Vidai Community is now available: one Rust binary for cost attribution, guardrails and multi-provider routing on every LLM call
Cost control on LLM traffic isn't an unsolved problem. Most teams I see have already wired up some version of it — a library for SDK translation, a custom OpenTelemetry exporter, Python middleware for spend caps, regex guards on prompts, a JSON of provider prices someone refreshes on Tuesdays. It works. Nobody chose that stack though. It grew.
Vidai Community is now available. One Rust binary doing those seven jobs in one pass, inside your own infrastructure.
- One-line integration. Change the
base_urlon your existing OpenAI / Anthropic / Google GenAI SDK. Keep your code, your tool-call format, your streaming semantics. Bidirectional native translation handles the rest — call with the Anthropic SDK, route to OpenAI, response comes back in Anthropic shape. - In-path cost attribution at return time. Per user, key, team, app and model. Hard per-team budgets that stop the next call when a cap breaks, not alerts about the call that already ran.
- Free public rate-card service. Pricing data for every major provider, refreshed on a standard cadence. Stop maintaining your own JSON.
- 25 MB. 1.95 ms median overhead. 21,803 RPS verified on a single node.
Free, no expiry, self-serve at vidai.uk/community · docs
If you spin it up and find the rough edges, hate one of the defaults, or wish a flag did something different: open an issue on the quickstart repo or reply here. The product gets sharper every time someone runs it on hardware that is not ours, and this is the kind of feedback we cannot get from inside our own stack.