This is the canonical reference for all config keys and defaults. For the getting-started walkthrough, see Configuration. File location
Override the config directory:
export NOTEWISE_HOME=/custom/path
Load order
| Priority | Source | Notes |
|---|
| 1 (lowest) | Code defaults | Built-in values in _constants.py |
| 2 | ~/.notewise/config.env | Written by notewise setup |
| 3 | Environment variables | Always override config file |
| 4 (highest) | CLI flags | Per-run only (e.g. --model, --output) |
Settings reference
DEFAULT_MODEL
Type: string · Default: gemini/gemini-2.5-flash
The LiteLLM-format model string used when --model is not passed on the CLI.
DEFAULT_MODEL=gpt-4o
DEFAULT_MODEL=anthropic/claude-3-5-sonnet-20241022
DEFAULT_MODEL=groq/llama3-70b-8192
API keys
Set the key matching the provider of your chosen model. Only the key(s) you actually use need to be set.
| Key | Provider |
|---|
GEMINI_API_KEY | Google Gemini / Vertex AI |
OPENAI_API_KEY | OpenAI |
ANTHROPIC_API_KEY | Anthropic |
GROQ_API_KEY | Groq |
XAI_API_KEY | xAI (Grok) |
MISTRAL_API_KEY | Mistral AI |
COHERE_API_KEY | Cohere |
DEEPSEEK_API_KEY | DeepSeek |
TEMPERATURE
Type: float, range 0.0–1.0 · Default: 0.7
LLM sampling temperature. Lower values produce more deterministic output; higher values produce more varied output.
TEMPERATURE=0.3 # focused, consistent
TEMPERATURE=0.7 # balanced (default)
TEMPERATURE=0.9 # creative
MAX_TOKENS
Type: integer > 0 · Default: (unset — uses the model’s own default)
Maximum tokens per LLM response. Set this to control costs or fit within a model’s output token limit.
OUTPUT_DIR
Type: path · Default: ./output
Directory where study notes are written. Relative paths resolve from the current working directory.
OUTPUT_DIR=~/study-notes
OUTPUT_DIR=/data/notewise-output
MAX_CONCURRENT_VIDEOS
Type: integer > 0 · Default: 5
Maximum videos processed in parallel during a batch or playlist run.
MAX_CONCURRENT_VIDEOS=3 # gentler on APIs
MAX_CONCURRENT_VIDEOS=10 # faster for large batches
YOUTUBE_REQUESTS_PER_MINUTE
Type: integer > 0 · Default: 10
Rate limit for YouTube HTTP requests shared across all concurrent workers. Reduce if you encounter rate-limit errors.
YOUTUBE_REQUESTS_PER_MINUTE=5
YOUTUBE_COOKIE_FILE
Type: path · Default: (unset)
Path to a Netscape-format .txt cookies file for private, age-gated, or members-only videos.
YOUTUBE_COOKIE_FILE=/home/user/youtube-cookies.txt
Code-only defaults
These constants are not exposed in config.env. Change them by editing src/notewise/_constants.py.
| Constant | Default | Description |
|---|
DEFAULT_CHUNK_SIZE | 4000 tokens | Max tokens per transcript chunk before splitting |
DEFAULT_CHUNK_OVERLAP | 200 tokens | Token overlap between consecutive chunks |
DEFAULT_CHAPTER_MIN_DURATION | 3600 seconds | Min duration to activate chapter-level generation |
DEFAULT_MAX_CONCURRENT_CHAPTERS | 3 | Max parallel chapter generation tasks |
DEFAULT_LANGUAGES | ["en"] | Default transcript language preference list |
TRANSCRIPT_MAX_RETRIES | 3 | Retry attempts for transcript fetch |
LLM_NUM_RETRIES | 3 | Retry attempts for LLM API calls |
HTTP_MAX_RETRIES | 3 | Retry attempts for YouTube HTTP requests |
MAX_FILENAME_LENGTH | 100 | Max characters in generated output filenames |
CLI overrides
CLI flags take the highest priority — they override both config.env and environment variables for the duration of that single run.
| Config key | CLI flag |
|---|
DEFAULT_MODEL | --model / -m |
OUTPUT_DIR | --output / -o |
TEMPERATURE | --temperature / -t |
MAX_TOKENS | --max-tokens / -k |
YOUTUBE_COOKIE_FILE | --cookie-file / --cookies |
| (transcript languages) | --language / -l |
Example config.env
# ~/.notewise/config.env
DEFAULT_MODEL=gemini/gemini-2.5-flash
OUTPUT_DIR=~/study-notes
MAX_CONCURRENT_VIDEOS=3
TEMPERATURE=0.7
# MAX_TOKENS=2000
GEMINI_API_KEY=your_key_here
# OPENAI_API_KEY=sk-...
# ANTHROPIC_API_KEY=sk-ant-...
Deprecated keys
These keys are silently ignored if present in config.env — they existed in earlier versions and were removed:
YOUTUBE_USE_OAUTHYOUTUBE_SAVE_OAUTH_TOKENYOUTUBE_OAUTH_TOKEN_FILEYOUTUBE_AUTO_REFRESH_OAUTH_TOKEN
Deprecated keys are ignored silently. Remove them from old config files to avoid confusion.