Refresh API specs, add query parameter support, bump to 0.10.0

[cretz]

🚀 What

• Regenerated the client from the latest management, inference, and model-config API specs
• GET methods now accept a typed request= model of their query parameters (pagination, filters, log/metric ranges, etc.), which was previously ungenerated and unusable
• Bumped version to 0.10.0
  • Was 0.9.1 (hence branch name) but updated to 0.10.0 when we chose to rename body kwarg to request for non-GET API calls

💥 Two backwards-incompatible changes mentioned below

💻 How

• Pulled updated API specs and regenerated all models and clients
• Fixed the model postprocessor's RootModel regex, which mishandled nested brackets in wrapped types
• Preprocess synthesizes one *Request schema per GET operation's query params (named after the method, e.g. get_users -> GetUsersRequest), reusing each param's own schema so datamodel-code-generator types every field (enums, arrays, nullable, constraints) and shares existing enum models
• clientgen renders that model as a single keyword-only argument, serialized to the query string; the arg is optional when no field is required
• Errors at generation time if any operation ever has both a request body and query params (our contract: GET is query-only, other methods body-only)
• Request bodies and query params are now serialized with exclude_unset so only caller-set fields go on the wire and the server applies its own defaults
• 💥 Renamed the body= keyword argument to request= on every client method
• 💥 Unset fields are no longer sent (previously every field, including defaults and explicit nulls, was serialized); callers relying on the client to send default values must now set them explicitly

🔬 Testing

• Added unit tests for a GET with query params (including that an unset non-null default like limit is dropped) and for omitting the request entirely
• Verified serialization matches the baseten backend's parsing: repeated params for lists, true/false for bools, enum values, and omitted-means-server-default

[CLAassistant]

All committers have signed the CLA.

[Copilot]

Pull request overview

This PR updates the SDK to match refreshed OpenAPI specs and cuts a new 0.9.1 release by regenerating the management/inference clients & models, updating the apigen post-processor, and aligning tests with new API constraints.

Changes:

• Bump package version to 0.9.1 and update lockfile.
• Update OpenAPI specs and regenerate management API models/client, including new resources (e.g., endpoints, metrics/log query filters, patches, etc.).
• Extend model config schema/models (e.g., EgressRestrictions) and adjust apigen post-processing for nested bracket cases.

Reviewed changes

Copilot reviewed 9 out of 12 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
uv.lock	Updates editable package version to 0.9.1.
pyproject.toml	Bumps SDK version to 0.9.1.
tests/client/test_management.py	Updates test fixture to include newly-required Secret.id.
scripts/apigen/specs/inference.json	Updates inference auth scheme to BearerAuth and updates security scheme definition.
scripts/apigen/specs/config.schema.json	Adds EgressRestrictions schema and extends weights source scheme docs (cw://).
scripts/apigen/postprocess.py	Tweaks RootModel postprocessing regex to better handle nested brackets / class boundaries.
baseten/client/modelconfig/_models.py	Adds EgressRestrictions model and wires it into Runtime.
baseten/client/modelconfig/init.py	Exports EgressRestrictions.
baseten/client/managementapi/_models.py	Large regen: adds/updates many schemas (secrets, logs/metrics requests, endpoints, patching, etc.).
baseten/client/managementapi/_client.py	Large regen: adds/updates endpoints and request methods (sync + async).
baseten/client/managementapi/init.py	Updates export surface to include newly generated models.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.