mstr-robotics

A Python toolkit for automating MicroStrategy / Strategy One via its REST API: reading out project objects, comparing environments, driving migrations, running regression tests, and exporting dashboards to Open Semantic Interchange (OSI) YAML so BI content can be consumed by LLM agents.

Built against a real MicroStrategy estate and shared as-is. The API surface is not stable yet — expect changes between 0.x releases.

Highlights

OSI export. osi_exporter walks a MicroStrategy dossier and emits an OSI semantic model — datasets, fields, metrics and relationships — plus a dashboard document describing chapters, pages and visualizations. MicroStrategy's native visualization types are mapped onto neutral OSI ones (BarChart → bar_chart, KpiWidget → kpi_card, …), and add_ai_context() annotates any node with free-text context for downstream agents.

MCP servers. mstr_robotics.mcp_servers exposes the toolkit over the Model Context Protocol, so an assistant can find the right dashboard for a natural-language question, resolve objects by folder path, execute a report, and answer from the result together with its OSI context.

Environment comparison. json_compare diffs object definitions between two MicroStrategy environments with configurable path filtering and checksums — the basis for "what actually changed between dev and prod".

Getting started

Requires Python 3.12+ and network access to a MicroStrategy Library REST endpoint.

git clone https://github.com/magerdaniel/MSTR_Robotics.git
cd MSTR_Robotics
python -m venv .venv
.venv\Scripts\Activate.ps1        # Windows;  source .venv/bin/activate on Unix
pip install -e .

Then two steps, both covered in detail by docs/SETUP.md:

1. Deploy the Object Manager packages from Object_Manager_Packages/ — they create the cubes, reports and dossiers the notebooks read from.
2. Run notebooks/00_setup.ipynb top to bottom. It creates the output folders, copies the config templates, checks what you filled in, verifies the connection, and resolves your environment's object GUIDs automatically.

That last part is the one worth knowing about: deploying the packages creates objects that MicroStrategy assigns new GUIDs, but their names are fixed — so the setup notebook searches your project by name and writes the right GUIDs into config/jupyter_objects_d.yml instead of you hunting ~30 of them by hand.

Install extras

The base install stays small on purpose. Add only what you need:

pip install -e ".[rag]"       # OpenAI/Perplexity, FAISS, LangChain — chat & RAG notebooks
pip install -e ".[redis]"     # Redis-backed metadata analysis
pip install -e ".[azure]"     # Azure Blob staging for migrations
pip install -e ".[servers]"   # MCP servers
pip install -e ".[all]"       # everything above
pip install -e ".[dev]"       # ruff, vulture, jupyter

Paths

All data locations resolve relative to the repo by default and can be repointed with environment variables:

Variable	Default	Holds
MSTR_REPO_ROOT	auto-detected from config/user_d.yml	repo root
MSTR_OSI_DIR	<repo>/data/osi	generated OSI YAML
MSTR_OSI_SCHEMA_DIR	MSTR_OSI_DIR	osi-schema-with-dashboards.json
MSTR_OUTPUT_DIR	<repo>/output	exports, logs, MCP data

Module map

Module	Purpose
_connectors	MstrApi — REST session handling
read_out_prj_obj	Read schema, facts, attributes, prompts, reports, cubes
osi_exporter	Build OSI semantic models and dashboard documents
json_compare	Diff object definitions across environments
select_mig_objects	Change-log driven migration package building
regam	Regression testing against Platform Analytics data
cube_load	Parallel cube publish with follow-up chains
report, dossier, navigation	Report/dossier execution and prompt answering
user_rag	FAISS vector store, OpenAI and Perplexity clients
redis_db	Redis-backed BI metadata analysis
prepare_ai_data	Normalize MSTR metadata JSON for AI consumption
mcp_servers	MCP tool groups over the above
setup	First-run helpers driven by 00_setup.ipynb

Cube load chains

cube_load.handle_cube_load(conn, execution_list) publishes cubes in parallel and can trigger a follow-up cube once one finishes. Each entry takes a project_id, a cube_id, run_any_time, and an optional follow_up — see mstr_robotics/cube_load_sample.json for the shape. Running the module directly executes the demo chain in its __main__ block:

python -m mstr_robotics.cube_load

Notebooks

Start with 00_setup.ipynb — it configures everything else. After that:

Notebook	Does
jup_prj_obj_exporter.ipynb	Read objects out of a project
jup_schema_monitor.ipynb	Monitor schema changes over time
jup_migrate.ipynb	Build and run migration packages
jup_REGAM.ipynb	Regression testing against Platform Analytics
jup_load_rag_cubes.ipynb	Export metadata cubes for RAG
jup_chat_answer_prompt_page.ipynb	Answer BI questions over exported context
jup_mstr_admin.ipynb	Administrative queries

Object IDs come from config/jupyter_objects_d.yml rather than being hardcoded, so the same notebooks run against any environment. Outputs are stripped before commit.

License

See LICENSE.