A sleek, free AI chat interface powered by open-source models.
| Category | Details |
|---|---|
| Multi-Provider Models | OpenRouter + Groq model catalog in one selector |
| Free/Usable Filtering | OpenRouter strict-free + Groq account-usable chat models |
| Auto Refresh | Model availability updates automatically with stale-cache fallback |
| Dynamic Key Rotation | Per-provider key failover, cooldowns, and retry classification |
| Vision Support | Upload images and analyze with vision-capable models |
| Thinking Visualization | Reasoning streams shown with expandable thought chains |
| Rich Rendering | Markdown, LaTeX (KaTeX), syntax-highlighted code blocks |
| Chat History | Persistent local conversation history |
| Streaming Responses | Token streaming with responsive UI updates |
| Zero Framework | Pure HTML/CSS/JS, no frontend build step |
- Frontend - Vanilla HTML, CSS, JavaScript
- Backend - Vercel Serverless Functions (Node.js)
- AI Providers - OpenRouter and Groq
- Rendering - Marked, KaTeX, Highlight.js
- Node.js 18+
- OpenRouter API key(s)
- Groq API key(s)
# Clone the repo
git clone https://github.com/Amer-alsayed/chaqgpt.git
cd chaqgpt
# Install Vercel CLI
npm i -g vercel
# Create .env (JSON arrays are preferred)
echo OPENROUTER_API_KEYS_JSON='["sk-or-1","sk-or-2"]' > .env
echo OPENROUTER_API_KEY=your_openrouter_key >> .env
echo GROQ_API_KEYS_JSON='["gsk_1","gsk_2"]' >> .env
echo GROQ_API_KEY=your_groq_key >> .env
# Start dev server
vercel devThe app will be available at http://localhost:3000.
| Variable | Purpose |
|---|---|
OPENROUTER_API_KEYS_JSON |
JSON array of OpenRouter keys (preferred) |
OPENROUTER_API_KEY |
Single OpenRouter key fallback |
OPENROUTER_API_KEYS_FILE |
Optional path to newline-delimited OpenRouter keys file |
OPENROUTER_KEY_COOLDOWN_RATE_LIMIT_MS |
Optional, default 900000 |
OPENROUTER_KEY_COOLDOWN_AUTH_MS |
Optional, default 21600000 |
OPENROUTER_KEY_COOLDOWN_TRANSIENT_MS |
Optional, default 30000 |
OPENROUTER_MAX_FAILOVER_ATTEMPTS |
Optional, default 10 |
GROQ_API_KEYS_JSON |
JSON array of Groq keys (preferred) |
GROQ_API_KEY |
Single Groq key fallback |
GROQ_API_KEYS_FILE |
Optional path to newline-delimited Groq keys file |
GROQ_KEY_COOLDOWN_RATE_LIMIT_MS |
Optional, default 900000 |
GROQ_KEY_COOLDOWN_AUTH_MS |
Optional, default 21600000 |
GROQ_KEY_COOLDOWN_TRANSIENT_MS |
Optional, default 30000 |
GROQ_MAX_FAILOVER_ATTEMPTS |
Optional, default 10 |
GROQ_MODEL_PROBE_CONCURRENCY |
Optional, default 4 |
GROQ_MODEL_PROBE_MAX_ATTEMPTS |
Optional, default 2 |
The app auto-refreshes model lists. OpenRouter models are filtered to strict-free pricing; Groq models are filtered to chat-usable models for your keys.
chaqgpt/
+-- api/
¦ +-- chat.js
¦ +-- image.js
¦ +-- models.js
¦ +-- lib/
¦ +-- model-catalog.js
¦ +-- openrouter-models.js
¦ +-- groq-models.js
¦ +-- provider-key-pool.js
¦ +-- openrouter-key-pool.js
¦ +-- groq-key-pool.js
+-- assets/
¦ +-- css/style.css
¦ +-- js/app.js
+-- index.html
+-- server.js
+-- README.md
- API keys are never exposed to the browser.
- Provider calls are proxied through server-side handlers.
.envis ignored by git.
This project is licensed under the MIT License.