Install only what your task needs¶

PyCharter requires Python 3.11+. A local record check needs the Python package and a contract. Add connectors, services or an orchestrator when you use those capabilities.

Start in a virtual environment¶

python -m venv .venv
source .venv/bin/activate
python -m pip install pycharter

On Windows PowerShell, activate with .venv\Scripts\Activate.ps1. For the exact release used by the first-result examples, install "pycharter==0.0.67" instead. Check the installed version:

python -m pip show pycharter
pycharter --help

Continue with your first record check. You do not need to initialize a registry or start a service for that task.

Run a local file pipeline¶

Follow the local data-feed guide. For the public 0.0.67 release with SQLAlchemy 2.1, install its asyncio dependency explicitly:

python -m pip install "pycharter==0.0.67" "sqlalchemy[asyncio]"

This supplies greenlet, which pipeline imports require. The reviewed source checkout already includes sqlalchemy[asyncio] in its base dependencies.

Add capabilities progressively¶

Quote extras in shell commands. Combine the ones you need, for example:

python -m pip install "pycharter[pipeline,postgres]"

Extra	Use it for	Next guide
`pipeline`	Additional cloud, database and file-format connectors	Pipelines
`postgres`	PostgreSQL drivers for headless pipeline and database access	Configuration
`server`	HTTP API, UI hosting and worker dependencies	API and UI
`ai`	MCP server and optional model-provider SDKs	MCP setup
`streaming`	Streaming transports, brokers and collaboration dependencies	Streaming scope
`semantic`	Vocabulary/ontology integrations	Business concepts
`prefect`, `airflow` or `dagster`	The adapter and dependencies for your chosen orchestrator	Prefect
`otel`	OpenTelemetry API integration; supply your own SDK/exporter	Observability
`cli`	Richer terminal output	CLI reference
`docs`	Building or serving the documentation	Contributing

Extras install dependencies. They do not provision databases, brokers, hosted compute, secrets or schedules. Follow the selected integration's setup and capability status. Prefer a single orchestrator adapter that matches your environment; a broad install is unnecessary for first value.

Use a source reference deliberately¶

Current source requires Pydantic >=2.11.0,<3.0; update older explicit pins when moving to this source version. Its minimum/current wheel checks execute the README's local examples. This source qualification does not mean a newer public release or every optional deployment profile has been qualified.

The bounded FMP reference and its Prefect fixture deployment currently need the matching source checkout. Their reference extra and application assets are ahead of the public 0.0.67 release. Follow those guides' exact setup and scope; installing the public wheel does not install the reference application.

For contribution and editable source setup, use CONTRIBUTING.md. For service state, database selection and environment settings, use configuration.

If your first command fails¶

Check that python -m pip show pycharter and pycharter refer to the same virtual environment. Read the relevant command's --help, confirm the guide's package version, and install the extra required by the selected connector. For the 0.0.67 pipeline error mentioning greenlet, use the explicit asyncio install above. Keep credentials out of command output shared with others.