Semantic search
Query in natural language; results rank by relevance. Optional hybrid mode combines dense vectors with sparse BM42 + RRF fusion for sharper keyword behaviour.
Semantic storage
A per-user vault for keys, descriptions, and values — with semantic search powered by embeddings, optional hybrid dense + sparse retrieval, and a clean HTTP API (JWT or API keys).
Store structured records with a stable key, human-readable description, and arbitrary value text. Search uses the text you care about — not only exact string matches.
Query in natural language; results rank by relevance. Optional hybrid mode combines dense vectors with sparse BM42 + RRF fusion for sharper keyword behaviour.
Every vector and payload is scoped by user_id. JWT sessions and hashed API keys; optional AES-GCM encryption for sensitive payload fields at rest.
Use X-API-Key for scripts: same vault operations as the browser token. Export and import JSON for backups.
Tier-based record caps (Free / Pro / Team). Per-IP rate limits with Redis for multi-instance deployments; standard limit headers on responses.
Defaults shown below are enforced server-side via environment variables. Billing integration is not included — tiers are a roadmap hook for Stripe or similar.
Individuals getting started
Power users & small teams
Larger vaults
Base URL is your deployment origin. Prefer versioned paths under /v1 (mirrors unversioned routes). Interactive schema: /docs.
curl -s -X POST "$ORIGIN/v1/search" \
-H "Authorization: Bearer $JWT" \
-H "Content-Type: application/json" \
-d '{"query":"wifi password","limit":10,"offset":0}'
curl -s -X POST "$ORIGIN/v1/records" \
-H "X-API-Key: sv_..." \
-H "Content-Type: application/json" \
-d '{"key":"wifi","description":"home","value":"secret","tags":[]}'