Skip to content

AI_GUIDE.md

Latest commit

 

History

History
141 lines (96 loc) · 2.89 KB

AI_GUIDE.md

File metadata and controls

141 lines (96 loc) · 2.89 KB
Raw

Zypheron AI Guide

This guide covers the current AI engine, provider setup, and model workflow in Zypheron CLI.

Overview

Zypheron's AI features are powered by the Python-side AI engine used by the Go CLI and TUI.

Current AI-enabled flows include:

  • chat in the CLI
  • AI-guided dork enhancement
  • model/provider selection in the TUI
  • AI-assisted analysis and workflow support
  • MCP integration for external AI clients

AI Engine Basics

Useful commands:

zypheron ai start
zypheron ai stop
zypheron ai status
zypheron ai providers
zypheron ai doctor
zypheron ai test --provider ollama

The CLI can auto-start the AI engine when needed for some flows.

Installing AI Dependencies

If you used the source bootstrap, this is usually handled for you. If not:

zypheron install-deps --all

If you need to work directly inside zypheron-ai/, you can still create a virtual environment manually, but the CLI installer is now the preferred path.

API Keys and Provider Setup

Store hosted-provider keys with the CLI:

zypheron config set-key anthropic
zypheron config set-key openai
zypheron config set-key deepseek
zypheron config set-key gemini
zypheron config set-key kimi

See configured providers:

zypheron config get-providers

Environment variables are also supported for many providers, but the CLI key flow is the preferred user path.

TUI Model Selection

The TUI supports switching between local and hosted models.

Current behavior:

  • selecting a hosted model without a configured API key prompts for the key immediately
  • the selected provider/model is persisted
  • the selection remains active until changed again in settings
  • local models such as Ollama do not require hosted API keys

Supported Provider Types

The current project supports a mix of hosted and local providers.

Common provider groups:

  • Anthropic
  • OpenAI
  • Google Gemini
  • DeepSeek
  • Moonshot Kimi
  • Ollama

Exact available models depend on the current model registry and your build.

Ollama Setup

For local models:

ollama pull llama3.2:3b
ollama pull mistral:latest
zypheron config set ai.provider ollama
zypheron config set ai.model llama3.2:3b

Then verify:

zypheron ai test --provider ollama

Common AI Workflows

zypheron chat "Summarize this target profile"
zypheron dork "admin login panels" --ai-guided
zypheron ai status

In the TUI:

zypheron

Then select your model/provider from the model selector.

Troubleshooting

If AI is not working:

zypheron ai doctor
zypheron ai status
zypheron config get-providers
zypheron doctor

Common causes:

  • missing provider API key
  • local model runtime not running
  • Python dependencies not installed
  • system keyring unavailable, requiring environment variables instead

MCP

If you want to expose Zypheron tooling to an external AI client, see MCP_INTEGRATION.md.