Skip to content

Configuration

Everything is an environment variable prefixed SCRIBE_, read once at start by scribe/config.py. Put them in your shell profile or a file you source before scribe serve.

Speech to text

Variable Default Notes
SCRIBE_STT_BACKEND whisper whisper or vosk
SCRIBE_WHISPER_MODEL small.en tiny.en base.en small.en medium.en large-v3, or a path to a CTranslate2 model folder
SCRIBE_STT_DEVICE auto auto picks cuda if available, else cpu
SCRIBE_STT_COMPUTE auto float16 on GPU, int8 on CPU. int8_float16 saves VRAM
SCRIBE_STT_LANGUAGE en Empty string for auto-detect (multilingual models only)
SCRIBE_STT_BEAM 5 Beam size; 1 is faster and slightly worse
SCRIBE_VOSK_MODEL models/vosk-model-small-en-us-0.15 Folder of an extracted Vosk model

The scribe also sets ORT_DISABLE_TELEMETRY=1 for you. That stops the onnxruntime library, which Whisper uses to skip silence, from reporting usage to Microsoft. Set it to 0 before starting if you want the library's own default back. All open source, no cloud has the measurement.

Language model

Variable Default Notes
SCRIBE_LLM_BASE_URL http://localhost:11434/v1 Ollama. vLLM is usually :8000/v1, llama.cpp :8080/v1, LM Studio :1234/v1
SCRIBE_LLM_MODEL empty Empty uses the first model the server lists
SCRIBE_LLM_API_KEY not-needed Local servers ignore it. A cloud service needs a real one; see Cloud models and PHI
SCRIBE_LLM_TEMPERATURE 0.2 Keep low
SCRIBE_LLM_MAX_TOKENS 2048 Enough for most notes. A note that hits the cap is flagged as cut off
SCRIBE_LLM_TIMEOUT 300 Seconds. CPU inference can be slow
SCRIBE_LLM_THINKING off on lets reasoning models think; slower, rarely better for this task
SCRIBE_ALLOW_REMOTE_LLM unset Set to 1 to permit a non-private address. Read Lesson 7 and Cloud models and PHI first

Storage

Variable Default Notes
SCRIBE_DATA_DIR scribe_data Relative to the working directory
SCRIBE_RETENTION_DAYS 7 Unsigned sessions older than this are removed when the review app starts, hourly while it runs, and by scribe purge
SCRIBE_DELETE_AUDIO_ON_SIGN 1 Set 0 to keep audio after signing

Audio and server

Variable Default Notes
SCRIBE_MIC system default Device index or name from scribe mics
SCRIBE_HOST 127.0.0.1 Do not change without reading Lesson 7
SCRIBE_PORT 8765
SCRIBE_ALLOWED_HOSTS empty Extra names the review app answers to, comma separated. It always answers to localhost, 127.0.0.1 and SCRIBE_HOST. Only needed off localhost; Lesson 7

Example profile

# ~/.config/scribe.env  ->  source ~/.config/scribe.env
export SCRIBE_WHISPER_MODEL=medium.en
export SCRIBE_LLM_BASE_URL=http://localhost:11434/v1
export SCRIBE_LLM_MODEL=qwen3:8b
export SCRIBE_DATA_DIR=$HOME/scribe_data
export SCRIBE_MIC="USB PnP Audio Device"