CASE EVIDENCE / A RECORD
Tunnello / Hello Agent team 公开的代码代理与软件工程案例，来源为 公开代码库，复核于 2026-06-27T07:30:26Z。
原始记录：Tunnello ChatBI offers a Qwen3 Max Preview powered conversational BI agent for natural-language SQL and charts
Chinese Brief
Tunnello / Hello Agent team 公开的代码代理与软件工程案例，来源为 公开代码库，复核于 2026-06-27T07:30:26Z。 Model Atlas 将它标记为 A 类证据，因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。 Model Atlas 不把 benchmark、教程、发布说明或集合页包装成真实案例。
任务
这是一个围绕软件工程任务执行的真实任务，公开材料可以回溯到具体使用者和具体产物。 原始资料写作：ChatBI is a LangGraph-based conversational business-intelligence application that turns natural-language questions into SQLite SQL, executes queries, returns results, and generates visual charts through MCP tools; its A…
公开材料提供公开代码、README 或项目配置，可用于核验任务结果、项目形态和模型绑定关系。 原始资料写作：The public repo includes screenshots and a Bilibili demo video for querying product categories, order trends, and chart visualizations from a local database through the ChatBI interface.
Qwen3 Max 在该案例中承担软件工程任务执行相关的生成、分析、编排或实现角色。 原始资料写作：Qwen3 Max family acts as the agent LLM that routes tasks, reasons over schema retrieved by RAG, generates SQL, calls execution/chart tools, and composes the final streamed BI answer.
当前判断基于公开材料；若产物下线、仓库变更或模型参与比例仅来自作者自述，需要在引用前重新复核。 原始资料写作：The exact selectable model shown in the API docs is qwen3-max-preview; this is treated as a Qwen3 Max family case. Demo video is public and reachable, but individual inference logs are not published.