v1.3.1 — Use your own AI server (Ollama, LM Studio, vLLM…)

Run ViralMint on your own AI server, plus three fixes. Full detail in CHANGELOG.md.

✨ New

• Use any OpenAI-compatible AI server (#43). Settings → AI Provider → OpenAI-compatible: enter the server's base URL and model name; the API key is optional. Works with Ollama, LM Studio, vLLM, llama.cpp, DeepSeek, Groq or a company gateway.
  • Ollama example: base URL http://localhost:11434/v1, model exactly as ollama list shows it (e.g. qwen3:14b).
  • From .env: CUSTOM_AI_BASE_URL, CUSTOM_AI_MODEL, and CUSTOM_AI_API_KEY if your server needs one.
  • Small local models (1–3B) often can't produce the structured replies some features need; those features then fall back to simpler behaviour. 7B+ is a better starting point.

🐛 Fixed

• Linux tray launcher no longer crashes on start — thanks @chiragjangiir (#46).
• Long jobs survive development mode — DEBUG=true no longer restarts the server mid-download/render, and the Windows "Cannot launch npm" Motion Graphics install error is gone (#45).
• A cancelled Motion Graphics upgrade no longer looks uninstalled (#45).
• The chat assistant counts an OpenRouter key in .env as a configured provider.

📦 Install

Download the installer for your platform below, or run from source (python run.py). The macOS build is Apple Silicon only; Intel Macs run from source for now (#44). Bring your own key — Anthropic, OpenAI or OpenRouter — or point it at your own server. No telemetry; keys go straight from your machine to the provider.

2,999 tests run on every commit.