Codemaxxing Docs

Architecture mapped into focused modules

The old single giant diagram has been split into focused pages. Each topic explains the request path, runtime path, and why the system is structured this way.

Asset Pipeline + HMR

Save file → durable write → HMR event → preview module reload

Save triggers HMR, then Preview fetches assets on-demand through esbuild

Read architecture ->

Worker Execution

Preview API → bundle worker code → isolate run → response payload

API request → read & bundle worker/ → spawn V8 isolate → return response

Read architecture ->

WebSocket Feedback

Runtime logs + errors → coordinator normalization → multi-client fan-out

Logs & errors flow from Sources → Coordinator → Editor

Read architecture ->

Dependency Resolution

Import parse → metadata lookup → CDN rewrite → bundled dependency graph

Bare imports resolved via esm.sh CDN at bundle time

Read architecture ->

Git Integration

Git UI action → Host route → DO mount → isomorphic-git → status broadcast

isomorphic-git running inside Durable Object with SQLite storage

Read architecture ->

AI Agent System

Chat request → guarded dispatch → tool/model loop → streamed project updates

Autonomous coding agent with 24 tools, running in a dedicated Durable Object

Read architecture ->

Test Runner

Run request → bundle harness + tests → isolate execution → synchronized results

In-browser test execution with a built-in describe/it/expect harness

Read architecture ->

Deploy Pipeline

Deploy trigger → file collection → bundle + upload → script publish

One-click production deployment to the edge

Read architecture ->

Snapshots + Revert

Pre-mutation capture → durable snapshot → revert command → reload broadcast

Point-in-time backups of files before AI agent modifications

Read architecture ->

Real-time Collaboration

Participant events → project coordinator → peer updates

Multi-user editing with cursor sharing and file edit broadcasting

Read architecture ->