Environment Configuration

Configuration and Authentication

Moltext provides a flexible configuration system to accommodate different workflows, ranging from cloud-based LLM processing to local-first raw extraction. You can configure the tool using CLI flags or environment variables.

API Key Management

If you are not using --raw mode, Moltext requires access to an LLM to normalize and compress documentation. By default, it is configured for OpenAI.

You can provide your API key in two ways:

1. Environment Variable (Recommended): Define OPENAI_API_KEY in your shell or a .env file.
2. CLI Flag: Pass the key directly using the -k or --key option.

# Using the CLI flag
moltext https://docs.example.com -k sk-proj-xxx

Environment Variables (.env)

Moltext automatically loads configuration from a .env file located in your current working directory. This is the preferred method for persistent configurations.

Create a .env file with the following variables:

# Required for OpenAI processing
OPENAI_API_KEY=sk-your-api-key-here

# Optional: Override the default model (defaults to gpt-4o-mini)
# MODEL_NAME=gpt-4

# Optional: Override the default OpenAI base URL
# BASE_URL=https://api.openai.com/v1

Local Inference Configuration

To use Moltext with local LLM providers (such as Ollama, LM Studio, or LocalAI), you must override the base-url. When a custom base-url is detected that does not point to OpenAI, the API key requirement is relaxed (a placeholder key is used internally).

# Example: Using Ollama
moltext https://docs.example.com \
  --base-url http://localhost:11434/v1 \
  --model llama3

Configuration Hierarchy

Moltext resolves configuration settings in the following order of precedence:

1. Explicit CLI Flags: (e.g., --key, --model, --base-url)
2. Environment Variables: (e.g., OPENAI_API_KEY)
3. Application Defaults:
   • Base URL: https://api.openai.com/v1
   • Model: gpt-4o-mini
   • Output Path: context.md
   • Page Limit: 100

Running without Configuration (Raw Mode)

If you prefer to skip LLM processing entirely and obtain structured Markdown without an API key or local model, use the --raw flag. This bypasses all authentication logic and returns the clean, normalized content directly.

moltext https://docs.example.com --raw

Execution Limits

To prevent infinite loops or excessive API usage when parsing large documentation sites, use the --limit flag to cap the number of pages crawled.

# Limit the crawl to 50 pages
moltext https://docs.example.com --limit 50