Open Model Inference - Together AI
API capability pack for bounded Together AI REST API operations against https://api.together.ai/v1 only. Enroll a Together AI bearer API-key connection named togetherai; no credential is embedded in this manifest. The pack reads the model catalog, creates non-streaming chat completions, legacy text completions, embeddings, rerank responses, prompt-only image generations, and reads deployment, deployment-log, and hardware inventory for open-model inference, RAG preparation, relevance review, image ideation, model selection, and dedicated-capacity operations. Generation, embedding, rerank, and image calls are read-tier inference calls that consume Together AI quota but do not create durable customer resources through this pack. The pack intentionally excludes streaming, tool/function calling, arbitrary extra request fields, OpenAI Assistants/Threads/Runs, file upload/download/delete, batches, audio speech/transcription/translation, video generation, image edits/inpainting/reference images/image LoRAs, disable_safety_checker, model upload, fine-tune job mutation, deployment create/update/delete, endpoint/secret/storage mutation, code interpreter/sandbox APIs, external base URLs, custom host switching, and arbitrary Together AI API passthrough.
What this pack installs
- Together AI chat completion brief Recipe
- Together AI RAG relevance brief Recipe
- Together AI image generation brief Recipe
- Together AI operations digest Recipe
- togetherai Ingredient
Trust & control
What installing this whole pack would let it do. Recued grants these permissions at install — review them there before approving.