dsh-gateway-billing

Self-hosted LLM gateways often route a single conversation through multiple models, yet figuring out the actual cost used to mean digging through logs or writing throwaway scripts. dsh-gateway-billing brings that number right under the DSH composer: configure one pricingUrl in your profile and you get a live, per-conversation breakdown of total cost, the models and call counts involved, and the input / output / cache token split. Multi-model conversations expand into separate lines automatically.

The plugin fetches a New API / One API-format price list from the gateway you configure, caches it server-side with a 10-minute refresh cycle, and serves snapshots to the dock via RPC so no conversation triggers a redundant fetch. Both token-based pricing (quota_type=0) and flat-rate pricing (quota_type=1) are supported. Model entries missing from the price list are called out with a visible counter instead of being silently zeroed out. Unconfigured and fetch-failed are treated as two distinct states, each with its own clear message, so nothing is swallowed.

Ideal for developers or teams running a self-hosted LLM gateway that exposes a New API / One API-compatible /pricing endpoint and want cost visibility without coupling to any specific vendor. The plugin hard-codes no gateway address, leaves the pricing source entirely in your hands, and ships under the MIT license.