Fresh-AI

AI-powered autocomplete plugin for the Fresh editor. Provides inline ghost-text suggestions with multi-provider support, RAG-based workspace context, and intelligent multi-line completion.

Features

• Multi-provider support -- OpenAI-compatible, Anthropic, and Ollama
• RAG workspace indexing -- semantic code search for context-aware completions
• Ghost text suggestions -- inline preview with partial acceptance (word/line)
• Intelligent indentation -- re-indents multi-line suggestions to match context
• Persistent cache -- embedding cache survives editor restarts

Quick Start

1. Clone into your Fresh plugins directory:

   git clone <repo-url> ~/.config/fresh/plugins/ai-autocomplete

2. Set your API key(s):

   export MISTRAL_API_KEY="your-key"     # For Codestral (default autocomplete)
   export OPENAI_API_KEY="your-key"       # For embeddings (default)

3. Restart Fresh -- the plugin auto-initializes.

Documentation

Topic	Description
Getting Started	Installation, first config, verifying it works
Configuration	All settings with explanations and examples
Commands	All commands with keybinding info
Providers
OpenAI-Compatible	Codestral, Mistral, OpenAI setup
Anthropic	Claude setup
Ollama	Local model setup
Features
RAG Indexing	How workspace indexing works
Partial Acceptance	Word/line acceptance
Caching	Persistence and cache invalidation
Reference
Troubleshooting	Common issues, logging, debugging
Architecture	Data flow, state machine, components
Contributing	Dev setup, testing, adding providers

Development

npm install
npm test              # Run tests (vitest)
npm run lint          # ESLint
npm run typecheck     # TypeScript type checking