What it does ✨
- Answers direct prompts or performs tasks using one or more local files as context
- Sends requests to an OpenAI-compatible model endpoint, including optional local model servers
- Generates structured multi-file patches with diffs, optional review, and configurable output suffixes
- Processes files recursively in tree mode with include/exclude filters, concurrent requests, retries, and resumable sessions
- Supports plan mode plus text, JSON, JSON Schema Structured Outputs, and streamed text responses
- Supports configuration profiles, reusable prompt-library entries, custom system prompts, reasoning controls, and model parameters
- Provides optional local session logging, API storage controls, usage statistics, token estimates, and cost estimates
- Creates a context-controlled, continuously verified workflow that maximizes cost effectiveness
Requirements 📦
- python 3.10 or newer and the pip package manager
- make from GNU Make or compatible for a managed installation
- Network access to your chosen OpenAI-compatible endpoint
- An API key for the endpoint
- Optional: a local OpenAI‑compatible server
Install / Update 🛠️
git clone https://github.com/Southland-Systems/mince.git
cd mince
make installUpdate
cd mince
make updateUninstall
cd mince
make uninstall-userManual install
(cd mince && cp -a mince ~/.local/bin/ && chmod +x ~/.local/bin/mince \
&& pip install -U -r requirements.txt)First run 🚀
mince --initThis creates ~/.local/state/mince/config.json.
Basic Usage 💡
Ask a direct question without file context: