AI Prompt Cost Audit Tool

OmniNode builds delegation and routing infrastructure for AI agents. We help teams send each task to the right-sized model instead of paying frontier prices for everything. We need a public, free Token Audit Tool on omninode.ai that lets a developer paste or upload their LLM usage logs and see exactly where their money goes and what they could recover. We already have a working single-file HTML prototype with five views: a model price catalog, a prompt cost explorer, a delegation modeler, a usage-log audit import, and a calibration panel. Your job is to turn it into a production tool: clean TypeScript, real pricing data, a robust import engine, a full audit trail per row, exportable reports, tests, and a deploy on our site. What you will build: 1. A typed calculation engine, separated from the UI and covered by unit tests 2. Live model pricing pulled from the public LiteLLM pricing registry, versioned and refreshed daily 3. An import engine for CSV and JSON exports from OpenAI, Anthropic, LiteLLM, OpenRouter and OpenTelemetry GenAI traces, with a column-mapping step when headers do not match 4. A per-row audit trail that shows measured versus estimated values, the pricing version used, and the savings math 5. CSV, JSON and PDF export of the audit report 6. A responsive, accessible UI in our brand, embedded at a route on omninode.ai 7. Product analytics events (PostHog) and a short README and methodology page Hard requirements: • All log processing happens in the browser. User data never leaves their machine. • Every number shown is labeled measured or estimated. • Code is delivered to our GitHub org and released under the MIT license. You will get: the prototype source, our audit standard, brand assets, sample log files, and a 30 minute kickoff call. Skills: React, TypeScript, Next.js or Vite, Data Visualization, CSV/JSON parsing, Vitest or Jest, PostHog, LLM APIs, Tailwind CSS