[Feature Request] AI chatbot component with MCP database access

[avinashkurup]

What are you building with SQLPage?
I'm building internal data dashboards and admin tools where business users need to ask natural-language questions about the underlying data — things like "what were the top products last month?" or "why did revenue drop on Tuesday?" SQLPage is a perfect fit for the UI layer, but users still need a conversational interface to explore data without writing SQL themselves.

What is your problem?
There's no native way to embed an interactive AI chatbot into a SQLPage app. Users who want a natural-language interface over their data have to leave the SQLPage paradigm entirely and build a custom frontend. This breaks the core promise of SQLPage — that you can build a full, useful web app with only SQL.

What are you currently doing?
Workarounds currently require:

• Writing a custom Typescript node widget outside of SQLPage's component system
• Setting up a separate backend proxy to forward messages to an AI API (to avoid exposing API keys in the browser)
• Manually managing conversation history in Typescript
• Wiring it all together with the fetch component or raw <script> tags in a shell component

This is brittle, hard to maintain, and out of reach for the non-developer audience SQLPage is designed for.

Describe the solution you'd like
An official external component — sqlpage-chat — modeled after sqlpage-spreadsheet, that renders a full chat UI and connects to any OpenAI-compatible HTTP API. Usage would be fully declarative via SQL:

SELECT 'sqlpage-chat' AS component,
       'https://api.openai.com/v1/chat/completions' AS api_url,
       'Bearer ' || $OPENAI_API_KEY AS authorization_header,
       'gpt-4o' AS model,
       'You are a helpful assistant for this sales dashboard.' AS system_prompt,
       'Ask me anything about your data...' AS placeholder;

For richer use cases, it would be nice to support MCP (Model Context Protocol) so the AI can autonomously query the live database — rather than requiring the developer to pre-inject static context into the system prompt:

SELECT 'sqlpage-chat' AS component,
       'https://api.anthropic.com/v1/messages' AS api_url,
       'Bearer ' || $ANTHROPIC_API_KEY AS authorization_header,
       'claude-sonnet-4-5' AS model,
       'postgresql://user:pass@localhost/mydb' AS mcp_database_url,
       'You are a SQL analyst. Use your tools to answer questions accurately.' AS system_prompt;

Key capabilities: streaming responses (stdio support initially), conversation history, error handling, and a clean chat bubble UI that inherits the SQLPage/Tabler theme.

Describe alternatives you've considered

• Injecting DB context into the system prompt via SQL subquery — works for small datasets but breaks on large schemas, gives the model stale data, and bloats the prompt on every turn.
• Using the fetch component — not interactive; it fires once on page load and doesn't support streaming or back-and-forth conversation.
• Embedding an iframe — fully disconnects from SQLPage's auth, theme, and data layer; defeats the purpose.
• Building a standalone chatbot app separately — loses the tight integration with the SQLPage data layer and forces users to maintain two separate apps.

Additional context

• This would live in its own repo (sqlpage/sqlpage-chat) as a CDN-hosted JS/TS bundle, following the sqlpage-spreadsheet pattern.
• The component targets any OpenAI-compatible endpoint, making it provider-agnostic (OpenAI, Anthropic, Ollama, LM Studio, etc.).
• MCP is an open standard by Anthropic (https://modelcontextprotocol.io) now widely supported across AI providers and tools — it's the cleanest way to give the model live, structured access to the SQLPage database without custom tooling.
• API calls default to client-side for simplicity; a server-side proxy mode could be added for environments where API keys must stay server-side.