v1.3.2

Added

• Search connectors — Perplexity, Google Custom Search, Bing
• OCR connectors — Azure Vision, Google Vision, AWS Textract, Tesseract
• Image Generation connectors — OpenAI (gpt-image-1), FLUX (Together AI), Stable Diffusion, Ideogram
• Speech & Audio connectors — ElevenLabs, OpenAI TTS, Azure Speech, Google TTS
• Video Generation connectors — Runway, Kling, Pika (async job polling included)
• Music Generation connectors — Suno, Udio (async job polling included)
• CLI --param key=value flag — inject agent parameters at runtime; substitutes {{param_name}} placeholders in agent prompts (repeatable flag)
• onToolResult hook — tool call results now printed to terminal with ↳ prefix so you can see exactly what each tool returned
• Cross-platform runtime — oe-runtime-win.exe (Windows), oe-runtime-linux (Linux), oe-runtime-macos (macOS) all built automatically on release
• Connector catalog now at 2,673 entries across 45+ categories

Fixed

• OpenAI image generation: removed unsupported style and response_format params that caused 400 errors
• Image generation errors now returned as strings to the LLM instead of throwing (agent sees the error and can retry)
• Default OpenAI image model changed from dall-e-3 to gpt-image-1