[robgruen]

Summary

Add backward-compatible Python text/image prompts through the existing Chat Completions model and translator-context interfaces, implementing the accepted P2 scope.

• Export precise PromptContent, TextPromptContent, ImagePromptContent, and ImageUrl dictionary types while preserving existing string callers.
• Share prompt normalization and structural validation across HTTP and configured-client models. Reject malformed parts before authentication or dispatch, without echoing prompt content. Preserve exact references, optional detail, ordering, and caller-owned data; images are supported only in user messages.
• Preserve string-only translation requests, existing schema/request placement, repair roles, strict result validation, and rejected-output payloads. Keep completion results text-only with separate HTTP response typing.
• Include the owner-approved optional temperature setting across model factories: retain 0.0 by default; None omits the parameter for reasoning-model compatibility.
• Add deterministic offline behavior/static-type coverage and Python documentation examples. No dependency-manifest or lockfile changes, required provider SDK, image retrieval/processing, or Responses backend.

Fixes #517.

Verification

• hatch run test on Python 3.11.18: 418 passed, 3 Python-version-dependent skips; 19 snapshots passed. Rerun before publication.
• npm run check with Node 24.22.1: 0 errors, 0 warnings. Rerun before publication.
• git diff --check: passed.
• Live Azure Chat Completions smoke tests with identity authentication and temperature=None: 10 endpoint requests passed across the HTTP/token-provider model and an actual configured AsyncAzureOpenAI adapter. Covered strings, low/high-detail mixed image prompts, immediate typed image translation, and image-preserving repair with a locally injected invalid first candidate.
• Ordinary regression tests remain offline and require no live credentials or provider SDKs.

Scope notes

The current prompt arrangement is unchanged; this does not adopt #241 or implement the separate Python Responses work in #518. Existing TypeScript image types and the contributed image example are unchanged.

[Copilot]

🟢 Approval recommended

No active review issues are reported for the backward-compatible Python text/image prompt changes.

No open issues 🎉

What changed in this PR

Adds backward-compatible Python text/image prompts to TypeChat’s existing Chat Completions models and translator context, addressing #517 without changing the prompt layout or adding a Responses backend.

Changes:

• Exports typed prompt parts and shares structural validation across HTTP and configured-client models.
• Adds optional temperature configuration, including omission with None, while keeping 0.0 as the default.
• Adds offline regression tests, static-type coverage, and documentation for image prompts and repair behavior.

File	Description
site/​src/​docs/​python/​basic-usage.md	Documents image prompts, supported limits, and temperature configuration.
python/​tests/​test_prompt_types.py	Checks public prompt types and documentation example consistency.
python/​tests/​test_multimodal_model.py	Covers multimodal validation, forwarding, repair, and existing model controls.
python/​tests/​test_create_language_model.py	Checks temperature forwarding across HTTP model factories.
python/​src/​typechat/​_internal/​model.py	Implements prompt types, shared validation, temperature options, and text-only response checks.
python/​src/​typechat/​__init__.py	Exports the new public prompt types.
python/​README.md	Adds image-prompt examples and temperature guidance.

_{Review effort: Balanced}