eagle-typesense — RETIRED

This service is retired and this repository is archived. Nothing below is running.

Search for EPIC is served by eagle-search, which puts the same data in Azure AI Search and adds document content search — searching the text inside PDFs, not just their titles.

Retired on 2026-08-10, after both remaining deployments had served zero searches for 24 hours:

• Helm releases uninstalled from 6cdc9e-dev and 6cdc9e-test.
• eao-nginx no longer proxies /api/public/typesense or /api/typesense, and TYPESENSE_ENABLED is out of the runtime config.
• eagle-api no longer carries the proxy controller, the client, the scoped-search-key endpoint or the typesense dependency. Its /api/search never used Typesense — it has always run the Mongo aggregators — so removing all of it changed no behaviour.

The rest of this README describes how the service worked, and is kept for that reason alone.

---

Typesense search engine and MongoDB sync service for the EPIC platform.

Architecture

eagle-public/admin ──HTTP──▶ eagle-api (typesenseProxy.js)
                                  │ injects allowed_roles filter
                                  ▼
                             Typesense server ◀── typesense-sync (this repo)
                                                        │
                                                  MongoDB (epic collection)
                                                  MinIO (PDF extraction)
                                                  PostgreSQL (popularity scores)

Components

Component	Description
Typesense server	Full-text search engine (deployed via Helm)
Change Stream listener	Real-time sync from MongoDB → Typesense (src/index.js)
Full sync CronJob	Nightly zero-downtime re-index (src/full-sync.js)
Popularity sync	Nightly score update from penguin-analytics (src/popularity-sync.js)

Collections

Collection	Source Schema	Purpose
documents	Document	Project documents (PDFs, reports)
projects	Project	EA projects
activities	RecentActivity	News/updates feed
notifications	ProjectNotification	Project notifications
document_chunks	DocumentChunk	Full-text search within PDFs

Quick Start

# Install dependencies
npm install

# Copy environment config
cp .env.example .env

# Start Change Stream listener (real-time sync)
npm start

# Run full re-index
npm run full-sync

# Run popularity score update
npm run popularity-sync

Environment Variables

See .env.example for all configuration options.

Required

Variable	Description
TYPESENSE_HOST	Typesense server hostname
TYPESENSE_PORT	Typesense server port (default: 8108)
TYPESENSE_API_KEY	Admin API key
MONGODB_HOST	MongoDB hostname
MONGODB_DATABASE	Database name (default: epic)

Optional

Variable	Description
MONGODB_USERNAME	MongoDB auth username
MONGODB_PASSWORD	MongoDB auth password
MONGODB_DIRECT	Use directConnection (for port-forwarding)
MINIO_HOST	Object store for PDF extraction
PENGUIN_DB_HOST	PostgreSQL for popularity scores

Deployment

Deployed to OpenShift via Helm chart in helm/.

# Deploy to dev
helm upgrade --install typesense ./helm \
  --namespace 6cdc9e-dev \
  --values ./helm/values-dev.yaml

# Deploy to test/prod
helm upgrade --install typesense ./helm \
  --namespace 6cdc9e-{test,prod} \
  --values ./helm/values-{test,prod}.yaml

Access Control

All documents get an allowed_roles field populated from MongoDB's read array. Typesense scoped search keys embed a filter_by: "allowed_roles:=[<roles>]" constraint so clients physically cannot bypass access control. Fail-closed: documents without a read array default to ['sysadmin'] visibility only.