FastbuildAI: The 3-Minute Guide to Running Your Own AI Chat Platform Locally
“
A straight-to-the-point tutorial for developers, product managers, and curious learners who want a private ChatGPT-style site without writing backend code.
Table of Contents
-
What Is FastbuildAI? -
Why Does It Save You Weeks of Work? -
The 3-Minute, No-Code Launch Checklist -
First-Time Login: Where to Click Next -
Features That Work Today -
Roadmap: What the Team Still Plans to Ship -
FAQ: Real Questions From Early Users -
System Map: One Diagram to Understand the Stack
1. What Is FastbuildAI?
FastbuildAI is an open-source starter kit for building AI-powered web applications.
It ships with:
Layer | Technology | What It Does for You |
---|---|---|
Backend API | NestJS 11 + TypeORM | REST and real-time endpoints |
Database | PostgreSQL 17 | Stores users, chats, payments |
Cache & Queue | Redis | Keeps responses fast |
Front End | Vue 3 + Vite | Responsive chat UI and admin panel |
Container | Docker | One command spins up the whole stack |
Think of it as the WordPress of AI apps: you download it, change a few settings, and you have a working product.
2. Why Does It Save You Weeks of Work?
Most teams spend days—or weeks—on tasks that FastbuildAI already handles:
Common Task | Without FastbuildAI | With FastbuildAI |
---|---|---|
User registration & login | Build JWT flow, password reset, OAuth | Already built—edit logo and colors |
Admin panel | Hand-code CRUD screens, roles, permissions | Pre-built dashboard |
Pay-as-you-go billing | Integrate Stripe or Alipay, build ledgers | Add keys to .env , done |
Model switching | Write routing logic, key rotation | Toggle models in the UI |
Responsive chat UI | Code React/Vue components, streaming | Ready to use |
3. The 3-Minute, No-Code Launch Checklist
Prerequisites
-
Docker and Docker Compose installed (official docs cover macOS, Windows, Ubuntu). -
A terminal and a browser.
Step 1: Clone the Repository
git clone https://github.com/fastbuildai/fastbuildai.git
cd fastbuildai
Step 2: Copy the Environment File
cp .env.production.local.example .env.production.local
Open .env.production.local
in any text editor.
The only line you must change is:
OPENAI_API_KEY=sk-your-real-key-here
Leave the database and Redis URLs untouched for a first run on your laptop.
Step 3: Start Everything
docker compose \
-p fastbuildai \
--env-file ./.env.production.local \
-f ./docker/docker-compose.yml \
up -d
You will see:
Creating fastbuildai_postgres ... done
Creating fastbuildai_redis ... done
Creating fastbuildai_api ... done
Creating fastbuildai_web ... done
Step 4: Wait Two Minutes
The first run downloads images and initializes the database. Grab coffee.
Step 5: Open Your Browser
Visit http://localhost:4090
.
You should see a clean login page.
4. First-Time Login: Where to Click Next
Default Super-Admin Account
-
Username: admin
-
Password: FastbuildAI&123456
Five-Minute Orientation
-
Admin Panel
URL:http://localhost:4090/admin
Menu: Models → Add Model-
Name: gpt-3.5-turbo
-
Base URL: https://api.openai.com/v1
-
Key: (use the same key from .env
)
Click Save.
-
-
Chat Interface
URL:http://localhost:4090/chat
-
Choose the model you just added. -
Type “Hello” and press Enter.
A streaming reply confirms the stack is healthy.
-
-
User Balance
In the admin panel, open Users → Edit next to your own account.
Add 10 USD of credits for local testing.
5. Features That Work Today
5.1 Multi-Model Chat
-
Any provider that mimics OpenAI’s REST format works (OpenAI, Azure, Anthropic via adapters, local LLMs with the same shape). -
Typing indicator, markdown support, code highlighting, and dark mode are all built in.
5.2 MCP Calls
-
MCP stands for Model Context Protocol. -
It lets the model pull extra context—think CSV files, JSON APIs, or database rows—without manual prompt stuffing. -
You register an MCP tool in the admin panel and reference it inside a chat session.
5.3 User Top-Up System
-
Each new account receives 5 USD in credits. -
When balance hits zero, the UI shows a Recharge button. -
Stripe and Alipay are pre-wired; paste your keys into .env
and restart.
5.4 Model Management
-
Toggle availability per user group. -
Example: allow only paying customers to use GPT-4, while free users stay on GPT-3.5.
5.5 Screenshot Tour
Figure 1: Clean chat interface with sidebar history and markdown rendering.
Figure 2: One-click enable/disable for each model.
Figure 3: Admin grid with balances, roles, and last login.
Figure 4: Stripe card form embedded inside the app.
Figure 5: Live CPU, memory, and token usage charts.
(Images 6-9 show the same screens in dark mode and on mobile.)
6. Roadmap: What the Team Still Plans to Ship
The backlog is public in the repo README.
Legend: ✅ released, ⬜ in progress.
Feature | What It Will Do | Why You Might Care |
---|---|---|
Knowledge Base ⬜ | Upload PDF, TXT, MD; ask questions about the content | Build a private ChatPDF for your team |
Smart Agents ⬜ | Let the AI break a big request into smaller tasks and run them | Auto-GPT style goal completion |
Workflow Builder ⬜ | Drag-and-drop nodes to chain multiple AI calls | Low-code automation |
Plug-in System ⬜ | Install third-party extensions from a marketplace | Extend without touching core code |
7. FAQ: Real Questions From Early Users
Q1: Do I have to use Docker?
Yes. The maintainers provide only Docker images to guarantee the same environment on macOS, Windows, and Linux servers.
Q2: Can I deploy to a cloud server?
Absolutely.
-
Copy the .env.production.local
file. -
Replace localhost
inDATABASE_URL
andREDIS_URL
with your server’s IP or internal Docker network names. -
Run docker compose up -d
on the cloud host. -
Point your domain to port 4090 or add an Nginx reverse proxy.
Q3: How do I change the logo or colors?
-
Logo: replace apps/web/public/logo.png
(512 × 512 px works best). -
Primary color: edit --color-primary
inapps/web/src/styles/variables.css
, then rebuild the image or mount the folder for dev mode.
Q4: Where is user data stored?
-
PostgreSQL container: /var/lib/postgresql/data
-
Docker volume is mapped to ./docker/postgres
on the host. -
Back up with:
docker exec fastbuildai_postgres \
pg_dump -U postgres fastbuildai > backup.sql
Q5: What happens when free credits run out?
-
Users see a “Recharge” banner. -
Admins can top-up any account manually or let users pay via Stripe/Alipay.
Q6: Does Chinese (or any non-English) text work?
Yes. The UI and the API layer are UTF-8 end-to-end. Tested with Chinese, Japanese, and emojis.
Q7: How do I add my own fine-tuned model?
-
Treat it like any other OpenAI-compatible endpoint. -
In the admin panel, add a new model row: -
Name: my-finetune
-
Base URL: https://api.yourhost.com/v1
-
Key: your custom key.
-
8. System Map: One Diagram to Understand the Stack
graph TD
subgraph Browser
A[Vue 3 SPA - Chat UI]
B[Vue 3 SPA - Admin UI]
end
subgraph API
C[NestJS Server<br/>REST + SSE]
end
subgraph Storage
D[(PostgreSQL<br/>TypeORM)]
E[(Redis<br/>Cache & Queue)]
end
A <-->|HTTPS / SSE| C
B <-->|HTTPS| C
C <-->|SQL| D
C <-->|KV| E
-
Static front ends are served by an Nginx container (included). -
NestJS handles auth, billing, chat streams, and model routing. -
PostgreSQL keeps relational data, including chat history. -
Redis caches per-user rate limits and short-lived conversation context.
Closing Thoughts
FastbuildAI is not another wrapper around OpenAI—it is a full-stack foundation you can deploy in minutes and extend for years.
Whether you need an internal support bot, a paid SaaS, or a research playground, the heavy lifting (auth, payments, multi-tenant model routing, responsive UI) is already done.
Next steps:
-
Run the three-minute checklist above. -
Swap in your own models or keys. -
Watch the repo for Knowledge Base and Workflow Builder releases.
If something breaks, open an issue with docker logs fastbuildai_api
and the maintainers usually reply within a day.
Enjoy building.