HistoryLab Vector Search API

A Cloudflare Worker providing semantic search across ~5 million declassified historical documents.

Quick Start

Base URL: https://vector-search-worker.nchimicles.workers.dev

Search for Documents

curl -X POST "https://vector-search-worker.nchimicles.workers.dev/api/search" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "queries": "Cuban Missile Crisis",
    "topK": 5
  }'

Fetch Full Document

curl -X GET "https://vector-search-worker.nchimicles.workers.dev/api/document/{r2Key}" \
  -H "Authorization: Bearer YOUR_API_KEY"

Documentation

See API_DOCUMENTATION.md for complete documentation including:

• Authentication
• All endpoints and parameters
• Filtering (date ranges, document IDs)
• Response schemas
• Code examples (curl, Python)
• Query tips

API Endpoints

Endpoint	Method	Description
/api/search	POST	Semantic search
/api/document/{r2Key}	GET	Fetch full document
/api/health	GET	Health check
/api/help	GET	API docs (JSON)

Key Features

• Semantic Search - Natural language queries using BGE embeddings
• Date Filtering - Filter by year/month or exact date (YYYYMM or YYYYMMDD)
• Full Document Retrieval - Get complete document text and metadata
• Multiple Queries - Search multiple concepts in one request

Document Sources

~5 million documents from:

• CIA declassified documents
• State Department Central Foreign Policy Files
• Foreign Relations of the United States (FRUS)
• Presidential Daily Briefings
• NATO, UN, World Bank archives
• And more...

Development

Prerequisites

• Node.js 18+
• Wrangler CLI

Deploy

wrangler deploy

Set API Key

wrangler secret put API_SECRET_KEY_2

Technical Stack

• Runtime: Cloudflare Workers
• Embeddings: @cf/baai/bge-base-en-v1.5 (768 dimensions)
• Vector DB: Cloudflare Vectorize
• Storage: Cloudflare R2 + KV