asterchat

Aster — account-backed live AI chat

Aster is a responsive chat website with open registration through Manus sign-in. Anyone with the site link can join using a Manus account; the first successful sign-in automatically creates their Aster account. There is no Aster invite list.

Each account can submit 100 user messages per UTC day, shared across browsers and devices. The server reserves each message atomically in the managed database, so simultaneous requests cannot exceed the cap. The database stores only a SHA-256 pseudonymous account key and daily counts. It does not store chat prompts or replies. The current transcript exists in the page’s memory and is lost on reload or sign-out.

Aster streams replies from the Manus AI chat-completions service. Provider credentials stay server-side. AI requests use the project’s service identity and consume the project’s AI credits. Aster applies a 4,000-character per-message limit and a bounded recent conversation context.

Requirements

• Python 3.12+
• MySQL-compatible managed database with TLS enabled
• A Manus OAuth configuration and an AI service access token or project API key

Local development

Copy the environment template and fill in your project values:

cp .env.example .env
python3 -m venv .venv
. .venv/bin/activate
pip install -r requirements.txt
python scripts/migrate.py
uvicorn app_server:app --host 0.0.0.0 --port 3000 --no-access-log

Use the managed Cloud Preview or published URL for a real OAuth callback. A plain loopback URL is not automatically accepted by the Manus sign-in portal. OAuth callbacks are allowlisted in PUBLIC_APP_ORIGINS to the current published and managed Preview origins; add any custom domain there before using it.

Git and deployment

This repository is ready to be pushed to GitHub and imported into Vercel.

GitHub:

git init
git add .
git commit -m "Initial Aster release"
git branch -M main
git remote add origin <your-github-url>
git push -u origin main

Vercel:

1. Import the repository in Vercel.
2. Set the build command to pip install -r requirements.txt if needed or leave the default.
3. Add the same environment variables from .env.example to Vercel project settings.
4. Deploy.

The included vercel.json routes requests to the FastAPI app in api/index.py, and the frontend remains served from the root public/ assets.

Application structure

• public/index.html and public/aster-chat.js provide the landing page, Manus sign-in, quota display, and streaming chat UI.
• app_server.py provides OAuth callback/session APIs, quota enforcement, health checks, and static files for development.
• api/index.py exports the FastAPI app for Vercel deployment.
• aster/auth.py validates the webdev_app_session HS256 project JWT and implements the OAuth state/identity flow.
• aster/db.py stores opaque account keys and enforces per-day quotas using atomic MySQL updates.
• aster/llm.py streams Manus AI responses and sanitizes rendered Markdown.
• migrations/ and scripts/migrate.py define and apply the additive schema.
• Dockerfile builds the hosted Python server. Published routes send /api/* to the server and page/assets to the static build.

Privacy and limits

Aster persists account-key hashes and per-account UTC daily message counts only. It does not persist conversation text in its database. Messages are sent to Manus AI to generate replies; this statement describes Aster’s storage, not the AI service’s separate handling. Sign out clears the browser session cookie. A new UTC allowance begins at 00:00 UTC.