ZK Proof
🚈
Kurier
API
zkVerify
Verified
Result

Documents are anchored on-chain once. Every search triggers a fresh ZK proof verified via Kurier → zkVerify — provably, without revealing the documents.

📂 Browse the Example Document Catalog

See what documents are in the demo corpus — titles, sources, and descriptions.

📡 API Reference

Build with the API — agents use the REST interface, humans browse the interactive docs below.

🔎 Search the Demo Corpus

📁 Search Within Collection

Collections

Collection Description Documents
Loading…

API Example: Query with Metadata Filtering

curl -X POST https://militarymanuals.ai/api/query \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "collection": "army",
    "query": "How do I conduct urban operations?",
    "metadata_filter": {
      "category": "field-manual",
      "classification": "unclassified"
    }
  }'

Discovery

For Content Owners

Publish your documents once. Others can search and verify them — provably — without ever downloading the originals. Your IP stays private; your corpus stays useful.

For Agents & Developers

Build AI systems that cite their sources with cryptographic certainty. Every answer can come with a ZK proof that it was drawn from your committed corpus, not fabricated.

Run this for your own corpus

The full open-source implementation is available on GitHub — pipelines, ZK circuits, smart contracts, and frontend all included. Built on Horizen EVM and zkVerify.

🔗 How It Works — ZK-RAG

ZK-RAG (Zero-Knowledge Retrieval-Augmented Generation) combines semantic search with cryptographic proofs so every result can be independently verified without revealing the underlying documents. Every passage in these search results is anchored to the Horizen blockchain at the time documents are registered. The ZK circuit proves the following without revealing the full document:

1
Poseidon Hash The passage text is hashed using the Poseidon function into a leaf hash.
2
Merkle Proof The circuit proves this leaf is a member of the Merkle tree for this document.
3
Chain Anchor The Merkle root was anchored on Horizen mainnet when the document was registered — verified on-chain.
4
ZK Verification Anyone can verify this proof on Horizen via zkVerify — without downloading the document or trusting a third party.

This allows third parties to cryptographically verify that a passage existed in a document at a specific block — provably — without requiring access to the original file. Click "🔗 Not verified" on any passage to generate and submit a ZK proof.