Fast Summarization CLI

Problem

meta-v2 targets a narrow workflow: you have a web page, a local file, or a YouTube link and need a readable summary without manually opening, copying, and pasting the content into a model. It bundles content retrieval, model selection, and output length control into a CLI, which fits scripts, note pipelines, or batch research curation.

How It Works

The core path is to use an external LLM for summarization:
- Model setup: set an environment variable such as OPENAI_API_KEY, ANTHROPIC_API_KEY, XAI_API_KEY, or GEMINI_API_KEY; if none is set, the default is google/gemini-3-flash-preview.
- Inputs: supports URLs, local files, and YouTube links; --extract-only limits processing to URL extraction.
- Output control: use --length for summary length, --max-output-tokens to cap output, and --json for machine-readable results.
- Fallbacks: --firecrawl can enable Firecrawl when page extraction is blocked, and --youtube can use Apify when APIFY_API_TOKEN is set.
An optional config file is ~/.summarize/config.json.

Boundaries

It depends on LLM API keys and is not a fully offline tool. YouTube and protected sites may require extra service tokens. The default model includes preview, so production use should track model availability.