[TASK] Refactor the SDK to follow the current OpenAPI spec

[edenreich]

Summary

Deprecate features that do not exist - simply delete them, no backward compatibility is needed.

The goal is to ensure there is no drift between the OpenAPI spec and the SDK, including the documentation.

Acceptance Criteria

• Discrepancies between current SDK and OpenAPI spec are fixed
• Image URL for vision models is implemented
• Missing providers are added
• Documentation is adjusted (if applicable)
• Tests are adjusted / added (if applicable)

[edenreich]

@claude Can you work on this?

[claude]

Claude finished @edenreich's task —— View job • claude/issue-22-20251124-1302 • Create PR ➔

---

Refactoring SDK to Match OpenAPI Spec

• Download and analyze the OpenAPI spec
• Compare SDK with OpenAPI spec to identify discrepancies
• Fix discrepancies between current SDK and OpenAPI spec
• Implement image URL support for vision models
• Add missing providers
• Update documentation if needed
• Update/add tests as needed
• Push changes and create PR

---

Summary:

After analyzing the current OpenAPI spec against the SDK, I found that the SDK is already well-aligned with the specification:

1. Providers ✅ - All 10 providers match: ollama, ollama_cloud, groq, openai, cloudflare, cohere, anthropic, deepseek, google, mistral

2. Image/Vision Types ✅ - Already implemented in generated types: ContentPart, TextContentPart, ImageContentPart, ImageURL with detail enum (auto, low, high)

3. All Schemas ✅ - Types are auto-generated from OpenAPI spec and match

Changes Made:

• Added documentation for Vision / Image Input in README.md with usage examples
• Added test case for vision/image content in chat completions
• Verified types regeneration produces identical output (SDK was up to date)

|