Toolkit to make Power BI semantic models and Fabric Data Agents AI/Copilot-ready: semantic model AI-readiness analyzers, a Data Agent config validator, a natural-language remediation agent, a Copilot diagnostics troubleshooter, and lakehouse data-source readiness checks.
Each tool lives in its own folder with a dedicated README.
This repository is built for teams searching for: Power BI Copilot readiness, Fabric Data Agent best practices, semantic model optimization for AI, Copilot diagnostics troubleshooting, DirectQuery tuning, and OneLake data-source readiness.
Developed by Farhan Soomro. These are free community utilities, released under the MIT License and provided "AS IS", with no warranty and no liability. See Author, license & disclaimer.
- For Power BI semantic model optimization: start at Start here, then review Tools and Search-friendly use cases.
- For Fabric Data Agent validation and remediation: go to What this toolkit offers, How to pick a tool, and Docs included.
- For Copilot troubleshooting and diagnostics: jump to Tools, How to pick a tool, and Reference documentation.
- For OneLake and lakehouse readiness: use How to pick a tool and Repository organization.
- Start here
- What this toolkit offers
- Search-friendly use cases
- Keywords
- Tooling formats included
- Who this is for
- Tools
- How to pick a tool
- Docs included
- Repository organization
- Suggested GitHub topics
- Suggested GitHub about text
- Reference documentation
- Author, license & disclaimer
🚀 New: Fabric Data Agent Analyzer v3 assesses a Fabric data agent and every connected source in one notebook. Import it into Fabric, set the agent name, and run through section 4. The Source Run Plan then shows which sources the agent actually uses and which source sections to run. See How to use the analyzer.
- Run FabricDataAgentAnalyzer to assess a Fabric Data Agent and every connected source, then use the remediation assets to fix the findings.
- Run SemanticModel-AI-Readiness-Analyzer to score semantic model readiness.
- Run Agent-Readiness-Validator to validate Data Agent configuration.
- If Copilot answers are poor, run Copilot-Diagnostics-Troubleshooter.
- For lakehouse-backed scenarios, run Lakehouse-Data-Source-Readiness.
- Semantic model AI-readiness analysis for naming, descriptions, DAX quality, schema clarity, and performance signals.
- Fabric Data Agent configuration validation against practical best practices.
- Natural-language remediation assets (agent + prompt + script) to translate findings into fix actions.
- Copilot diagnostics parsing and troubleshooting guidance to connect symptoms to model-level fixes.
- Lakehouse data-source readiness checks for agent and Copilot scenarios.
- Power BI Performance Analyzer JSON diagnostics: slow-visual root causes, DAX/DirectQuery/Direct Lake fallback analysis, data-quality checks, and a self-contained HTML report.
- Power BI Performance Analyzer JSON export analyzer and slow report troubleshooting
- Direct Lake fallback to DirectQuery detection from Performance Analyzer
- Power BI Copilot semantic model best practices
- How to prepare semantic model for Copilot and AI
- Fabric Data Agent configuration validator and readiness checklist
- Copilot diagnostics export analyzer for Power BI
- Natural-language remediation for semantic model issues
- DirectQuery best practices checker and max connection tuning
- OneLake shortcut audit and automation for Fabric
- DAX export to Excel paginated-style reporting workflow
- Fabric AI readiness tools and docs for analytics teams
Power BI, Microsoft Fabric, Copilot, Fabric Data Agent, semantic model, AI readiness, Copilot readiness, semantic model analyzer, data agent validator, diagnostics troubleshooter, DirectQuery best practices, DAX optimization, Performance Analyzer, report performance, Direct Lake fallback, OneLake shortcut automation, lakehouse readiness, Tabular Editor, model governance, analytics engineering.
- Jupyter notebooks for analyzers, diagnostics, assessments, and guided workflows.
- PowerShell scripts for repeatable semantic model setting changes and utilities.
- Agent and prompt definition files for remediation workflows.
- Markdown playbooks and checklists for implementation guidance.
- Power BI developers optimizing semantic models for Copilot.
- Fabric data engineers building or validating Data Agents.
- Analytics platform teams troubleshooting Copilot response quality.
- BI governance teams standardizing AI-readiness checks and documentation.
| Tool | What it does |
|---|---|
| FabricDataAgentAnalyzer | Fabric Data Agent Analyzer (v3). Assesses a data agent and every connected source (Lakehouse, Warehouse, SQL DB, Mirrored DB, Eventhouse, semantic model, Graph, Ontology, Azure AI Search) with readiness, coverage, and release gates. Its Source Run Plan shows which sources the agent uses and which sections to run. Also includes a natural-language remediation agent (agent + prompt + Tabular Editor script) for semantic-model findings. |
| SemanticModel-AI-Readiness-Analyzer | Validates a Power BI semantic model for Copilot with 18+ severity-weighted checks (naming, descriptions, DAX, AI Data Schema, performance). |
| Agent-Readiness-Validator | Validates a Fabric Data Agent's configuration (benchmarks, instructions, examples, join risk, failure diagnosis) against Microsoft best practices. |
| Copilot-Diagnostics-Troubleshooter | Parses a Copilot diagnostic export into a prioritized troubleshooting report and a shareable Markdown report, mapping each signal to a model fix. |
| Lakehouse-Data-Source-Readiness | Assesses a Fabric Lakehouse data source for agent/Copilot readiness. (Work in progress.) |
| PowerBI-Performance-Analyzer-Diagnostics | Turns a Power BI Performance Analyzer JSON export into a prioritized root-cause report (DAX, DirectQuery / Direct Lake fallback, rendering, waiting) with overlap-safe timings, data-quality rules, event trees, a timeline, and HTML/Markdown/CSV/JSON outputs. |
| Analyzer | Dataflow Gen2 Analyzer. Checks Fabric Dataflow Gen2 items against the performance best-practices and the copy-activity / dataflow / Spark decision guide (46 rules, refresh-history telemetry, indicative CU cost), then recommends fixes and, where it fits, a better engine (Copy job, Spark, Eventstream). Outputs HTML/Markdown/JSON/CSV. |
| Ontology | Fabric IQ ontology for budget vs. actual (TMDL). Generic budgeting entity types (BudgetLine, Account, OrgUnit, Scenario, CalendarDate), relationships, business rules and 47 metrics, live-bound to a Power BI semantic model. Includes a one-command import script and an end-user guide to import it (script, Git integration or REST API) and use it with the ontology agent, MCP clients and data agents. |
- Configuring a Fabric Data Agent? Start with FabricDataAgentAnalyzer for the whole agent and its sources, then use Agent-Readiness-Validator and Lakehouse-Data-Source-Readiness.
- Preparing a semantic model for Copilot? Start with SemanticModel-AI-Readiness-Analyzer, then use FabricDataAgentAnalyzer to analyze and remediate.
- Copilot answering incorrectly? Download the diagnostics and run Copilot-Diagnostics-Troubleshooter.
- Report page is slow? Export a Performance Analyzer capture and run PowerBI-Performance-Analyzer-Diagnostics.
- PowerBI_Report_Authoring_Unified_Checklist.md - Report-authoring checklist.
- Docs/A Practical Guide to Building and Deploying Data Agents.pdf - Practical implementation guide for data agents.
- Docs/autonomous_fabric_-_agentic_data_quality_using_fabric_data_agents.pdf - Agentic data-quality architecture and workflows.
- Docs/build_smarter_analytics_assistants_with_fabric_data_agents_and_copilot_studio.pdf - Building analytics assistants with Data Agents and Copilot Studio.
- Docs/mastering_fabric_data_agents_from_setup_to_success.pdf - End-to-end setup and adoption guidance.
The repository now separates packaged tools from supporting assets:
- SemanticModel-AI-Readiness-Analyzer/ and related tool folders: production analyzers and validators.
- Notebooks/ - Working notebooks grouped by scenario (semantic model optimization, OneLake, Copilot/agent insights, business demos).
- Scripts/ - Reusable automation scripts grouped by purpose (model settings and utilities).
- Ontology/ - Fabric IQ ontology definitions (TMDL) built from semantic models, with an import script and end-user guide.
- Notebooks/SemanticModel-Optimization/ - Model quality, performance, and schema-shaping notebooks.
- Notebooks/Lakehouse-OneLake/ - OneLake shortcut governance and automation.
- Notebooks/Copilot-Agent-Insights/ - Copilot usage mining and agent quality analysis.
- Notebooks/Business-Demos/ - Architecture and dashboard-oriented demos.
- Scripts/PowerBI-Model-Settings/ - Semantic model connectivity and DirectQuery tuning scripts.
- Scripts/Utilities/ - Helper scripts for operational tasks.
Add these topics in repository settings to improve GitHub search visibility:
- power-bi
- microsoft-fabric
- copilot
- semantic-model
- fabric-data-agent
- ai-readiness
- directquery
- onelake
- dax
- tabular-editor
- analytics-engineering
- business-intelligence
Use this repository subtitle/About text in GitHub settings for search visibility:
Power BI and Microsoft Fabric toolkit for Copilot and AI readiness: semantic model analyzers, Fabric Data Agent validator, natural-language remediation agent, Copilot diagnostics troubleshooter, DirectQuery tuning, and OneLake/lakehouse readiness checks.
- Use Copilot with semantic models
- Prepare your data for AI (Power BI)
- Best practices for configuring your data agent
Developed by Farhan Soomro (@SoomroFarhanH). These notebooks, scripts, and agents are free utilities for the wider Power BI and Microsoft Fabric community.
- License: the code is released under the MIT License. You can use, copy, modify, and share it, as long as you keep the copyright notice.
- Disclaimer: Provided "AS IS", without warranty of any kind, express or implied, including but not limited to the warranties of merchantability, fitness for a particular purpose, and non-infringement. This is a personal community project. It is not an official Microsoft product and is not supported or endorsed by Microsoft or any employer. Review the code and test it in a non-production environment before use. You use it at your own risk. In no event shall the author be liable for any claim, damages, data loss, service disruption, capacity or licensing cost, or other liability arising from its use. Results are guidance only and should be validated by a qualified person.
- Third-party content: the PDFs in Docs/ and at the repository root, and any linked Microsoft Learn content, belong to their respective owners and aren't covered by the MIT License. Code adapted from community sources is credited inline (for example, the Prep for AI retrieval adapted from Sandeep Pawar, Fabric.guru) and stays subject to its original terms.
- Trademarks: Microsoft, Power BI, Microsoft Fabric, Copilot, and related names are trademarks of the Microsoft group of companies. They're used here only to identify the products these tools work with.