Tableau → Power BI Migration Agents

The problem

Migrating enterprise dashboards from Tableau to Power BI is slow, repetitive manual work — every chart, calculated field, filter and axis has to be rebuilt by hand in a different tool with a different query language.

Why it needed solving

Dashboard migration effort scales linearly with dashboard count, and organizations moving BI platforms have hundreds of them. Most of the work is mechanical translation, which is exactly what parsing plus an LLM can absorb.

The approach

Parse the Tableau workbook format directly to recover dashboard structure, then use LLM agents to generate the equivalent DAX and Power Query (M) for Power BI, exposed through a FastAPI service.

Architecture

• Parser layer — custom parsers for Tableau (.twb) XML files extracting chart types, calculated fields, filters and axis configurations.
• Agent layer — LangGraph/LangChain agents mapping extracted structures to Power BI equivalents.
• Generation layer — LLM-driven DAX query and Power Query (M) script generation.
• Service layer — FastAPI endpoints exposing the chatbot and automation agents to the rest of the pipeline.

Implementation

• Wrote custom .twb XML parsers rather than relying on an LLM to read raw workbook files, so structural extraction stayed deterministic.
• Handed the parsed structures to LangGraph/LangChain agents for the parts that genuinely need judgment — expression translation.
• Generated DAX and Power Query (M) output targeted at Power BI's semantics rather than transliterating Tableau syntax.
• Built the FastAPI service layer so the migration agents were callable from the broader pipeline.

AI/ML components

Challenges & how they were handled

Results