Skip to main content

​
Config file location

NoteWise reads its configuration from: 
~/.notewise/config.env
This file is created by notewise setup. It uses a simple KEY=VALUE format, one setting per line. Lines beginning with # are treated as comments.

​
Overriding the state directory

All persistent state (config, cache, logs) lives under ~/.notewise/ by default. To use a different directory, set the NOTEWISE_HOME environment variable before running any notewise command: 
export NOTEWISE_HOME=/path/to/custom/dir
NoteWise will read $NOTEWISE_HOME/config.env and write all state files there instead.

​
Managing your config

NoteWise provides three commands to work with the config file: 
# Run the interactive setup wizard to create or rewrite config
notewise setup

# Display the current resolved configuration (secrets masked)
notewise config

# Open the config file in your $EDITOR (or the OS default editor)
notewise edit-config
CLI flags always take precedence over values in the config file. For example, passing --model gpt-4o on the command line overrides whatever DEFAULT_MODEL is set to in config.env.

​
Load order

Settings are resolved in this order, with later sources overriding earlier ones:
  1. Code defaults (the values documented below)
  2. ~/.notewise/config.env (or $NOTEWISE_HOME/config.env)
  3. Environment variables

​
All configuration keys

​
LLM settings

​
DEFAULT_MODEL
string
default:"gemini/gemini-2.5-flash"
The LiteLLM-format model string used when --model is not passed on the command line. Any model supported by LiteLLM can be used here. Examples: gpt-4o, claude-3-5-sonnet-20241022, groq/llama3-70b-8192.
​
TEMPERATURE
number
default:"0.7"
LLM sampling temperature. Accepts values between 0.0 (deterministic) and 1.0 (more creative). Lower values produce more consistent, focused notes; higher values may introduce more varied phrasing.
​
MAX_TOKENS
number
Maximum number of tokens per LLM response. When unset, the model’s own default limit applies. Increase this value if notes are being truncated for very long video segments.

​
API keys

Set the key that corresponds to your chosen model’s provider. Only the key for the provider you are using needs to be present.
​
GEMINI_API_KEY
string
API key for Google Gemini (and Vertex AI) models. Required when DEFAULT_MODEL starts with gemini/ or vertex/. Get a free key at aistudio.google.com.
​
OPENAI_API_KEY
string
API key for OpenAI models (e.g. gpt-4o, o3-mini).
​
ANTHROPIC_API_KEY
string
API key for Anthropic Claude models (e.g. claude-3-5-sonnet-20241022).
​
GROQ_API_KEY
string
API key for Groq-hosted models (e.g. groq/llama3-70b-8192).
​
XAI_API_KEY
string
API key for xAI Grok models (e.g. xai/grok-2).
​
MISTRAL_API_KEY
string
API key for Mistral models (e.g. mistral/mistral-large-latest).
​
COHERE_API_KEY
string
API key for Cohere models (e.g. command-r-plus).
​
DEEPSEEK_API_KEY
string
API key for DeepSeek models (e.g. deepseek/deepseek-chat).

​
Output

​
OUTPUT_DIR
string
default:"./output"
Directory where generated Markdown files are written. Relative paths are resolved from the current working directory at the time you run notewise process. Can be overridden per-run with --output / -o.

​
Concurrency

​
MAX_CONCURRENT_VIDEOS
number
default:"5"
Maximum number of videos processed in parallel when handling a playlist or batch file. Reduce this value if you hit rate limits from your LLM provider or YouTube.
​
YOUTUBE_REQUESTS_PER_MINUTE
number
default:"10"
Rate limit applied to YouTube transcript and metadata requests. Reduce this value if you encounter 429 Too Many Requests errors from YouTube.

​
Transcript and language

​
YOUTUBE_COOKIE_FILE
string
Path to a Netscape-format cookies .txt file. Required to access age-gated or login-required videos. Can be overridden per-run with --cookie-file.
The default transcript language (en) is a code default and is not configurable via config.env. Override it per-run with --language / -l.

​
Chunking and chapter generation (code defaults only)

These settings are fixed code defaults and are not read from config.env. They apply automatically on every run.
SettingDefaultDescription
DEFAULT_CHUNK_SIZE4000 tokensMaximum tokens per transcript chunk sent to the LLM
DEFAULT_CHUNK_OVERLAP200 tokensToken overlap between consecutive chunks
DEFAULT_CHAPTER_MIN_DURATION3600 secondsMinimum duration to trigger per-chapter notes (1 hour)
MAX_CONCURRENT_CHAPTERS3Maximum chapters processed in parallel

​
Example config file

A minimal ~/.notewise/config.env for a user using the default Gemini model: 
GEMINI_API_KEY=your_gemini_api_key_here
DEFAULT_MODEL=gemini/gemini-2.5-flash
OUTPUT_DIR=./output
TEMPERATURE=0.7
MAX_CONCURRENT_VIDEOS=5
YOUTUBE_REQUESTS_PER_MINUTE=10
To switch to OpenAI GPT-4o instead: 
OPENAI_API_KEY=your_openai_api_key_here
DEFAULT_MODEL=gpt-4o
OUTPUT_DIR=./output
TEMPERATURE=0.7
Never commit your config.env file to version control. It contains your API keys. The file lives in your home directory (~/.notewise/) precisely to keep it out of project repositories.

Build docs developers (and LLMs) love

Get started for freeTalk to us