fix(client): drop v1-incompat discriminator on InputItem (b48)

[scale-ballen]

Summary

Drop Field(discriminator=\"type\") from InputItem.__root__ in gen/openai.py. Under pydantic v2 + v1 compat, the discriminator on this union triggers:

pydantic.v1.errors.ConfigError: Field 'type' is not the same for all submodels of 'Item'

at module-load time, breaking any consumer that imports llmengine (e.g. egp-api-backend).

Background

This is the exact fix @brandon.allen applied in #776 (1d50898 'feat: support openai 2 in python client'). A later commit (bd8a98f 'fix: address python client review feedback') ran codegen which wiped the manual fix. b47 (this repo's main) still carries the bug; b48 re-applies the patch.

Impact

• JSON in/out: identical. Every submodel has a Literal[...] type field, so validation still selects the correct submodel by value.
• Validation: O(n) try-each instead of O(1) jump. Negligible for typical usage.
• Error messages on invalid input: generic union error instead of 'invalid discriminator value'.
• Downstream: egp-api-backend (the immediate consumer) doesn't use InputItem at all — it only uses Completion and Model. Zero runtime impact there.

Other 9 discriminator=\"type\" sites in gen/openai.py are left intact; they don't recurse into RootModel unions so they don't hit the v1-compat failure.

Local verification

Built wheel, installed into a pydantic==2.11.9 venv (matches egp-api-backend's pin):

• Before: from llmengine import Completion, Model → ConfigError
• After: imports cleanly; ChatCompletionV2Request/Response work; all 9 Item submodels import; InputItem/Item/ItemResource/OutputItem1 import; InputMessage instantiation + validation roundtrips

Follow-ups (separate tickets)

• Fix OpenAPI schema or datamodel-code-generator config so next regen doesn't re-introduce this
• Add import smoke test to llm-engine CI under pydantic 2 + v1 compat so regressions fail fast

Test plan

• Local import smoke test under pydantic==2.11.9
• Publish b48 to CodeArtifact
• egp-api-backend CI passes with b48 pin (scaleapi/scaleapi#140646)

🤖 Generated with Claude Code

Greptile Summary

Re-applies the manual fix from #776 that was accidentally reverted by a codegen run: drops discriminator="type" from InputItem.__root__ because Item is itself a __root__-based union (RootModel) and does not expose a uniform type field, causing pydantic v1-compat to raise ConfigError at module-load time. The version is bumped to b48 accordingly. Validation correctness is unaffected; union resolution falls back to sequential try-each instead of O(1) discriminator lookup, which is negligible for typical payloads.

Confidence Score: 5/5

Safe to merge — the one-line fix correctly resolves a module-load crash with no JSON or validation regressions.

The change is a targeted, well-understood removal of a pydantic v1-compat incompatible option. The PR description documents root cause, local verification, and downstream impact. The only remaining feedback is a P2 style suggestion to add an explanatory comment/TODO ticket so the next codegen run is easier to catch.

No files require special attention; consider adding the inline comment with a ticket number before the next codegen run to avoid re-introducing the bug.

Important Files Changed

Filename	Overview
clients/python/llmengine/data_types/gen/openai.py	Removes discriminator="type" from InputItem.root to fix pydantic v1-compat ConfigError; fix is correct because Item is itself a root union with no uniform type field
clients/python/pyproject.toml	Version bumped from 0.0.0.beta47 to 0.0.0.beta48 to publish the fix; change is straightforward

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[InputItem.__root__] --> B{Union resolution}
    B -->|before fix| C[ConfigError at import\nItem has no uniform type field]
    B -->|after fix - sequential try-each| D[EasyInputMessage]
    B -->|after fix - sequential try-each| E[Item - __root__ union\nwith its own discriminator]
    B -->|after fix - sequential try-each| F[ItemReferenceParam]
    E --> G[InputMessage / OutputMessage / FileSearchToolCall / ...]

Loading

Prompt To Fix All With AI

This is a comment left during a code review.
Path: clients/python/llmengine/data_types/gen/openai.py
Line: 11517-11520

Comment:
**Missing workaround comment and TODO ticket link**

This is an intentional manual fix to a generated file that will be silently reverted the next time codegen runs (as happened between #776 and bd8a98f). Without an inline comment explaining *why* the discriminator was removed, future reviewers (or the codegen follow-up ticket) have no breadcrumb. Per the repo's conventions, complex workarounds should carry an explanatory comment, and TODO comments should include a ticket number so they're searchable.

```suggestion
class InputItem(BaseModel):
    # TODO(<ticket>): Remove Field() workaround once codegen is fixed to omit
    # discriminator="type" here. Item is itself a __root__ union (RootModel) and
    # does not expose a uniform `type` field, so pydantic v1-compat raises
    # ConfigError: "Field 'type' is not the same for all submodels of 'Item'"
    # at module-load time when discriminator="type" is present. Validation still
    # works correctly via sequential union try-each.
    __root__: Annotated[
        Union[EasyInputMessage, Item, ItemReferenceParam], Field()
    ]
```

**Rule Used:** Add comments to explain complex or non-obvious log... ([source](https://app.greptile.com/review/custom-context?memory=928586f9-9432-435e-a385-026fa49318a2))

**Learned From**
[scaleapi/scaleapi#126958](https://github.com/scaleapi/scaleapi/pull/126958)

How can I resolve this? If you propose a fix, please make it concise.

_{Reviews (1): Last reviewed commit: "fix(client): drop v1-incompat discrimina..." | Re-trigger Greptile}

Context used:

• Rule used - Add comments to explain complex or non-obvious log... (source)

Learned From
scaleapi/scaleapi#126958

• Rule used - Include ticket numbers in TODO comments to make cl... (source)

Learned From
scaleapi/scaleapi#126926