qdrant-rest-cli

A full-coverage CLI for the Qdrant vector database REST API. Every endpoint in Qdrant's OpenAPI spec, exposed as a typed command. Generated from Qdrant's official OpenAPI spec using openapi-cli-gen.

Why

Qdrant ships excellent client SDKs (Python, Rust, Go, JS), but no REST CLI. For shell scripting, Makefile targets, CI pipelines, and interactive ad-hoc queries, most people drop to raw curl and hand-build JSON payloads.

This CLI gives you the entire Qdrant REST API as flat shell commands with typed flags — collections, points, search, snapshots, distributed cluster ops — without writing any Python or curl boilerplate. When Qdrant adds endpoints, a regeneration picks them up.

Install

pipx install qdrant-rest-cli

# Or with uv
uv tool install qdrant-rest-cli

Setup

Point it at your Qdrant instance (default is http://localhost:6333):

export QDRANT_REST_CLI_BASE_URL=http://localhost:6333

If your instance uses an API key:

export QDRANT_REST_CLI_API_KEY=your-api-key

The CLI sends it as the api-key header automatically, matching Qdrant's security scheme.

HTTPS with Self-Signed Certs

If your Qdrant instance uses TLS with a self-signed or internal certificate:

# Skip certificate verification
export QDRANT_REST_CLI_VERIFY_SSL=false

# Or point to your CA bundle
export QDRANT_REST_CLI_CA_CERT=/path/to/ca-bundle.pem

Quick Start

All commands below have been verified against a live Qdrant instance.

# Server health
qdrant-rest-cli service root
qdrant-rest-cli service healthz

# List collections
qdrant-rest-cli collections get-collections

# Create a collection (4-dim vectors, cosine distance)
qdrant-rest-cli collections create \
  --collection-name pets \
  --vectors '{"size": 4, "distance": "Cosine"}'

# Upsert points with payloads (use --root for the full JSON request body)
qdrant-rest-cli points upsert --collection-name pets --root '{
  "points": [
    {"id": 1, "vector": [0.9, 0.1, 0.1, 0.1], "payload": {"name": "cat",  "species": "mammal"}},
    {"id": 2, "vector": [0.1, 0.9, 0.1, 0.1], "payload": {"name": "dog",  "species": "mammal"}},
    {"id": 3, "vector": [0.1, 0.1, 0.9, 0.1], "payload": {"name": "bird", "species": "avian"}}
  ]
}'

# Count points
qdrant-rest-cli points count --collection-name pets

# Semantic search — return results WITH their payloads
qdrant-rest-cli search points \
  --collection-name pets \
  --vector '[0.85, 0.15, 0.1, 0.1]' \
  --limit 2 \
  --with-payload true

# Scroll through all points
qdrant-rest-cli points scroll --collection-name pets --limit 10

# Get a collection's info
qdrant-rest-cli collections get-collection --collection-name pets

# Delete the collection
qdrant-rest-cli collections delete --collection-name pets

Discover All Commands

# Top-level groups
qdrant-rest-cli --help

# Commands in a group
qdrant-rest-cli collections --help

# Flags for a specific command
qdrant-rest-cli search points --help

Output Formats

Every command accepts --output-format:

qdrant-rest-cli collections get-collections --output-format table
qdrant-rest-cli collections get-collections --output-format yaml
qdrant-rest-cli collections get-collections --output-format raw

Command Groups

Passing Complex JSON Bodies

Most Qdrant write endpoints take deeply nested request bodies (lists of points, complex filters, hybrid queries). For these, pass the entire body as a JSON string via --root:

qdrant-rest-cli points upsert --collection-name pets --root '{"points":[...]}'

qdrant-rest-cli search points --collection-name pets --root '{
  "vector": [0.85, 0.15, 0.1, 0.1],
  "limit": 5,
  "with_payload": true,
  "filter": {"must": [{"key": "species", "match": {"value": "mammal"}}]}
}'

Flat endpoints (like collections create) accept typed flags directly — use whichever is clearer for a given call.

Real Example: End-to-End Vector Search

$ qdrant-rest-cli collections create \
    --collection-name movies \
    --vectors '{"size": 4, "distance": "Cosine"}'
{"result": true, "status": "ok", "time": 0.012}

$ qdrant-rest-cli points upsert --collection-name movies --root '{
    "points": [
      {"id": 1, "vector": [0.9, 0.1, 0.1, 0.1], "payload": {"title": "The Matrix"}},
      {"id": 2, "vector": [0.1, 0.9, 0.1, 0.1], "payload": {"title": "Titanic"}}
    ]
  }'
{"result": {"operation_id": 0, "status": "completed"}, "status": "ok"}

$ qdrant-rest-cli search points --collection-name movies \
    --vector '[0.85, 0.15, 0.1, 0.1]' \
    --limit 2 \
    --with-payload true
{
  "result": [
    {"id": 1, "score": 0.998, "payload": {"title": "The Matrix"}},
    {"id": 2, "score": 0.299, "payload": {"title": "Titanic"}}
  ],
  "status": "ok"
}

How It Works

This package is a thin wrapper:

• Embeds the Qdrant OpenAPI spec (spec.yaml)
• Delegates CLI generation to openapi-cli-gen at runtime
• Default base URL: http://localhost:6333

Since it's spec-driven, new Qdrant endpoints show up automatically on regeneration — no manual wrapping to fall behind.

License

MIT. Not affiliated with Qdrant — this is an unofficial community CLI built on top of their public OpenAPI spec.