Showboat Ecosystem Expands with Chartroom and Datasette Integration for AI-Assisted Documentation

Revolutionizing AI Documentation: Showboat’s New Ecosystem Powers Real-Time Code Insights

In a significant leap forward for AI-assisted programming, developer and open-source advocate Simon Willison has introduced two new tools—Chartroom and datasette-showboat—that extend the capabilities of his Showboat CLI framework. Originally launched to help coding agents generate self-documenting Markdown files with embedded code outputs and screenshots, Showboat has now evolved into a dynamic ecosystem that bridges AI-generated analysis with real-time visualization and remote collaboration.

According to Willison’s detailed blog post, Showboat v0.6.0 introduced a groundbreaking remote publishing feature that allows AI agents to stream document updates to a designated web endpoint as they work. This innovation eliminates the traditional wait for code execution to complete before viewing results. By setting the SHOWBOAT_REMOTE_URL environment variable, developers can direct Showboat’s output—whether it’s a code snippet, screenshot, or chart—to a live server, enabling instant feedback loops during AI-driven development sessions.

Datasette-Showboat: Real-Time Document Streaming to a Personal Dashboard

The datasette-showboat plugin is a masterstroke of ecosystem integration. Built as a Datasette extension, it provides two endpoints: /-/showboat for viewing all generated documents and /-/showboat/receive for accepting incoming Showboat payloads. This allows developers to host their own lightweight, SQLite-backed dashboard that auto-updates as AI agents like Claude Code produce new documentation.

Willison demonstrates the tool’s utility by running a simple command to spin up a local Datasette instance with Showboat support, then configuring Claude Code to send document fragments to it. As the AI agent executes showboat init, showboat exec, and showboat image commands, the dashboard refreshes in real time, displaying each new chunk of documentation—including embedded screenshots of web interactions via Rodney, Willison’s companion browser automation tool. This transforms the AI coding process from a black box into a transparent, iterative workflow.

Chartroom: AI-Generated Visualizations at Your Fingertips

Equally transformative is Chartroom, a lightweight CLI tool built atop matplotlib that enables AI agents to generate publication-ready charts directly from CSV, TSV, JSON, or SQL query results. With commands like uvx chartroom bar --csv, agents can now produce visual summaries of data analysis—such as sales trends or user behavior metrics—and embed them automatically into Showboat documents.

Chartroom goes beyond image generation: it can output alt text, HTML, or Markdown-formatted image tags with accessibility metadata. For instance, a bar chart of sales data generates alt text like: “Sales by Person. Bar chart of value by name — Alice: 42, Bob: 28, Charlie: 35, Diana: 51, Eve: 19.” This ensures AI-generated documentation is not only informative but also inclusive and standards-compliant.

Willison built Chartroom using his own click-app template and guided a Claude Code session through a TDD workflow, resulting in a tool that’s both powerful and simple enough to be understood from its --help output alone. The result is a seamless pipeline: AI queries a database, Chartroom renders the insight, Showboat packages it with context, and datasette-showboat publishes it live.

The Broader Implications

The convergence of these tools signals a maturation of the AI-assisted programming paradigm. No longer are developers merely asking AI to write code—they’re collaborating on documentation, visualization, and deployment in real time. The remote publishing mechanism, implemented via a simple HTTP POST endpoint, opens the door to future integrations: CI/CD pipelines, Slack notifications, or even AI-driven audit trails.

As Willison notes, these tools operate on loose conventions—any tool that outputs text or images can integrate with Showboat. This modular philosophy suggests a future where AI agents don’t just code, but curate, visualize, and document their own work, creating living records of computational thought.