Datasette Agent
Datasette Agent is an AI assistant tool built on large language models (LLMs), designed for integration with the Datasette data exploration platform. It enables users to interact with databases through natural language, automatically converting user queries into SQL and returning results. The tool is extensible, supports multiple LLM backends, and is applicable to data analysis and database exploration scenarios, lowering the barrier for non-technical users working with structured data.
Timeline (last 90 days)
When the panel feature works together with the datasette-apps plugin, users can use a prompt to generate a database-driven app from scratch and modify it directly in the chat window next to the app.
Datasette Agent released version 0.5a0, which introduces a core feature called panel that docks the AI chat window alongside the Datasette page, forming a workflow of chatting, viewing and editing at the same time.
0.5a0 is still an alpha version, and its features and stability may still be iterating.
The Datasette Agent project is open source on GitHub, with the version tag 0.5a0.
In the example, the agent generates the app by calling tools such as app_create and app_render, and the chat history on the left keeps a collapsed record of each tool call and version entry points (View app, Edit).
The "View app" button shown by the Agent tool and the links generated by `#` references open the app in full screen within the Datasette Agent panel.
Typing `#` in the chat input box of Datasette Agent triggers app name autocompletion, with candidates drawn from the apps the current user has permission to view.
In the release notes, Simon Willison mentioned that Datasette Agent has even bigger upgrades queued up.
After 0.1a9, starting Datasette Agent no longer requires additionally declaring the model; the plugin directly adopts LLM's default model setting.
Before 0.1a9, running Datasette Agent required manually passing the model parameter, for example `-s plugins.datasette-agent.default_model gpt-6-astra`.
1 more timeline events
All Facts (12)
Datasette Agent 利用 LLM 的工具调用（Function Calling）能力形成 ReAct（Reasoning + Acting）循环来执行数据查询
90%UnverifiedIn the example, the agent generates the app by calling tools such as app_create and app_render, and the chat history on the left keeps a collapsed record of each tool call and version entry points (View app, Edit).
50%UnverifiedWhen the panel feature works together with the datasette-apps plugin, users can use a prompt to generate a database-driven app from scratch and modify it directly in the chat window next to the app.
50%UnverifiedThe Datasette Agent project is open source on GitHub, with the version tag 0.5a0.
50%Unverified0.5a0 is still an alpha version, and its features and stability may still be iterating.
50%UnverifiedThe "View app" button shown by the Agent tool and the links generated by `#` references open the app in full screen within the Datasette Agent panel.
50%UnverifiedTyping `#` in the chat input box of Datasette Agent triggers app name autocompletion, with candidates drawn from the apps the current user has permission to view.
50%UnverifiedAfter 0.1a9, starting Datasette Agent no longer requires additionally declaring the model; the plugin directly adopts LLM's default model setting.
50%UnverifiedIn the release notes, Simon Willison mentioned that Datasette Agent has even bigger upgrades queued up.
50%UnverifiedBefore 0.1a9, running Datasette Agent required manually passing the model parameter, for example `-s plugins.datasette-agent.default_model gpt-6-astra`.
50%UnverifiedData agent allows users to connect company data and mine insights and build interactive dashboards through natural language conversation.
50%UnverifiedDatasette Agent released version 0.5a0, which introduces a core feature called panel that docks the AI chat window alongside the Datasette page, forming a workflow of chatting, viewing and editing at the same time.
50%