Build-a-Bot is an educational tool developed by the Stanford Accelerator for Learning's Accelerator Studio. This interactive platform empowers users to deconstruct, design, and experiment with AI chatbots through an intuitive block-based interface.

Purpose

Build-a-Bot serves as a hands-on introduction to AI system design, allowing users to experiment with chatbot design concepts in a visual, accessible way. The tool is specifically designed to support in-person workshop facilitation, making complex AI concepts tangible for learners of various backgrounds.

How It Works

Users can:

• Create custom chatbots by assembling blocks in the visual workspace
• Define system prompts that shape the AI's behavior and responses
• Configure starter messages to guide user interactions
• Test their creations in real-time through the integrated chat interface
• Publish their bots to a shared gallery for others to try and remix

Why Chatbots?

Chatbots are often the way that people first experience generative AI. They provide an interactive and intuitive interface for users to engage with AI technology, making it accessible and user-friendly.

This tool helps demystify how AI chatbots work by visualizing the relationship between system prompts and AI behavior. Through experimentation and iteration, users develop a deeper understanding of:

• How AI systems interpret instructions
• The importance of interative design in AI development
• Ethical considerations in AI development
• The creative potential of human-AI collaboration

• System Prompt
• Reference Material

A Start Message helps people to understand how they can interact with your chatbot. It provides initial guidance or examples of questions users can ask, making it easier for them to begin the conversation.

Great job! You've created a System Prompt that defines your chatbot's personality and behavior.

Save Your Work

Download your system prompt to keep it safe and use it with other AI tools:

Download as Markdown

Share Your Creation

Publish your chatbot to the gallery so others can try it out and learn from your design:

Publish to Gallery

---

Try It Out

Want to see how ChatGPT would respond using your system prompt?

Try on ChatGPT

Privacy

No personal data is being collected through this application. Your privacy is important to us, and we've designed this tool with that in mind.

If you choose to publish your chatbot to our public Chatbot Gallery, the system prompt and start messages are saved to our database so that the gallery can function. Other users will be able to see and remix your system prompt, so be sure to exclude any personal information.

AI Model Information

This application uses OpenAI's GPT-4o-mini model for generating responses. This is a powerful language model known for its ability to understand and generate human-like text based on the input it receives.

Data Usage

While we use the GPT-4o-mini model to generate responses, no data from your interactions are stored or used for any purpose other than providing immediate responses within your current chat session. When reference documents are attached, relevant excerpts are sent to the AI model as part of each chat message to provide context for responses.

Local Storage

Some information, such as your system prompt and reference document text, may be stored in your browser's local storage for convenience. This data remains on your device and is only used to customize your chatbot.

Reference Documents

When you upload a document (PDF, DOCX, TXT, or MD) or link a Google Doc using a Reference Document block:

• The text content is extracted in your browser. Original files are never uploaded to our servers.
• For Google Docs, the document is fetched once as markdown and stored as text. The Google Doc link is not retained.
• Document text is split into chunks and sent to OpenAI's embedding API to generate semantic search vectors. This happens once when the document is uploaded.
• The extracted text and search vectors are stored in your block configuration and persisted in your browser's local storage.
• When you chat, your message is also sent to the embedding API so it can be compared against the stored document vectors to find the most relevant content.
• If you publish your bot, the extracted text and search vectors are saved to our database and become publicly accessible to anyone who uses or remixes your bot.
• Do not upload documents containing personal, confidential, or copyrighted material if you intend to publish your bot.

Build-a-Bot Privacy Policy
OpenAI Enterprise Privacy Policy

Select a previous bot configuration to restore it. The current configuration will be saved before loading.