Quickstart

This quickstart uses only surfaces found in the repository. It assumes the API, Qdrant and Gel/EdgeDB are running locally.

1. Start data services

cd repos/VectorHub-dev
docker compose -f backend1/Client/docker-compose.yml up edgedb qdrant

2. Start the API

cd repos/VectorHub-dev/backend1/Client
python server_endpoints.py

3. Verify the API

curl http://localhost:8081

4. Query semantic context

POST /query is confirmed in server_endpoints.py and documented in VectorHub-client/README.md.

curl -X POST "http://localhost:8081/query" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -d '{
    "query": "teste",
    "collection_name": "teste",
    "with_payload": true,
    "score_threshold": 0.5,
    "limit_wo_XID": 5,
    "limit_w_XID": 2,
    "limit_w_hits": 15,
    "query_filter": {
      "must": [
        { "key": "system_metadata.tags_name", "match": { "value": "contract" } }
      ]
    }
  }'

The exact results depend on an existing collection and indexed content.

Known limits

• Collection creation exists through /collections, but the route uses RPC dispatch with { "fun": "...", "data": ... }.
• Auth setup is real, but public onboarding flows depend on configured secrets and users.
• The portal docs do not ship sample credentials.