What problem does it solve? Teams shipping LLM-powered features often have no visibility into what each request costs or which callsites generate spend. This Skill integrates a repository with the Caveman gateway so every LLM request is measured and priced, without changing application behavior. ## Core Features & Use Cases - Callsite Discovery: Scans dependency files and source code for LLM SDKs (OpenAI, Anthropic, Vercel AI SDK, LangChain, LiteLLM) and raw provider HTTP calls before making any change. - Gateway Wiring: Redirects each callsite's base URL through the gateway path /w/<app> with the x-cave-api-key header, supporting both stored provider keys and bring-your-own-key (byok) modes. - Verified Integration: Sends one real minimal verification request and reports the actual HTTP status and token usage, with verbatim failure templates for unreachable gateways, invalid keys, or routing errors. - Use Case: A developer with a Python support bot using the OpenAI SDK runs this Skill to route all calls through the gateway, then sees per-request costs grouped under the app's slug on the Caveman dashboard. ## Quick Start Ask the agent to set up Caveman in this repository using the provided gateway URL, API key, provider key mode, and dashboard URL.
