What's the difference between NoAuth and WithAuth?

NoAuth is a minimal 2-router streaming service (9 API routes) with no login wall — ideal for embedding into an existing app. WithAuth adds JWT authentication, RBAC, OTP email verification, Fernet-encrypted password reset tokens, and an admin dashboard — 6 router files and 26 API routes total. The relationship is additive: WithAuth includes everything NoAuth has.

Can I test this without any LLM API keys?

Yes — a built-in Mock provider streams realistic timed test tokens with zero API cost, ideal for local development, UI testing, and CI/CD pipelines. Switch to a real provider (OpenAI, Anthropic, Gemini, or Groq) by setting the corresponding API key.

How does cancellation work when a user closes the tab?

The server polls client connectivity in the background. If a user closes their tab or loses connection mid-generation, the server detects the broken connection within roughly 250ms and cancels the active asyncio task via RunManager, preventing wasted LLM token spend.

Which databases are supported?

MongoDB (via Motor/PyMongo) or PostgreSQL (via SQLAlchemy 2 + Alembic migrations), selected per tier. Both offer full feature parity for session and chat history persistence.

NoAuth tier — clean landing screen, dark mode

1 / 7

AI Chatbotsintermediate

BoilerShelf Streaming Chatbot

Full streaming SaaS platform with JWT auth, backed by MongoDB

$69one-time payment

Everything in NoAuth + MongoDB, plus:
JWT authentication with bcrypt hashing and refresh cookies
Email OTP verification and Fernet-encrypted password reset tokens
Role-based access control (admin / user)
Admin dashboard with user management and audit logs
6 router files total (adds auth, users, admin, audit_logs) — 26 API routes
Next.js frontend expands to 4 pages (adds login, signup, admin)

Instant download

Secure checkout

Refund if we can't fix it

Commercial license

Have a question before buying? Contact us

5 minSetup Time

80 hrsDev Time Saved

At a glance

Real-time streaming conversational AI assistants with token-by-token renderingMulti-provider LLM switching per request (OpenAI, Anthropic, Gemini, Groq)Local development and CI/CD testing with zero API cost via the built-in Mock modelProduction AI backends requiring client disconnect detection and active run cancellation to prevent wasted tokens

Setup in about 5 minutes

Saves roughly 80 dev hours

Infra estimate: $0 with the built-in Mock provider and free/self-hosted database tiers; pay-as-you-go LLM API usage once you switch to a real provider

Built with

ReactTypeScriptTailwind CSSFastAPIPythonMongoDBPostgreSQLSQLAlchemy

Product details

About this product

BoilerShelf Streaming Chatbot

A production-grade developer starter kit for high-concurrency, ultra-low-latency real-time LLM token streaming — the responsive ChatGPT-style experience where text appears token-by-token as it's generated, not all at once.

Core capabilities

Server-Sent Events (SSE) streaming: standardized event lifecycle (stream_start, token, metadata, error, stream_end).
Real-time client cancellation: a RunManager tracks every active generation task — clicking "Stop" or closing the tab immediately aborts upstream LLM generation, preventing wasted token spend.
5 LLM backends: OpenAI, Anthropic Claude, Google Gemini, Groq, and a built-in zero-cost Mock model for local development and CI without any API keys.
Redis rate limiting: atomic IP-based fixed-window limiting with graceful fallback if Redis is unavailable.

Four tiers, two independent choices

Pick your database (MongoDB or PostgreSQL) and your auth level (NoAuth or WithAuth) — all four combinations ship the identical streaming engine:

NoAuth tiers: a minimal 2-router backend, ideal for embedding into existing apps or internal tools.
WithAuth tiers: adds full JWT authentication, RBAC, OTP email verification, Fernet-encrypted credential handling, and an admin dashboard with audit logs.

This is a pure text-streaming product — no voice, no document RAG. If you need a voice-in/voice-out assistant instead, see the Voice Chatbot product.

streaming-chatbotsse-streamingllm-streaminglangchainmulti-tierfastapimongodbpostgresql

Key features

Real-Time SSE Token Streaming

Standardized Server-Sent Events streaming with a structured event lifecycle: stream_start, token, metadata, error, and stream_end.

Active Disconnect Detection & Run Cancellation

Background task polling client connectivity to cancel LLM tasks immediately on disconnect, plus an explicit POST /runs/{run_id}/cancel endpoint backed by a real RunManager class.

Zero-Cost Mock LLM Provider

A built-in MockChatModel streams realistic timed test tokens — full local testing of streaming, UI animations, and cancellation without any paid API keys.

Multi-Provider LLM Factory

Unified LangChain abstraction supporting OpenAI, Anthropic Claude, Google Gemini, Groq, and the Mock provider, switchable per request.

JWT Authentication & Admin Dashboard (WithAuth tiers)

Bcrypt password hashing, refresh-token cookies, email OTP verification, Fernet-encrypted password reset tokens, and an admin dashboard with user management and audit logs.

Redis Rate Limiting with Graceful Fallback

Atomic IP-based fixed-window rate limiter that bypasses checks gracefully if Redis is offline or unconfigured.

What's included & Fit

Included

FastAPI backend source code for the tier you purchase

Next.js + TypeScript frontend source code, wired to the backend API

LangChain multi-provider streaming service factory (OpenAI, Anthropic, Gemini, Groq, Mock)

Session and chat history persistence layer

Redis rate limiting service with automatic fallback handling

Active run management and client disconnect polling service

Docker & Docker Compose files for local development

OpenAPI specification and Postman collection

Not included

Paid LLM provider API credits (Mock provider works with zero cost)

Managed database subscriptions (MongoDB Atlas, managed Postgres, Redis Cloud)

Cloud hosting fees

Target audience

AI app builders who need a battle-tested SSE streaming backend without building cancellation mechanics from scratch

Enterprise teams embedding a high-performance LLM chat console into internal or customer-facing apps

SaaS founders needing a ready-to-monetize chat product with logins, roles, and persistence out of the box

Perfect for

Real-time streaming conversational AI assistants with token-by-token rendering

Multi-provider LLM switching per request (OpenAI, Anthropic, Gemini, Groq)

Local development and CI/CD testing with zero API cost via the built-in Mock model

Production AI backends requiring client disconnect detection and active run cancellation to prevent wasted tokens

Not suitable for

Projects needing voice input/output (see the Voice Chatbot product instead)

Retrieval-augmented generation pipelines requiring vector databases and document chunking

BoilerShelf Streaming Chatbot · WithAuth + MongoDB

$69

one-time payment

Instant download Secure checkout Refund if we can't fix it Commercial license

Details

commercial License

Email Support

Languages

ReactTypeScriptTailwind CSSFastAPIPythonMongoDBPostgreSQLSQLAlchemyAlembicRedisLangChain

Status

Version1.0.0

Last updatedNov 2026

Impact

80+ hours saved

Estimated dev time

~5 min setup

Time to first run

$69