Use multiple LLM backends in a single crate, simple builder-based configuration, and built-in prompt chaining & templating.
Terminal score: 0–1000 raw, weighted across 4 dimensions. Public score: 0–10 normalized (shown in the 30-day stars chart above).
