Installation¶
Install the package¶
Supported Python versions:
3.123.13
What gets installed¶
The package includes:
- the Python library
- the
batchorCLI - the built-in OpenAI provider integration
- SQLite and Postgres storage implementations
It does not provision external infrastructure for you. If you use Postgres or a shared artifact root, you still manage those resources yourself.
Authentication¶
For Python API usage, authentication resolution is:
OpenAIProviderConfig(api_key=...)OPENAI_API_KEY
The Python library does not auto-load .env.
The CLI loads a local .env as a convenience for operator usage, then resolves OPENAI_API_KEY.
Storage defaults¶
BatchRunner()uses SQLite durability by default.BatchRunner(storage="memory")uses ephemeral in-memory state.PostgresStorage(...)is available as an opt-in durable control-plane backend when you need shared state.
Artifact defaults follow storage:
- SQLite defaults to a local artifact root beside the SQLite database
- explicit
temp_root=orartifact_store=can override that - Postgres users should usually provide an explicit shared
LocalArtifactStore(...)
If you use Postgres across machines or fresh processes, configure a shared artifact root so replayable request artifacts remain visible everywhere they need to be.
Where to go next¶
- Read Use Cases for concrete end-to-end examples.
- Read Architecture for the runtime model and diagrams.
- Read Storage & Runs for durable-run and artifact semantics.
- Read Python API if you are integrating
batchorinto application code. - Read CLI Usage if you want file-backed operator workflows.