René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"bl omni --audio always 400: sends audio_url, endpoint requires input_audio","articleBody":"## `bl omni --audio` always 400: sends `audio_url`, OpenAI-compatible endpoint requires `input_audio`\n\n### Summary\n`bl omni --audio \u003cfile\u003e` fails for **every** Qwen-Omni model with HTTP 400. The CLI builds the audio content part as type `audio_url`, but the DashScope OpenAI-compatible `/compatible-mode/v1/chat/completions` endpoint only accepts `input_audio` for audio. So audio transcription via the CLI is completely broken.\n\n### Environment\n- bailian-cli **1.3.0**\n- Windows 11, Node (npm global install)\n- Auth: api-key from `~/.bailian/config.json`\n\n### Repro\n```bash\nbl omni --model qwen3.5-omni-flash --audio ./clip.wav --text-only \\\n  --system \"transcribe this audio verbatim\" --message \"transcribe this audio.\" --output json\n```\n\n### Actual\n```json\n{\n  \"error\": {\n    \"code\": 1,\n    \"message\": \"Invalid value: audio_url. Supported values are: 'text','image_url','video_url' and 'video'.\",\n    \"http_status\": 400,\n    \"api_code\": \"invalid_request_error\"\n  }\n}\n```\nReproduced identically on `qwen3.5-omni-flash`, `qwen3-omni-flash`, `qwen-omni-turbo`, `qwen2.5-omni-7b` — i.e. not model-specific, it's the request shape.\n\n### Root cause\nThe CLI almost certainly emits:\n```json\n{ \"type\": \"audio_url\", \"audio_url\": { \"url\": \"data:audio/wav;base64,...\" } }\n```\nThe OpenAI-compatible Qwen-Omni API instead expects the `input_audio` shape:\n```json\n{ \"type\": \"input_audio\", \"input_audio\": { \"data\": \"data:audio/wav;base64,...\", \"format\": \"wav\" } }\n```\n\n### Fix\nChange the audio content part builder in the `omni` command from `audio_url` -\u003e `input_audio`, with `{ data, format }` (format inferred from file extension: wav/mp3/opus/pcm). I verified the `input_audio` shape works against the same endpoint/key with a hand-rolled streaming request (text-only modality + `stream_options.include_usage`) across all four models above, so only the CLI's content-part construction needs changing.\n\nRef: https://help.aliyun.com/zh/model-studio/qwen-omni (OpenAI-compatible audio input uses `input_audio`).","author":{"url":"https://github.com/a2fapower","@type":"Person","name":"a2fapower"},"datePublished":"2026-06-12T07:49:28.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":1},"url":"https://github.com/54/cli/issues/54"}