Skip to content

Configuration

Module: src/sbobinator/config.py
Summary API keys: src/sbobinator/summary_config.py

ASR constants

Name Value Description
DEFAULT_MODEL nvidia/parakeet-tdt-0.6b-v3 ASR model
SAMPLE_RATE 16000 Hz for NeMo
CHUNK_THRESHOLD_SEC 1800 (30 min) Chunking threshold

Local LLM (Qwen)

Name Value
LOCAL_LLM_FOLDER qwen2.5-3b-instruct
LOCAL_LLM_GGUF_FILE qwen2.5-3b-instruct-q4_k_m.gguf
MIN_RAM_GB 16

Functions: local_gguf_path(), local_llm_available(), system_ram_gb().

Enums

SummaryLength

auto, short, normal, detailed

SummaryMode (legacy DB)

extractive, abstractive — deprecated, not used by the LLM summary engine.

Path functions

Function Description
data_dir() SBOBINATOR_DATA or ./data
models_dir() NEMO_CACHE_DIR or ./models
local_model_path() Parakeet .nemo path

Python dependencies

Source of truth: pyproject.toml → see requirements/README.md.

Environment variables

See Environment and .env.example.