Qdrant Ops

REST API Patterns

All operations use curl. Pipe through python3 -m json.tool for readable output.

List all collections

curl -s -H "api-key: ${KEY}" "${BASE}/collections" | python3 -m json.tool

Get collection info (schema, counts, config)

curl -s -H "api-key: ${KEY}" "${BASE}/collections/<name>" | python3 -m json.tool

Create a collection

curl -s -X PUT -H "api-key: ${KEY}" -H "Content-Type: application/json" \
  "${BASE}/collections/<name>" \
  -d '{
    "vectors": {
      "size": 1536,
      "distance": "Cosine"
    }
  }' | python3 -m json.tool

For named/hybrid vectors (dense + sparse):

-d '{
  "vectors": {
    "dense": {"size": 1536, "distance": "Cosine"},
    "sparse": {}
  }
}'

Delete a collection

curl -s -X DELETE -H "api-key: ${KEY}" "${BASE}/collections/<name>" | python3 -m json.tool

Points — Browse & Inspect

Scroll (paginate all points)

curl -s -X POST -H "api-key: ${KEY}" -H "Content-Type: application/json" \
  "${BASE}/collections/<name>/points/scroll" \
  -d '{
    "limit": 10,
    "with_payload": true,
    "with_vector": false
  }' | python3 -m json.tool

Use "offset": <next_page_offset> from the response to paginate.

Get points by ID

curl -s -X POST -H "api-key: ${KEY}" -H "Content-Type: application/json" \
  "${BASE}/collections/<name>/points" \
  -d '{"ids": ["<uuid-or-int>"], "with_payload": true, "with_vector": false}' \
  | python3 -m json.tool

Count points (with optional filter)

curl -s -X POST -H "api-key: ${KEY}" -H "Content-Type: application/json" \
  "${BASE}/collections/<name>/points/count" \
  -d '{"exact": true}' | python3 -m json.tool

Points — Upsert & Delete

Upsert points

curl -s -X PUT -H "api-key: ${KEY}" -H "Content-Type: application/json" \
  "${BASE}/collections/<name>/points" \
  -d '{
    "points": [
      {
        "id": "<uuid-or-int>",
        "vector": [0.1, 0.2, ...],
        "payload": {"text": "...", "source": "..."}
      }
    ]
  }' | python3 -m json.tool

Delete points by ID

curl -s -X POST -H "api-key: ${KEY}" -H "Content-Type: application/json" \
  "${BASE}/collections/<name>/points/delete" \
  -d '{"points": ["<id1>", "<id2>"]}' | python3 -m json.tool

Delete points by filter

curl -s -X POST -H "api-key: ${KEY}" -H "Content-Type: application/json" \
  "${BASE}/collections/<name>/points/delete" \
  -d '{
    "filter": {
      "must": [{"key": "document_name", "match": {"value": "tshelpdesk"}}]
    }
  }' | python3 -m json.tool

Update payload on existing points

curl -s -X POST -H "api-key: ${KEY}" -H "Content-Type: application/json" \
  "${BASE}/collections/<name>/points/payload" \
  -d '{
    "payload": {"new_field": "value"},
    "points": ["<id1>"]
  }' | python3 -m json.tool

Search

Vector similarity search

curl -s -X POST -H "api-key: ${KEY}" -H "Content-Type: application/json" \
  "${BASE}/collections/<name>/points/search" \
  -d '{
    "vector": [0.1, 0.2, ...],
    "limit": 5,
    "with_payload": true,
    "with_vector": false
  }' | python3 -m json.tool

Filtered search

-d '{
  "vector": [...],
  "limit": 5,
  "filter": {
    "must": [{"key": "metadata.source_type", "match": {"value": "pdf"}}]
  },
  "with_payload": true
}'

Search with named vector (hybrid collections like tro-child-hybrid-v1)

-d '{
  "vector": {"name": "dense", "vector": [...]},
  "limit": 5,
  "with_payload": true
}'

Recommend (find similar to existing point)

curl -s -X POST -H "api-key: ${KEY}" -H "Content-Type: application/json" \
  "${BASE}/collections/<name>/points/recommend" \
  -d '{
    "positive": ["<point-id>"],
    "limit": 5,
    "with_payload": true
  }' | python3 -m json.tool

Filtering Reference

Filters go inside "filter": { ... } blocks. Conditions combine with must (AND), should (OR), must_not (NOT).

{
  "must": [
    {"key": "metadata.source_type", "match": {"value": "pdf"}},
    {"key": "chunk_index", "range": {"gte": 0, "lte": 100}}
  ],
  "should": [
    {"key": "document_name", "match": {"value": "tshelpdesk"}},
    {"key": "document_name", "match": {"value": "tsjobaid"}}
  ]
}

Indexes & Performance

Create payload index (for fast filtered search)

curl -s -X PUT -H "api-key: ${KEY}" -H "Content-Type: application/json" \
  "${BASE}/collections/<name>/index" \
  -d '{"field_name": "metadata.source_type", "field_schema": "keyword"}' \
  | python3 -m json.tool

Field schema types: keyword, integer, float, bool, text, geo, datetime

Collection aliases

# Create alias
curl -s -X POST -H "api-key: ${KEY}" -H "Content-Type: application/json" \
  "${BASE}/aliases" \
  -d '{"actions": [{"create_alias": {"alias_name": "prod", "collection_name": "<name>"}}]}' \
  | python3 -m json.tool

# List aliases
curl -s -H "api-key: ${KEY}" "${BASE}/aliases" | python3 -m json.tool

Cluster & Health

# Health check
curl -s "${BASE}/healthz"

# Cluster info
curl -s -H "api-key: ${KEY}" "${BASE}/cluster" | python3 -m json.tool

# Collection telemetry
curl -s -H "api-key: ${KEY}" "${BASE}/telemetry" | python3 -m json.tool

Snapshots (Backup/Restore)

# Create snapshot
curl -s -X POST -H "api-key: ${KEY}" "${BASE}/collections/<name>/snapshots" | python3 -m json.tool

# List snapshots
curl -s -H "api-key: ${KEY}" "${BASE}/collections/<name>/snapshots" | python3 -m json.tool

Quick Discovery (run in parallel to inspect everything)

BASE="${QDRANT_API_URL}" && KEY="${QDRANT_API_KEY}"

# All collections with counts
curl -s -H "api-key: $KEY" "$BASE/collections" | python3 -m json.tool

# Details on a specific collection
curl -s -H "api-key: $KEY" "$BASE/collections/tro-child-hybrid-v1" | python3 -m json.tool

# Sample points from a collection
curl -s -X POST -H "api-key: $KEY" -H "Content-Type: application/json" \
  "$BASE/collections/tro-child-hybrid-v1/points/scroll" \
  -d '{"limit": 3, "with_payload": true, "with_vector": false}' | python3 -m json.tool