Roadmap

[crossps]

Tracking planned work for LLM Bridge & Cache. Ideas, feedback, and PRs welcome — see CONTRIBUTING.md.

Near term

• Native provider adapters beyond OpenAI/Anthropic (Gemini, Mistral, OpenRouter) — keep them dependency-free.
• Per-model routing overrides and provider fallback (retry on a second provider when one errors).
• Optional on-disk usage/cost tracking with a /metrics view.

Later

• Pluggable injection presets (named profiles you can switch per request via a header).
• Configurable cache-breakpoint strategy (how many breakpoints, where).
• A tiny status web page at / for at-a-glance cache-refresh activity.

Non-goals

• Runtime dependencies — this stays standard-library only.
• Bundling or storing any credentials — bring your own key, always.