Repository navigation
CharacterBench 0.2.2-alpha
Pre-release
Pre-release
CharacterBench 0.2.2-alpha fixes local model access on Macs whose system proxy intercepts localhost requests. Ollama model discovery, diagnostics, and evaluation now connect directly to local endpoints.
Download the ZIP, unzip it, then run:
python3 characterbench.py doctor
python3 characterbench.py uiThe offline demo works without a model. Live character tests require a running local model such as Ollama, or a configured OpenAI-compatible endpoint.
The package passed its built-in self-tests and privacy scan. Scores remain diagnostic signals for character consistency.