AI Agent Hub
Back to plugins
🖥️

dsh-plugin-nlbi

Client Updated 2026.08.25

Run the following command in DeepSeek Harness:

dsh plugin install 1byteone/dsh-plugin-nlbi

Paste the following prompt into your AI chat to install this plugin:

Install it by running dsh plugin install 1byteone/dsh-plugin-nlbi in DeepSeek Harness; the source code is available at https://github.com/1byteone/dsh-plugin-nlbi

About this plugin

Ask DeepSeek Harness something like "top regions by sales last month" and you used to get a wall of raw SQL or a flat table of numbers. dsh-plugin-nlbi turns that single Q&A into a lightweight BI session: an editable SQL draft, an interactive data grid, an auto-generated chart (bar, line, pie, or KPI cards), and a bookmarkable report you can re-run anytime—all inside the chat window and a sidebar panel.

The core pipeline is Text2SQL plus guardrails. Your natural-language question is converted to SQL by the LLM, validated at the AST level with node-sql-parser, checked against a regex whitelist, and confirmed read-only before execution. Results render as ECharts visualizations with SVG fallback. Beyond a single query, the plugin adds report saving / re-running / Markdown export, a sidebar SchemaTree for browsing table structures, a paginated data grid with sorting and WHERE filtering, and a sidebar natural-language input that returns results instantly without switching back to the chat. It inherits full connection management and per-session database switching from dsh-mysql, and layers on new Agent tools such as nl_query, sql_to_chart, and save_report.

Safety is baked in at every step: nl_query is read-only regardless of the connection's write flag, dual-layer validation (AST + regex) guarantees no write leak, and table whitelisting, single-statement enforcement, a 15-second timeout, and a 2000-row cap provide defense in depth. Generated SQL appears as a draft that must be explicitly clicked to run, protecting you from model hallucinations. It is built for developers, analysts, and DBAs who query MySQL frequently inside Harness conversations and want a "one question to a chart" workflow without ever leaving the chat window.

Use Cases

  • Ask a data question in chat and receive editable SQL, an interactive grid, and an auto-generated chart in one reply
  • Run instant natural-language queries from the sidebar without ever switching back to the conversation
  • Bookmark routine queries as reports, then re-run or export them as Markdown with a single click

Best For

  • Developers who run frequent MySQL queries inside DeepSeek Harness conversations
  • Data analysts who prefer asking a plain-language question over hand-writing SQL
  • DBAs who want a lightweight BI workbench in Harness to cut down on DBeaver or Metabase context-switching