209 ·GitHub
What is this?
DSH Data Agent is a DeepSeek Harness (DSH) plugin that lets you connect to your database and analyze data through natural-language conversation, automatically writing and running SQL, generating charts and dashboards, and surfacing business insights.
DSH Data Agent is a DeepSeek Harness (DSH) plugin that lets you connect to your database and analyze data through natural-language conversation, automatically writing and running SQL, generating charts and dashboards, and surfacing business insights.Is this a standalone app, or does it plug into something else?It is a plugin that runs inside DeepSeek Harness (DSH), adding a data-analysis mode you reach from the DSH Web console or its terminal UI (dsh-tui), not a separate program you run on its own.
It is a plugin that runs inside DeepSeek Harness (DSH), adding a data-analysis mode you reach from the DSH Web console or its terminal UI (dsh-tui), not a separate program you run on its own.How do I install and start using it?Install the plugin from npm in a terminal, then in the DSH Web console start a new chat, switch to Data Mode and configure the database connection, or in dsh-tui type /preset data-agent and /database connect to begin asking questions.
Install the plugin from npm in a terminal, then in the DSH Web console start a new chat, switch to Data Mode and configure the database connection, or in dsh-tui type /preset data-agent and /database connect to begin asking questions.Do I need DeepSeek Harness already set up?Yes, it requires the DeepSeek Harness (DSH) runtime plus local or network access to the target database, whether a local SQLite file or a LAN/cloud database.
Yes, it requires the DeepSeek Harness (DSH) runtime plus local or network access to the target database, whether a local SQLite file or a LAN/cloud database.Is it safe to point this at a production database?Yes if you use a read-only setup: the project recommends a read-only database account with read-only mode enabled, which strictly blocks high-risk statements like UPDATE, DELETE and DROP.
Yes if you use a read-only setup: the project recommends a read-only database account with read-only mode enabled, which strictly blocks high-risk statements like UPDATE, DELETE and DROP.Which databases can I connect to?It supports relational databases (MySQL, PostgreSQL, SQLite, Oracle, Microsoft SQL Server) and OLAP warehouses (ClickHouse, Apache Doris, Apache Hive, Apache Impala), plus local SQLite files that work out of the box without any server.
It supports relational databases (MySQL, PostgreSQL, SQLite, Oracle, Microsoft SQL Server) and OLAP warehouses (ClickHouse, Apache Doris, Apache Hive, Apache Impala), plus local SQLite files that work out of the box without any server.How do I share the charts it generates with people who don't use DSH?Each analysis can produce a self-contained offline HTML report saved under the analysis-reports/ folder, with charts and data tables built in, so anyone can open it in a browser without installing software or needing network access.
Each analysis can produce a self-contained offline HTML report saved under the analysis-reports/ folder, with charts and data tables built in, so anyone can open it in a browser without installing software or needing network access.Can I use this if I don't know SQL at all?Yes, it's built for non-technical business users: you describe what you want in plain language and the agent finds the right tables, writes and runs the SQL, and returns readable charts and conclusions.
Yes, it's built for non-technical business users: you describe what you want in plain language and the agent finds the right tables, writes and runs the SQL, and returns readable charts and conclusions.What does the built-in data governance feature do?It has the AI scan your tables and columns and auto-generate plain-language business definitions, which you can review, correct, or extend with custom metrics (like 'net revenue = order amount - refunds') for consistent future analysis.
It has the AI scan your tables and columns and auto-generate plain-language business definitions, which you can review, correct, or extend with custom metrics (like 'net revenue = order amount - refunds') for consistent future analysis.What are similar projects?Chat2DB, dbx, usql, clearai-dsh, sqlit, mcp-toolbox, navop, redash
Chat2DB, dbx, usql, clearai-dsh, sqlit, mcp-toolbox, navop, redashWhich topics is it in?data-governance, oracle, sql-server, clickhouse, analytics, dsh, dsh-plugin, deepseek-harness, mysql, data-visualization, tui, sql
data-governance, oracle, sql-server, clickhouse, analytics, dsh, dsh-plugin, deepseek-harness, mysql, data-visualization, tui, sql