obdt — OneBudd Dataset Tool

CLI tool to upload audio files to the OneBudd Emotion Labeler API and get back a labeled dataset as a ZIP file.

Install

npm install -g @onebudd/obdt

Requires Node.js 18+

First-time Setup

Run any command for the first time and it will prompt you for your API key and server URL — these are saved to ~/.obdt/config.json and reused automatically.

obdt "audio.mp3"

  🔧 First-time setup — configure your OneBudd connection:

🔑 OneBudd API key: <your-api-key>
🌐 Server URL (e.g. https://emotion.onebudd.com): https://emotion.onebudd.com

  ✔ Config saved

Commands

Process an audio file (main command)

obdt "path/to/audio.mp3"

Uploads the file, waits for processing, prints an emotion summary, and downloads the labeled dataset as a ZIP — all in one command.

By default the audio is treated as English. For other supported languages, pass --language <code>:

# English (default — no flag needed)
obdt "audio.mp3"

# Marathi
obdt "audio.mp3" --language mr

The language code controls which Deepgram transcription model is used and how much weight the transcript + LLM ranking gets vs. the acoustic model when picking the final emotion label — non-English audio leans more heavily on transcript/LLM signal since the acoustic model was trained on English speech.

  ◆ OneBudd Emotion Labeler
  ─────────────────────────────────────
  File:  interview.mp3  (14.2 MB)
  URL:   https://emotion.onebudd.com

  Uploading  ████████████████████████████████████████  100%  14.2 MB
  ✔ Uploaded  →  Job a3f91b2c…  (482.3s of audio)
  Processing  ▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓  ✔  23s elapsed
  ✔ Done

  Emotion Summary
  ─────────────────────────────────────
  neutral     ████████████████       72%  (38 chunks)
  happy       ████                   18%  (9 chunks)
  sad         ██                     10%  (5 chunks)

  ✔ All done!
  Saved to: /Users/you/emotion_results_a3f91b2c.zip

What's inside the ZIP:

emotion_results_a3f91b2c.zip
├── neutral/
│   ├── chunk_001.wav
│   └── chunk_002.wav
├── happy/
│   └── chunk_001.wav
├── emotion_log.csv       ← timestamps, transcripts, labels per chunk
├── emotion_curve.png     ← emotion timeline chart
└── metadata.csv          ← file-level summary

Options

# Save ZIP to a specific location
obdt "audio.mp3" --out "/path/to/output.zip"

# Disable transcription (ON by default)
obdt "audio.mp3" --no-transcribe

# Disable LLM semantic analysis (ON by default)
obdt "audio.mp3" --no-llm

# Use a different server
obdt "audio.mp3" --url https://your-server.com

# Override API key for this run only
obdt "audio.mp3" --key <your-api-key>

# Skip downloading the ZIP (just process)
obdt "audio.mp3" --no-download

# Audio language — defaults to "en" if omitted. Also works with `obdt submit`.
obdt "audio.mp3" --language mr

Supported --language codes:

| Code | Language | |------|----------| | en | English (default) | | mr | Marathi |

Check your quota

Each API key has a 40-minute lifetime audio quota.

obdt quota

  ◆ OneBudd API Quota
  ─────────────────────────────────────
  ████████░░░░░░░░░░░░░░░░░░░░░░  30%

  Used:       8.0 min  / 40 min total
  Remaining:  32.0 min

Check server health

obdt health

Submit without waiting (get job ID only)

obdt submit "audio.mp3"

Returns a job_id immediately. Use it later:

obdt status <job-id>
obdt download <job-id>
obdt download <job-id> --out "/path/to/output.zip"

Limits

• Max 30 minutes of audio per file
• Max 40 minutes of audio total per API key (lifetime)
• Supported formats: MP3, WAV, M4A, FLAC, OGG

Config file

Stored at ~/.obdt/config.json. You can edit it directly:

{
  "apiKey": "your-api-key-here",
  "serverUrl": "https://emotion.onebudd.com"
}

License

MIT © OneBudd