SDK Revamp with CI/CD & Bug Fixes

[Khurdhula-Harshavardhan]

Python SDK Revamp

CI/CD & Linting

• Added GitHub Actions workflow: .github/workflows/ci.yml to run Ruff and tests on push/PR.
• Added Ruff config: .github/ruff.toml to standardize style and rules.
• Removed unused biome.json.

Code Organization

• Normalized and reordered imports across key modules for consistency/readability:
  • audio.py, classification.py, embedding.py, image_generation.py, prediction.py, prompt_engine.py.
  • See inline diffs for exact changes.

API & Bug Fixes

• Multipart uploads: Reworked AsyncRequest to use aiohttp.FormData, correctly set/remove Content-Type for multipart, and serialize param types reliably (bools, lists/dicts, etc.).
• Speech-to-text: Simplified path/header handling and enforce consistent parameter usage in audio.py (sync + async).
• Embedding v2: Removed unsupported token_overflow_mode="chunk"; restrict to valid values only.
• Image generation: Support responses as URL or bytes; clarify return type docs.
• Prediction typing: Allow float values in dataset payloads; fix minor typos and docstrings.

Docs & Examples

• Updated VOCR example image URL in README.md.

Breaking Changes

• embedding_v2: token_overflow_mode="chunk" is no longer accepted.

Testing Notes

• CI runs Ruff + tests across modules.
• Verified multipart behavior with aiohttp.FormData and header handling.
• Smoke-tested speech-to-text paths and image-generation return types.

Checklist

• CI passing (lint + tests)
• No user-visible (user exp) API changes except noted breaking change
• Docs+comments updated
• Imports standardized