# Guanlan

Guanlan is an AI-agent-first search and source-routing tool for the Chinese internet. It focuses on public search, webpage reading, Chinese hotnews, source classification, safe diagnostics, and explicit authentication boundaries.

## Documentation

- `README.md`: Canonical product documentation in Chinese
- `docs/agent-usage.md`: Agent-facing commands, routing rules, and fallback behavior
- `docs/local-llm.md`: How local models use Guanlan evidence
- `docs/troubleshooting.md`: Keychain, network, cookie, and platform diagnostics
- `docs/SOURCE_ATTRIBUTION.md`: Open-source references

## Core Commands

- `guanlan welcome`
- `guanlan capabilities`
- `guanlan search "keyword" --profile china --limit 50`
- `guanlan search "keyword" --profile china --scope party_central`
- `guanlan search "keyword" --profile china --scope ecommerce`
- `guanlan read "https://example.com/article"`
- `guanlan read "https://example.com/article" --backend direct`
- `guanlan research "keyword" --profile china --advisor`
- `guanlan hotnews today --limit 50`
- `guanlan doctor --trace`

## Safety

Guanlan defaults to public, read-only, low-disturbance workflows. Cookie, browser, Keychain, and login-state access should only happen after explicit user authorization.
