OpenAI Provider on Bifrost

1) Chat Completions

Primary chat path at /v1/chat/completions. See Chat Completions in Bifrost docs.

• Messages: text, image_url, input_audio; tool messages include tool_call_id
• Streaming: SSE with delta.content, delta.tool_calls, finish_reason, usage on final chunk
• cache_control stripped from messages, content blocks, and tools
• Reasoning: effort minimal/low/medium/high; minimal → low when routing to other providers

curl -X POST http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-4o-2024-11-20",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

2) Responses API

Structured output API at /v1/responses. Non-gpt-oss models use reasoning summaries; gpt-oss uses reasoning content blocks. See Responses API in Bifrost docs.

Supported tool types: functionfile_searchcomputer_use_previewweb_searchmcpcode_interpreterimage_generationlocal_shellcustomweb_search_preview. Action types zoom/region → screenshot. Response: id, status, output, usage.

curl -X POST http://localhost:8080/v1/responses \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-4o-2024-11-20",
    "input": "Hello",
    "max_output_tokens": 1024
  }'

3) Text Completions (Legacy)

Legacy API at /v1/completions — prefer Chat Completions for new work. Supports streaming. See Text Completions in Bifrost docs.

curl -X POST http://localhost:8080/v1/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-3.5-turbo-0125",
    "prompt": "Hello, my name is",
    "max_tokens": 50
  }'

4) Embeddings

/v1/embeddings — no streaming. See Embeddings in Bifrost docs.

curl -X POST http://localhost:8080/v1/embeddings \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/text-embedding-3-large",
    "input": "Hello world"
  }'

5) Speech (Text-to-Speech)

/v1/audio/speech — returns raw binary audio; streaming via SSE base64 chunks where supported. See Speech in Bifrost docs.

curl -X POST http://localhost:8080/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/tts-1",
    "input": "Hello world",
    "voice": "alloy"
  }' --output speech.mp3

6) Transcriptions (Speech-to-Text)

/v1/audio/transcriptions — multipart/form-data (not JSON). Formats: mp3, mp4, mpeg, mpga, m4a, wav, webm. Streaming supported. See Transcriptions in Bifrost docs.

curl -X POST http://localhost:8080/v1/audio/transcriptions \
  -F file=@audio.mp3 \
  -F model=openai/whisper-1

7) Image Generation

/v1/images/generations — pass-through parameters; streaming via SSE (image_generation.partial_image, image_generation.completed). See Image Generation in Bifrost docs.

curl -X POST http://localhost:8080/v1/images/generations \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/dall-e-3",
    "prompt": "A serene landscape",
    "n": 1,
    "size": "1024x1024"
  }'

8) Image Edit

/v1/images/edits — multipart/form-data with image[], optional mask; streaming via image_edit.partial_image / image_edit.completed. See Image Edit in Bifrost docs.

9) Image Variation

/v1/images/variations — multipart/form-data; no streaming. Only the first image is sent upstream. See Image Variation in Bifrost docs.

10) Files API

Upload, list, retrieve, delete, and download files. See Files API in Bifrost docs.

• GET /v1/files — list with purpose, limit, after, order
• GET /v1/files/{file_id} — metadata
• DELETE /v1/files/{file_id}
• GET /v1/files/{file_id}/content — download

curl -X POST http://localhost:8080/v1/files \
  -F "file=@document.pdf" \
  -F "purpose=assistants"

11) Batch API

Async batch jobs at /v1/batches. Statuses: validating, failed, in_progress, finalizing, completed, expired, cancelling, cancelled. See Batch API in Bifrost docs.

• GET /v1/batches/{batch_id} — retrieve
• POST /v1/batches/{batch_id}/cancel — cancel
• Results: download output file via Files API when status is completed; parse JSONL BatchResultItem lines

12) List Models

GET /v1/models — no request body. Model IDs in responses are prefixed with openai/; results aggregate across configured API keys. See List Models in Bifrost docs.

curl http://localhost:8080/v1/models

13) Video Generation

Sora-style video jobs at /v1/videos. Job statuses: queued → in_progress → completed / failed. See Video Generation in Bifrost docs.

curl -X POST http://localhost:8080/v1/videos \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/sora-2",
    "prompt": "A cat walking in the rain"
  }'

Common error codes

HTTP status to OpenAI error type mapping from Bifrost docs.