pse — Python Search Engine

Find Python packages across PyPI, Conda channels, and other compatible indexes.

pse is a cross-source command-line Python package search tool. It searches Python packages without installing packages or using a package manager at runtime.

Features

• Search PyPI by default, or select exact PyPI and Conda sources.
• Detect PyPI-compatible indexes and Conda channels from source URLs.
• Produce human-readable tables or structured JSON.

Usage

Using pse on-the-fly

You can run pse directly with uvx, pipx or pixi exec without a permanent install.

uvx pse pkg

pipx run pse pkg

pixi exec -s uv uvx pse pkg

Installing pse for repeated use

You can install pse as global command with uv tool install for repeated use:

uv tool install pse
pse pkg

Use-cases

Search PyPI by default:

pse pkg

Search only conda-forge:

pse pkg --source conda-forge

Search a PyPI-compatible source URL:

pse pkg --source https://download.pytorch.org/whl/cu126

Search a Conda channel URL:

pse pkg --source https://conda.anaconda.org/bioconda

Use multiple exact sources:

pse pkg --source pypi --source conda-forge --source https://download.pytorch.org/whl/cu126

Sources

PyPI

pse searches the PyPI Simple API by default. The pypi source name and source URLs that expose a compatible Simple index, such as the PyTorch wheel index, use the PyPI adapter.

Conda

pse searches Conda only when conda-forge or a detected Conda source URL is selected. It reads current_repodata.json.bz2 for the selected channel and platform, and also searches noarch unless the requested platform is already noarch.

The built-in conda-forge source maps to conda.anaconda.org. Full channel URLs are detected automatically; source URLs must point to channel roots.

The reduced current_repodata metadata is fast, but it may not contain every historical Conda package version.

JSON output

pse can produce structured JSON output for programmatic consumption and agent friendly use. The JSON output contains a results array with one entry per source. Each entry contains the source name, package name, and an array of available versions:

{
  "query": "torch",
  "platform": "win-64",
  "results": [
    {
      "source": "pypi",
      "name": "torch",
      "versions": [
        {
          "version": "2.13.0",
          "build": null,
          "platform": null
        }
      ]
    }
  ]
}

If a source fails, an errors array is added to the document.