Compare chatbot boilerplates
AI Wrapper vs Streaming vs Voice Chatbot
Compare the AI Wrapper Kit, Streaming Chatbot and Voice Chatbot boilerplates: response style, LLM providers, speech, authentication, databases and deployment.
BoilerShelf's three chatbot boilerplates each do one job well. The AI Wrapper Kit is the quickest way to put a multi-provider chat product on FastAPI. The Streaming Chatbot adds token-by-token responses with instant cancellation. The Voice Chatbot lets users speak and hear replies.
All three come with or without JWT authentication, and on MongoDB or PostgreSQL, so you pay only for the scope you need.
Side by side
At a glance
| AI Wrapper Kit | Streaming Chatbot | Voice Chatbot | |
|---|---|---|---|
| Editions and prices | $39 to $59 across 4 editions | $49 to $69 across 4 editions | $49 to $79 across 4 editions |
| How replies arrive | Complete reply per request over REST; no token-by-token streaming | Token-by-token streaming over server-sent events, with instant cancellation | Spoken replies; turn-based conversation with memory across the session |
| LLM providers | OpenAI, Anthropic, Google Gemini, Groq | OpenAI, Anthropic, Google Gemini, Groq, Mock | Groq |
| Speech | Not included | Not included | Local Whisper speech-to-text and Kokoro text-to-speech, so no cloud speech API fees |
| Interface | Chat workspace with a model selector | Streaming chat interface | Voice studio with a live waveform and microphone or text input |
| Authentication | NoAuth, or WithAuth with JWT, email OTP and an admin console | NoAuth, or WithAuth with JWT, roles and an admin dashboard | NoAuth, or WithAuth with JWT, roles and an admin dashboard |
| Databases | MongoDB or PostgreSQL edition | MongoDB or PostgreSQL edition | MongoDB or PostgreSQL edition |
| Difficulty | Beginner | Intermediate | Intermediate |
| Setup time | About 10 minutes | About 5 minutes | About 15 minutes |
| Deployment targets | Vercel, Render, Railway | Docker, Render, Railway, AWS ECS, DigitalOcean App Platform | Vercel, Render, Railway |
Which one
How to choose
Choose the AI Wrapper Kit to prototype an AI product fast
You want a working multi-provider chat backend and UI you can brand and extend. It is the simplest of the three. It does not stream tokens and it has no document upload or retrieval.
View AI Wrapper Kit · $39 to $59 across 4 editions
Choose the Streaming Chatbot for a ChatGPT-style experience
Replies appear as they are generated and the user can stop generation at any point. It includes a zero-cost mock model, so you can build and test the interface without spending on API calls.
View BoilerShelf Streaming Chatbot · $49 to $69 across 4 editions
Choose the Voice Chatbot when users talk to the assistant
Speech recognition and speech synthesis run on your own server, so there are no per-request speech fees. The language model integration is Groq only, so it is not the right base if you need another LLM provider.
View BoilerShelf Voice Chatbot · $49 to $79 across 4 editions
Which authentication edition?
NoAuth editions are a minimal service with no login wall: good for internal tools, prototypes and products where you add your own authentication. WithAuth editions add JWT authentication, roles and an admin dashboard, so you can run the chatbot as a multi-user product straight away.
Within each product, the MongoDB and PostgreSQL editions are the same product on a different database. Choose the database your application already uses.
What to use for document question answering
None of the three chatbots retrieves answers from your own documents. If you need that, look at the RAG boilerplates instead: the Verified RAG Delivery Kit for cited answers and evaluation, or the LangGraph RAG Agent for a self-correcting agent.