Fix IndexEmbed.as_dict() passing None values to CreateIndexForModelRequestEmbed

[MaxwellCalkin]

Summary

Fixes #497.

When creating an integrated index via create_index_for_model() with an IndexEmbed instance that only has required fields set (model, field_map), embed.as_dict() returns self.__dict__ which includes all attributes — including metric: None. This dict is then unpacked as CreateIndexForModelRequestEmbed(**parsed_embed), and the OpenAPI model's strict validation rejects None for the metric field (expects str, not NoneType).

Reproduction:

from pinecone import Pinecone, IndexEmbed

pc = Pinecone(api_key="...")
pc.db.index.create_for_model(
    name="test",
    cloud="aws",
    region="us-east-1",
    embed=IndexEmbed(model="multilingual-e5-large", field_map={"text": "my_text_field"})
)
# TypeError: Invalid type for variable 'metric'. Required value type is str ...

Fix: Filter out None values from the dict returned by embed.as_dict() before passing it to CreateIndexForModelRequestEmbed:

# Before:
parsed_embed = embed.as_dict()

# After:
parsed_embed = {k: v for k, v in embed.as_dict().items() if v is not None}

This ensures only explicitly set fields are forwarded to the OpenAPI model constructor, letting the server apply its own defaults for omitted optional fields like metric.

Test plan

• Create an index with IndexEmbed(model=..., field_map=...) (no metric) — should succeed instead of raising TypeError
• Create an index with IndexEmbed(model=..., field_map=..., metric="cosine") — should continue working as before
• Existing unit tests in tests/unit/models/test_index_embed.py pass unchanged

---

This PR was authored by Claude, an AI assistant by Anthropic, as part of an open-source contribution effort. See my work for more details.

---

Note

Low Risk
Small, localized change to request-building that only affects how optional IndexEmbed fields are serialized; low chance of unintended side effects beyond omitting previously-sent nulls.

Overview
Fixes create_index_for_model_request() to drop None values from IndexEmbed.as_dict() before constructing CreateIndexForModelRequestEmbed, preventing OpenAPI type validation errors when optional fields (e.g. metric) are unset.

^{Written by Cursor Bugbot for commit ae91e83. This will update automatically on new commits. Configure here.}