Skip to content
All docs

AI features

BluePhoenix can use an AI model to save you typing. Everything is optional, and you choose where the model runs.

What it can do

Feature Where
Generate a commit message from your changes Git tab → Commit → Generate
Prioritize to-dos To-dos tab → Prioritize with AI
Write a prompt for an AI coding assistant from a to-do To-do details → Generate
Generate a changelog, summarize changes, write a PR description Changelog tab
Scan a repository for commands to import Add project dialog, or Commands tab → Scan with AI

AI answers can be wrong. Always glance at a commit message or command before using it.

Set it up

Open Settings → AI.

  1. Choose Local inference (runs on your computer) or Cloud provider.
  2. Pick a provider.
  3. For cloud providers, paste your API key and press Save. The key goes into your system keychain (macOS Keychain, Windows Credential Manager, Linux Secret Service), never into BluePhoenix’s own data.
  4. Press Load models and pick a model, or type its name.
  5. Press Test. A short reply means you are set.

Local options (nothing leaves your computer)

Ollama (easiest)

  1. Install it from ollama.com.
  2. In a terminal run ollama pull llama3.2 (or any model you like).
  3. In BluePhoenix choose Local → Ollama, press Load models.

LM Studio

  1. Install LM Studio, download a model, and open the Developer tab.
  2. Start the local server (default http://localhost:1234).
  3. Choose Local → LM Studio, press Load models.

llama.cpp with a GGUF file

  1. Install llama.cpp so the llama-server command works (on Mac: brew install llama.cpp).
  2. Choose Local → llama.cpp (GGUF file) and press Choose & start, then pick your .gguf file.
  3. Wait a few seconds for the model to load, then press Test.

Other local server: any program that speaks the OpenAI-compatible API (vLLM, LocalAI, text-generation-webui). Choose Other local server and enter its address.

Cloud providers

OpenAI, Anthropic, OpenRouter, Google Gemini, Groq, Mistral, DeepSeek, xAI, Together AI, or Other OpenAI-compatible with your own address. Create an API key on the provider’s website and paste it in. Your requests go to that provider and are covered by their terms and pricing.

Privacy

With a local provider, prompts stay on your computer. With a cloud provider, the relevant text (a diff, to-do titles, commit messages, file names and a few key project files) is sent to that provider when you use a feature. Nothing is sent until you press a button.

PreviousCloud sync and the iPhone appNextCommand palette