Qdrant Manager

A general-purpose command-line tool for managing Qdrant vector database collections and documents. Simplifies common Qdrant management tasks through a CLI interface.

Features

• Create, delete, and list collections
• Get detailed information about collections
• Retrieve points from collections with flexible query options
• Batch operations on documents:
  • Add fields to documents
  • Delete fields from documents
  • Replace fields in documents
• Support for JSON path selectors for precise document modifications
• Multiple configuration profiles support

Installation

# From PyPI
pipx install qdrant-manager

# From source
git clone https://github.com/allenday/qdrant-manager.git
cd qdrant-manager
pipx install -e .

Configuration

When first run, qdrant-manager will create a configuration file at:

• Linux/macOS: ~/.config/qdrant-manager/config.yaml
• Windows: %APPDATA%\qdrant-manager\config.yaml

You can edit this file to add your Qdrant connection details and schema configuration:

default:
  connection:
    url: localhost
    port: 6333
    api_key: ""
    collection: my-collection
  
  vectors:
    size: 256
    distance: cosine
    indexing_threshold: 0
  
  # Optional payload indices for optimized searching
  payload_indices:
    - field: category
      type: keyword
    - field: created_at
      type: datetime
    - field: price
      type: float

production:
  connection:
    url: your-production-instance.region.cloud.qdrant.io
    port: 6333
    api_key: your-production-api-key
    collection: production-collection
  
  vectors:
    size: 1536  # For OpenAI embeddings
    distance: cosine
    indexing_threshold: 1000
  
  payload_indices:
    - field: product_id
      type: keyword
    - field: timestamp
      type: datetime

Each profile can define its own:

• Connection settings
• Vector configuration (size, distance metric, indexing behavior)
• Payload indices for optimized search performance

The YAML format makes it easy to maintain a clean, organized configuration across multiple environments.

You can switch between profiles using the --profile flag:

qdrant-manager --profile production list

You can also override any setting with command-line arguments.

Usage

qdrant-manager <command> [options]

Available Commands:

• create: Create a new collection
• delete: Delete an existing collection
• list: List all collections
• info: Get detailed information about a collection
• batch: Perform batch operations on documents
• get: Retrieve points from a collection
• config: View available configuration profiles

Connection Options:

--profile PROFILE  Configuration profile to use
--url URL          Qdrant server URL
--port PORT        Qdrant server port
--api-key API_KEY  Qdrant API key
--collection NAME  Collection name

Examples:

# List all collections
qdrant-manager list

# Create a new collection with custom settings
qdrant-manager create --collection my-collection --size 1536 --distance euclid

# Get info about a collection
qdrant-manager info --collection my-collection

# Retrieve points by ID
qdrant-manager get --ids "1,2,3" --with-vectors

# Retrieve points using a filter and save as CSV
qdrant-manager get --filter '{"key":"category","match":{"value":"product"}}' \
  --format csv --output results.csv

# Add a field to documents matching a filter
qdrant-manager batch --filter '{"key":"category","match":{"value":"product"}}' \
  --add --doc '{"processed": true}'

# Delete a field from specific documents
qdrant-manager batch --ids "doc1,doc2,doc3" --delete --selector "metadata.temp_data"

# Replace fields in documents from an ID file
qdrant-manager batch --id-file my_ids.txt --replace --selector "metadata.source" \
  --doc '{"provider": "new-provider", "date": "2025-03-31"}'

# Switch between profiles
qdrant-manager --profile production list

Changelog

v0.1.6

• Improved pagination for large result sets to prevent timeouts
• Fixed empty filter handling to correctly match all documents
• Added retry logic for failed batch retrievals
• Better cloud connection detection and handling
• Improved logging with detailed progress updates
• Changed invalid filter structure message from error to warning

v0.1.5

• Fixed packaging issue to include command modules

v0.1.4

• Added get command to retrieve and export points from collections
• Refactored CLI code into separate modules for better maintainability
• Improved test coverage to over 85%
• Fixed various bugs in tests and command handling

v0.1.3

• Fixed bug in collection creation with payload indices

v0.1.2

• Added comprehensive test coverage
• Improved error handling

v0.1.1

• Initial release with basic functionality
• Added configuration profiles support

License

Apache-2.0