Dev Console

This is a personal project for experimenting with alternative UIs designed for AI-native development of cloud-native applications from mobile devices. While existing tools are usable on some mobile devices, they are not ideal for this workflow.

An AI-chat-first development environment for software engineers who want to code from mobile devices or thin web clients while the heavy lifting runs on self-managed Linux infrastructure.

Features

• AI Agent Chat — copilot-style chat sessions that can read and write files in the workspace; proposed changes are staged for review before being written to disk
• Change Review — side-by-side / unified diff view with per-file Accept / Reject controls
• Terminal — full PTY session scoped to the workspace, accessible from the browser
• File Browser — lazy-loaded file tree with syntax-highlighted viewer and minimal in-browser editor
• GitHub OAuth — authentication via GitHub OAuth; access restricted to an operator-configured allowlist of GitHub users
• Mobile-friendly — React PWA with Tailwind CSS; works on 375 px viewports

Architecture

Client (React SPA / PWA)
       │  HTTPS + WebSocket
       ▼
Dev Console Server  (single Go binary)
  ├── Auth (GitHub OAuth, JWT sessions)
  ├── Workspace Manager
  ├── Agent Chat Manager  ──►  LLM API (OpenAI / Anthropic / Ollama)
  ├── Terminal Manager    ──►  PTY / shell (creack/pty)
  └── File I/O (fs, git)

The server embeds the compiled SPA via go:embed so no separate static-file server is needed.

Quick Start

Note: The project is under active development. The build tooling and example config described below are planned as part of Phase 1 (see docs/PLAN.md) and are not yet available in the repository.

Once Phase 1 is complete, the intended workflow will be:

1. Create a GitHub OAuth App and note the client ID / secret.

2. Create a config file based on the Configuration section below and fill in your values.

3. Build and run:

   make build
   ./dev-console --config dev-console.yaml

4. Open http://localhost:8080 in a browser, sign in with GitHub, and pick a workspace.

Configuration

# dev-console.yaml
server:
  listenAddr: ":8080"
  tls:
    certFile: "/etc/dev-console/tls.crt"
    keyFile:  "/etc/dev-console/tls.key"

auth:
  github:
    clientId:     "YOUR_GITHUB_CLIENT_ID"
    clientSecret: "${GITHUB_CLIENT_SECRET}"
    callbackUrl:  "https://console.example.com/callback"
  allowedGithubUsers:
    - "alice"
  sessionSecret: "${SESSION_SECRET}"
  sessionTtl:    "24h"

llm:
  provider: "openai"          # openai | anthropic | ollama
  apiKey:   "${OPENAI_API_KEY}"
  model:    "gpt-4o"
  allowedCommands:
    - "go"
    - "npm"
    - "make"
    - "git"

workspaces:
  - id:       "my-project"
    name:     "My Project"
    rootPath: "/srv/workspaces/my-project"

Secrets are read from environment variables and are never stored in the config file or logged.

Development

The Makefile-based workflow (make dev, make test, make build) described in the design and planning documents is not yet implemented in this repository. Until a Makefile is added, please refer to the project-specific setup and run instructions (and docs/PLAN.md) for the current development, testing, and build process.

Documentation

• Design Document
• Implementation Plan
• Wireframes

Docs Site

The site/ directory contains a Hugo static site built from the docs above, deployable to either Cloudflare Pages or Cloudflare Workers.

make site-serve   # serve locally with live reload
make site-build   # build static output to site/public/

Cloudflare Pages

Connect the repository to Cloudflare Pages via the dashboard and set these build settings (root directory: site):

Setting	Value
Build command	hugo --minify
Build output directory	public
HUGO_VERSION	0.146.0

Cloudflare Workers

The root wrangler.toml configures the site for deployment as a Cloudflare Worker with Static Assets. This approach gives you full Workers programmability on top of the static site if needed in the future.

Prerequisites:

npm install -g wrangler   # install the Wrangler CLI
wrangler login            # authenticate with your Cloudflare account

Deploy:

wrangler deploy           # builds the site and publishes the Worker

Wrangler runs make site-build (configured in wrangler.toml) before uploading, so no separate build step is needed. On first deploy, Wrangler creates the Worker in your Cloudflare account using the name defined in wrangler.toml (dev-console-site). Subsequent wrangler deploy calls update the existing deployment.

To preview the Worker locally before deploying:

wrangler dev

Roadmap

See docs/PLAN.md for the full phased implementation plan and docs/DESIGN.md for the complete product design document.

Phase	Goal
1	Minimal E2E terminal (auth + workspace + PTY + React shell)
2	File browsing (tree + viewer)
3	Basic agent chat (read-only tools)
4	Change proposal & diff review
5	Full agent toolset + manual file editor
6	Polish, PWA, TLS, observability

Security

• All routes require a valid GitHub OAuth session.
• File paths are validated against the workspace root to prevent path traversal.
• Agent run_command calls are restricted to an operator-configured allowlist.
• TLS is terminated by the server (cert/key config) or a reverse proxy.
• Secrets are passed via environment variables, never in config files.

See docs/DESIGN.md § Security Considerations for the full threat model.

License

This project is licensed under the MIT License.