Local with uv

This guide installs PrivateGPT from the cloned repository using uv, a fast Python package manager. Use this if you want to modify the source, run tests, or use hot-reload during development.

For the simplest install, use the package install instead.

Core install

The default source install uses core:

$
uv sync --frozen --extra core

Add feature-specific extras only when you need them, for example uv sync --frozen --extra core --extra queue.

Auto mode (zero config)

$
curl -LsSf https://astral.sh/uv/install.sh | sh

$
git clone https://github.com/zylon-ai/private-gpt
$
cd private-gpt
$
uv sync --frozen --extra core

$
OPENAI_API_BASE=http://localhost:11434/v1 uv run private-gpt serve

Detailed model settings (custom profile)

Use Detailed Model Configuration to generate settings-model.yaml, tune model settings in more detail, and run local uv with PGPT_PROFILES=model.

Development with hot-reload

When working on the source, run with --reload to restart automatically on file changes:

$
OPENAI_API_BASE=http://localhost:11434/v1 \
>
  PGPT_PROFILES=local \
>
  uv run private-gpt serve --reload

Or use the Makefile shortcut (Unix only):

$
OPENAI_API_BASE=http://localhost:11434/v1 make dev