Dev

[connorakey]

This pull request adds robust support for using cloud-based, OpenAI-compatible APIs for both text cleaning (disfluency removal) and transcription. It introduces new environment variables and configuration options, updates documentation, and implements backend logic to route requests to cloud providers when selected. The changes also improve the clarity of the text cleaning prompt and enhance workflow automation.

Cloud API integration and configuration:

• Added support for using any OpenAI-compatible API (e.g., OpenAI, Together AI, OpenRouter, Groq, Azure OpenAI) for both text cleaning and audio transcription by introducing the cloud option for LLM_PROVIDER and TRANSCRIPTION_PROVIDER. New environment variables (CLOUD_API_BASE_URL, CLOUD_API_KEY, TRANSCRIPTION_MODEL_NAME) enable flexible configuration. ([[1]](https://github.com/connorakey/openwhisper/pull/5/files#diff-361f2d0f55c49b270125820efb10f0d9433ced883e9d59b1f58c27b877dc2080L9-R25), [[2]](https://github.com/connorakey/openwhisper/pull/5/files#diff-8b801df796d0872b733d89acd897feeada5463acc50545adb56aa086048e46f5R6-R21), [[3]](https://github.com/connorakey/openwhisper/pull/5/files#diff-b57590968a12cee85a37c1b91d8cc7092cd8b68e50b242c39a178121db82a797L816-R859))
• Updated documentation (docs/API.md) with detailed instructions and example configurations for using cloud providers, including benefits and considerations. ([[1]](https://github.com/connorakey/openwhisper/pull/5/files#diff-b57590968a12cee85a37c1b91d8cc7092cd8b68e50b242c39a178121db82a797L816-R859), [[2]](https://github.com/connorakey/openwhisper/pull/5/files#diff-b57590968a12cee85a37c1b91d8cc7092cd8b68e50b242c39a178121db82a797R952-R1078))

Backend implementation:

• Modified text_cleaner.py to support sending disfluency removal requests to cloud APIs, including handling API endpoints, authorization headers, and response parsing for OpenAI-compatible chat endpoints. The cleaning prompt was rewritten for clarity and specificity. ([[1]](https://github.com/connorakey/openwhisper/pull/5/files#diff-5d99e68f5b6f59488b8b75a2a0db33f538c8ef197616a29809becd99674f4faeR10-R45), [[2]](https://github.com/connorakey/openwhisper/pull/5/files#diff-5d99e68f5b6f59488b8b75a2a0db33f538c8ef197616a29809becd99674f4faeR55-R61), [[3]](https://github.com/connorakey/openwhisper/pull/5/files#diff-5d99e68f5b6f59488b8b75a2a0db33f538c8ef197616a29809becd99674f4faeL49-R88))
• Updated transcription.py to support cloud-based audio transcription using OpenAI-compatible APIs, with logic to select the provider based on environment configuration. ([[1]](https://github.com/connorakey/openwhisper/pull/5/files#diff-8b801df796d0872b733d89acd897feeada5463acc50545adb56aa086048e46f5R6-R21), [[2]](https://github.com/connorakey/openwhisper/pull/5/files#diff-8b801df796d0872b733d89acd897feeada5463acc50545adb56aa086048e46f5R53-R91))

Dependency and workflow improvements:

• Added the requests library to backend dependencies for HTTP requests to cloud APIs. ([backend/pyproject.tomlR12](https://github.com/connorakey/openwhisper/pull/5/files#diff-9f2cf60079756e94b31dbaeb145297215236d3c0135647163c2e51b80fd81551R12))
• Enhanced the GitHub Actions workflow to support GPG-signed commits and clarified commit author configuration. ([[1]](https://github.com/connorakey/openwhisper/pull/5/files#diff-4221e7060cb5692b5e8656f8f092468895f4bc8804397e8de7cb4cfdf8d95024R45), [[2]](https://github.com/connorakey/openwhisper/pull/5/files#diff-4221e7060cb5692b5e8656f8f092468895f4bc8804397e8de7cb4cfdf8d95024R73-R89))

These changes enable flexible deployment scenarios, allowing users to leverage local or cloud resources for both LLM and transcription tasks, with comprehensive documentation and improved automation.