Cognos to Power BI Migration Tool

Migrate IBM Cognos reports, models, data modules, and dashboards to Microsoft Power BI automatically. An open-source, AI-assisted migration engine that converts Cognos sources into Power BI Project (PBIP) format using TMDL semantic models and PBIR report definitions.

What this project does

Organizations moving from IBM Cognos Analytics to Microsoft Power BI face a slow, manual, error-prone migration. This tool automates the heavy lifting:

1. Parse Cognos report specification XML, Framework Manager models, and data modules.
2. Normalize them into a vendor-neutral intermediate representation (IR).
3. Model the result as a Power BI star schema: classify fact and dimension tables, orient relationships, infer cardinality and cross-filter direction, and mark date tables.
4. Generate Power BI Project (PBIP) output: TMDL semantic models and PBIR reports.
5. Refine the output with an AI assistant (Claude, GitHub Copilot, or Codex) to translate expressions, layouts, and visuals that have no direct mechanical mapping.

The result is a Git-friendly Power BI project you can open in Power BI Desktop, review, and deploy to the Power BI / Microsoft Fabric service.

Why use it

• Faster migrations. Convert hundreds of reports in a fraction of the manual time.
• Consistent output. Deterministic, reviewable PBIP artifacts instead of hand-rebuilt reports.
• AI-assisted, not AI-dependent. The mechanical conversion works without any AI key; AI only refines what cannot be mapped deterministically, and never at the cost of a loadable model.
• Provider-agnostic AI. Use Azure OpenAI (Entra ID), Claude Code CLI, GitHub Copilot CLI, or OpenAI Codex CLI.
• Open and extensible. MIT-licensed, plugin-friendly parsers and generators.

Supported conversions

The star-schema modeling pass classifies fact, dimension, and date tables, hides foreign keys, resolves role-playing dimensions and ambiguous filter loops, and flags many-to-many and snowflake joins for review. Disable it with --no-infer-model. See data modeling for the full behavior and edge-case handling.

See the migration coverage matrix for detail on expressions, filters, and visual types.

Quick start

# Install
pip install cognos2powerbi

# Convert a single Cognos report specification to a Power BI project
cognos2pbi migrate ./examples/sample_report.xml --out ./out/SalesReport

# Convert a Cognos report to a Report Builder paginated report (.rdl)
cognos2pbi migrate ./examples/sample_report.xml --out ./out/SalesReport --format rdl

# Point the generated model at your database (refreshable PBIP)
cognos2pbi migrate ./examples/sample_report.xml --out ./out/SalesReport \
  --server sql01.contoso.com --database Sales

# Convert a Cognos Framework Manager model to a Power BI semantic model
cognos2pbi migrate-model ./examples/sample_model.xml --out ./out/SalesModel

# Infer a full star schema (fact/dimension roles, relationships, date tables)
cognos2pbi migrate-model ./examples/star_schema_model.xml --out ./out/RetailStar

# Convert a Cognos data module to a Power BI semantic model
cognos2pbi migrate-module ./examples/sample_data_module.json --out ./out/SalesAnalysis

# Convert a Cognos dashboard to a Power BI report
cognos2pbi migrate-dashboard ./examples/sample_dashboard.json --out ./out/SalesDashboard

# Convert a whole folder of mixed sources, with a coverage report
cognos2pbi migrate-batch ./examples --out ./out/Batch

# Open the generated project
#   ./out/SalesReport/SalesReport.pbip   ->  open in Power BI Desktop

Run from source instead:

git clone https://github.com/navintkr/cognos-to-powerbi.git
cd cognos-to-powerbi
pip install -e ".[dev]"
cognos2pbi migrate ./examples/sample_report.xml --out ./out/SalesReport

End-to-end demo (complex report on local SQL Server)

This walkthrough migrates a complex Cognos report, wires the generated model to a local SQL Server database, and opens a refreshable Power BI Project. Everything it needs ships in the repo:

• examples/complex_sales_report.xml - a Cognos report with three queries (FactSales, DimProduct, DimDate), typed columns, six aggregate measures (including translated arithmetic such as Gross Profit and Net Revenue), and three pages of list, crosstab, column, line, and pie visuals.
• examples/sql/setup_demo_db.sql - creates the matching CognosDemo database with sample data. Table and column names line up with the report, so the generated PBIP refreshes with no manual edits.

Step-by-step video

Walkthrough video: add your recording link here (for example a Loom or YouTube URL). Suggested flow to record: run the SQL script, run cognos2pbi migrate, open the .pbip in Power BI Desktop, set the Server / Database parameters, then Refresh.

1. Install the demo database

Requires a local SQL Server instance and sqlcmd. Using a default instance with Windows auth:

sqlcmd -S localhost -E -C -i examples/sql/setup_demo_db.sql

This creates CognosDemo with DimProduct (6 rows), DimDate (6 rows), and FactSales (12 rows).

2. Migrate the report and wire it to local SQL

cognos2pbi migrate examples/complex_sales_report.xml --out out/ComplexDemo `
  --source-type sqlserver --server localhost --database CognosDemo --schema dbo

Expected summary: 3 tables, 6 measures, 3 pages, 0 items to review. The generated model uses Server and Database parameters and a Sql.Database(Server, Database) partition per table.

3. Open and refresh in Power BI Desktop

1. Open out/ComplexDemo/complex_sales_report.pbip in Power BI Desktop.

2. If prompted, confirm the Server (localhost) and Database (CognosDemo) parameters.

3. Refresh. The measures evaluate against live data, for example:

   Total Revenue	Gross Profit	Net Revenue	Avg Unit Price	Total Quantity
   1,751,000.00	663,000.00	1,663,450.00	741.67	3,810

To point at a different server, change --server / --database, or edit the Server and Database parameters in Power BI Desktop after opening the project.

AI-assisted refinement (optional)

Enable an AI provider to translate complex Cognos expressions (calculated columns and measures) into Power BI DAX. Azure OpenAI runs over HTTPS with Microsoft Entra ID; the others shell out to their CLI.

# Azure OpenAI (Entra ID via az login; install extras first)
pip install "cognos2powerbi[azure]"
cognos2pbi migrate ./report.xml --out ./out/Report --ai azure

# Claude Code CLI
cognos2pbi migrate ./report.xml --out ./out/Report --ai claude

# GitHub Copilot CLI
cognos2pbi migrate ./report.xml --out ./out/Report --ai copilot

# OpenAI Codex CLI
cognos2pbi migrate ./report.xml --out ./out/Report --ai codex

If no AI provider is configured, the tool completes a deterministic conversion and flags items that need manual review. Either way the generated model stays loadable: an expression that cannot be translated remains a physical column and is listed in MIGRATION_REVIEW.md rather than emitted as invalid DAX. See docs/ai-providers.md.

How it works

flowchart LR
    A["Cognos artifacts<br/>report XML, FM model, data module, dashboard"] --> B["Parsers"]
    B --> C["Intermediate Representation<br/>vendor-neutral IR"]
    C --> D["PBIP Generators<br/>TMDL and PBIR"]
    D --> E["AI Refinement<br/>Claude, Copilot, Codex"]
    E --> F["Power BI Project<br/>.pbip"]

The intermediate representation decouples parsing from generation, so new Cognos inputs and new Power BI output formats can be added independently. See docs/architecture.md.

Project layout

cognos-to-powerbi/
├── src/cognos2powerbi/
│   ├── cli.py                 # Command-line interface
│   ├── core/
│   │   ├── ir/                # Vendor-neutral intermediate representation
│   │   ├── parsers/           # Cognos report/model parsers
│   │   ├── generators/        # PBIP (TMDL + PBIR) generators
│   │   ├── translate/         # Deterministic Cognos-to-DAX translation
│   │   ├── ai/                # Provider-agnostic AI adapter
│   │   └── pipeline.py        # Orchestration
│   └── api/                   # FastAPI backend (SaaS surface)
├── web/                       # Single-page web frontend
├── examples/                  # Sample Cognos inputs
├── docs/                      # Documentation
└── tests/                     # Test suite

Run the SaaS API locally

pip install -e ".[api]"
uvicorn cognos2powerbi.api.main:app --reload
# Open the web UI at http://127.0.0.1:8000/
# Auto-detects the source kind (report, model, module, dashboard) on upload.
#   POST a single source to http://127.0.0.1:8000/api/v1/migrate   (returns a PBIP zip)
#   POST a single source to http://127.0.0.1:8000/api/v1/analyze   (returns JSON review items)
#   POST many sources to    http://127.0.0.1:8000/api/v1/batch     (returns a zip + coverage report)

Roadmap

• Framework Manager model conversion to TMDL
• Expression translation library (Cognos to DAX)
• Parameterized data-source wiring for refreshable PBIP
• Data module conversion
• Dashboard to PBIR page mapping
• Hosted SaaS portal with upload, review, and download
• Batch / folder migration with a coverage report

Full roadmap: docs/roadmap.md.

Contributing

Contributions are welcome and wanted. Good first issues are labeled good first issue. Read CONTRIBUTING.md and the Code of Conduct to get started.

Security

Report vulnerabilities privately per our security policy. Do not open public issues for security reports.

License

Licensed under the MIT License.

Keywords

Cognos to Power BI, Cognos Power BI migration, IBM Cognos migration tool, convert Cognos reports to Power BI, Cognos report converter, Power BI PBIP TMDL PBIR generator, business intelligence migration, automated report migration.