AI Context Visualizer
A full-stack analytics tool that parses Claude Code JSONL session files and renders real-time, interactive visualizations of context window consumption. Features GPU-style animated bubble charts across five view modes (Source, Cache, Activity, Thread, Model), autocompaction tracking, and detailed token breakdowns by category.
The Problem
AI coding assistants consume context windows opaquely — developers have no visibility into how tokens are allocated across tool calls, file reads, user input, and model output, making it difficult to optimize long sessions.
The Solution
Built a Python/FastAPI backend that ingests and categorizes session tokens using deterministic rules, paired with a React/Canvas 2D frontend that renders interactive bubble charts, timeline views, and category breakdowns in real time.
Screenshots
