How gamakon/ultradim works
UltraDim is an analytical vector store; its principal workflow is to ingest vectors, build searchable structures, and run analytical queries such as exact nearest ranking, mapping, clustering, and classification. The repository exposes the engine through an MCP tool server and includes a substantial BloomMap product branch that turns hierarchical clusterings into data for poster and interactive visualizations. The engine implementation itself is not present in the sampled tree, so its internal layers and most MCP-to-engine calls cannot be mapped here; those links are documented at the product level.
Main entry points: bloommap_build.py Shared map builder · visualization.js Poster renderer · ultradim_mcp_server_v0_5_0.py MCP tools · hseg_chembl.py ChEMBL segmentation
Components 14
MCP interface
- MCP tools
mcp/ultradim_mcp_server_v0_5_0.py
UltraDim engine
- Vector analytics
BloomMap production
- ChEMBL segmentation
BloomMap/process/hseg_chembl.py - Movie segmentation
BloomMap/process/hseg_ml20m_movies.py - ChEMBL map data
BloomMap/process/build_bloommap_json_chembl.py - Movie map data
BloomMap/process/build_bloommap_json_ml20m.py - Shared map builder
BloomMap/process/bloommap_build.py - Galaxy enrichment
BloomMap/galaxy/build_galaxy.py
BloomMap rendering
- Poster renderer
BloomMap/renderer/visualization.js - Paper profiles
BloomMap/renderer/config/paper_profiles.js - SVG export
BloomMap/renderer/render.mjs - Sunburst export
BloomMap/galaxy/build_sunburst_svg.py
Other components
- AI assistant
- Analyst
Connections 14
- AI assistant → MCP tools: invokes tools
README.md - MCP tools → Vector analytics: wraps engine
README.md - Analyst → Vector analytics: runs analyses
- Vector analytics → ChEMBL segmentation: clusters vectors
- Vector analytics → Movie segmentation: clusters vectors
- ChEMBL segmentation → ChEMBL map data: writes tree inputs
- Movie segmentation → Movie map data: writes tree inputs
- ChEMBL map data → Shared map builder: uses shared logic
BloomMap/process/build_bloommap_json_chembl.py - Movie map data → Shared map builder: uses shared logic
BloomMap/process/build_bloommap_json_ml20m.py - Shared map builder → Poster renderer: supplies map data
- Shared map builder → Galaxy enrichment: supplies tree data
BloomMap/galaxy/build_galaxy.py - Poster renderer → Paper profiles: uses layout profiles
BloomMap/renderer/visualization.js - SVG export → Poster renderer: loads HTML view
BloomMap/renderer/render.mjs - Galaxy enrichment → Sunburst export: supplies hierarchy