Codistant

AI-powered code assistant
Chat with Llama 3.3 70B, get debugging help, and run multi-step code reviews - with persistent memory.

Live site · Open app · GitHub

---

Features

• Streaming AI Chat - Real-time streaming responses from Llama 3.3 70B via Workers AI
• Conversation Memory - Messages persist across page reloads using Durable Objects with SQLite
• Project Context - Set your project details so the AI gives tailored responses
• Multi-Step Code Review - Paste code and get a 3-step review via Cloudflare Workflows
• Landing + App - Marketing site at /, assistant at /app/
• Dark/Light Theme - Toggle between themes with a Cloudflare-inspired palette

Architecture

┌─────────────────────────────┐
│  Vercel (codistant.v-ai.org)│
│  /        Landing page      │
│  /app/    Chat UI           │
└─────────────┬───────────────┘
              │  fetch(API_BASE + /api/...)
    ┌─────────▼──────────┐
    │  Cloudflare Worker │  cf-ai-codistant.*.workers.dev
    └──┬─────────────┬───┘
       │             │
┌──────▼───┐  ┌──────▼──────────────┐
│ Durable  │  │ Workflow            │
│ Object   │  │ (CodeReview)        │
│ Session  │  │ Analyze → Issues →  │
│ + SQLite │  │ Improvements        │
└────┬─────┘  └─────────┬───────────┘
     └────────┬─────────┘
       ┌──────▼──────┐
       │ Workers AI  │
       │ Llama 3.3   │
       └─────────────┘

Tech Stack

Component	Technology
Frontend	Static HTML/CSS/JS on Vercel (codistant.v-ai.org)
LLM	Llama 3.3 70B (@cf/meta/llama-3.3-70b-instruct-fp8-fast) via Workers AI
State/Memory	Durable Objects with embedded SQLite
Workflow	Cloudflare Workflows
Backend	Cloudflare Workers (TypeScript)

Prerequisites

• Node.js >= 18
• A Cloudflare account (free tier works)
• A Vercel account for the frontend
• Wrangler CLI (via npx)

Setup & Running

1. Clone and install

git clone https://github.com/VishalPatil18/codistant.git
cd codistant
npm install

2. Authenticate with Cloudflare

npx wrangler login

3. Run the API locally

npm run dev

This starts the Worker (and optionally serves frontend/ assets) at http://localhost:8787.

frontend/config.js uses a same-origin API base when the UI is served from port 8787. On Vercel / production it points at:

https://cf-ai-codistant.vishalrp18100.workers.dev

4. Deploy

A. Deploy the Worker (API)

npm run deploy

Live API: https://cf-ai-codistant.<your-subdomain>.workers.dev

After CORS or API changes, redeploy the Worker before testing the Vercel site.

B. Deploy the frontend (Vercel)

1. Import the GitHub repo in Vercel
2. Set Root Directory to frontend
3. Framework Preset: Other (no build command)
4. Attach domain codistant.v-ai.org
5. Deploy

Routes:

Path	File
/	frontend/index.html (landing)
/app/	frontend/app/index.html (chat)

5. CORS

The Worker allows:

• https://codistant.v-ai.org
• *.vercel.app (preview deployments)
• Local static / wrangler origins (localhost:8787, :3000, :5500)

Streaming chat responses also include Access-Control-Allow-Origin.

API Endpoints

Method	Path	Description
POST	/api/chat	Send a chat message. Body: { message, sessionId }. Returns SSE stream.
GET	/api/history?sessionId=xxx	Get conversation history for a session.
DELETE	/api/history	Clear history. Body: { sessionId }.
POST	/api/context	Set user context. Body: { sessionId, key, value }.
GET	/api/context?sessionId=xxx	Get all saved context for a session.
POST	/api/review	Start a code review workflow. Body: { sessionId, code, language }. Returns { instanceId }.
GET	/api/review?instanceId=xxx	Poll code review workflow status.

Project Structure

codistant/
├── src/
│   ├── index.ts                  # Worker entry point & API router
│   ├── cors.ts                   # Allowed origins / CORS headers
│   ├── chat-session.ts           # Durable Object: conversation memory & AI chat
│   ├── code-review-workflow.ts   # Workflow: 3-step code review
│   └── types.ts                  # Shared TypeScript types
├── frontend/
│   ├── index.html                # Landing page
│   ├── landing.js                # Landing theme + hero mock
│   ├── styles.css                # Shared design system + landing styles
│   ├── config.js                 # CODISTANT_API_BASE
│   ├── logo.svg
│   ├── vercel.json
│   └── app/
│       ├── index.html            # Chat UI
│       └── app.js                # Client logic (SSE, sessions, markdown)
├── wrangler.toml
├── package.json
├── README.md
└── PROMPTS.md

How It Works

Chat Flow

1. User sends a message from /app/
2. The browser calls the Worker /api/chat
3. The Worker routes to the user's Durable Object (sessionId)
4. The DO stores the message, builds a prompt, and streams Workers AI over SSE
5. The full reply is saved to SQLite for later context

Code Review Flow

1. User pastes code and clicks Run Code Review
2. The Worker creates a Cloudflare Workflow instance
3. Three sequential AI steps run (with retries): analyze → find issues → suggest improvements
4. The browser polls until complete and shows the report

Manual deploy checklist

1. npm run deploy - publish Worker with latest CORS / API
2. Push or redeploy Vercel project (root frontend)
3. Confirm https://codistant.v-ai.org/ (landing) and /app/ (chat)
4. Smoke-test chat streaming, history, context, and code review from the Vercel origin
5. Commit locally when you are ready (this repo does not auto-commit)

Local Development Notes

• wrangler dev simulates Durable Objects / Workflows and proxies Workers AI (requires wrangler login)
• SQLite data for local DO state lives under .wrangler/
• Override the API target in the browser console if needed: window.CODISTANT_API_BASE = "https://..."