Regenerate code from 2025-10, implement schema/read_capacity in BYOCSpec
#614
Conversation
…Schema to use MetadataSchema, add support for schema and read_capacity to BYOCSpec and make sure things are wired properly in request_factory.py
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
…ess and BYOC to remove duplication, fix bug when calling parse_non_empty_args for BYOC in __parse_index_spec, add unit tests to cover schema parsing logic
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 2 potential issues.
Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.
| else: | ||
| # If not a dict, create with default filterable=True | ||
| fields[field_name] = MetadataSchemaFields(filterable=True) | ||
| byoc_args["schema"] = MetadataSchema(fields=fields) |
There was a problem hiding this comment.
Duplicated schema conversion bypasses existing __parse_schema helper
Low Severity
The new ByocSpec object schema handling (lines 358–371) duplicates the inline schema conversion logic from the ServerlessSpec object path (lines 316–328). Both could simply call __parse_schema(spec.schema), which already exists and is correctly used in the dict-based paths (lines 265 and 298) and in create_index_for_model_request (line 483). If __parse_schema is updated in the future, these inline copies won't benefit, creating a maintenance risk and inconsistency.
Additional Locations (1)
| raise PineconeApiAttributeError( | ||
| f"`{var_name}` is a read-only attribute. Use `from_openapi_data` to instantiate " | ||
| f"class with read only attributes." | ||
| ) |
There was a problem hiding this comment.
BYOC response deserialization missing read_capacity handling
High Severity
The PR adds read_capacity to ByocSpec and creates a new ByocSpecResponse model (typed with ReadCapacityResponse), but never updates the BYOC response deserialization in index_model.py. The serverless path uses ServerlessSpecResponse and manually deserializes nested read_capacity to work around "broken oneOf matching logic" in deserialize_model. The BYOC path still uses the request model ByocSpec (typed with ReadCapacity, not ReadCapacityResponse) and has no manual read_capacity handling. When a BYOC index with DRN is described or listed, read_capacity in the API response will likely fail to deserialize correctly.
1 participant
Problem
When the Python client was released for
2025-10, the spec did not yet includeBYOCSpec.read_capacity, so the ability to configure BYOC indexes with DRN is limited through the SDK.Additionally, I noticed we never pulled in
BYOCSpec.schemaeither, so that can be added in as well.Solution
2025-10spec to pull in most recent changes (make generate-oas). We shouldn't need to regenerate grpc for this as the changes are limited toBYOCSpecin the control plane.schemaandread_capacitytoBYOCSpec, and make sure things are wired properly inrequest_factory.py.BackupModelSchemato useMetadataSchema. This was due to an error in the spec wherein theMetadataSchemain the OpenAPI spec did not have anx-component-name. That was fixed here: https://github.com/pinecone-io/apis/pull/417. It's confusing to haveBackupModelSchemaused in all these areas that aren't backup-related.Also pulled in some dependency bumps based on dependabot alerts:
aiohttp(2.6.3)urllib3(3.13.3)filelock(3.21.0)virtualenv(20.31.1)Type of Change
Test Plan
CI - unit, integration, etc
Note
Medium Risk
Touches generated API models and request-building for index creation/configuration; type/model renames and new BYOC fields could cause subtle compatibility or serialization issues if downstream code relies on the old schema types.
Overview
Regenerates the OpenAPI client from the
2025-10spec (and bumps the codegen target inMakefile), updating the embedded API repo SHA and refreshing generated docs/comments.Adds BYOC parity with serverless for index configuration by supporting
read_capacityandschemaonByocSpecand wiring both throughPineconeDBControlRequestFactory(including shared__parse_schemahandling for both direct field maps and{fields: ...}wrapper format).Renames the OpenAPI model used for metadata indexing configuration from
BackupModelSchema/BackupModelSchemaFieldstoMetadataSchema/MetadataSchemaFieldsacross control-plane models and public SDK type hints, adds a generatedByocSpecResponsemodel, and updates unit tests to cover schema parsing and BYOC/serverless spec variants. Also includes dependency lockfile updates (e.g.,aiohttp,urllib3,filelock,virtualenv) and a Ruff config tweak.Written by Cursor Bugbot for commit 6525664. This will update automatically on new commits. Configure here.