Integrations¶
These integrations extend Consist for container execution, data ingestion, and configuration management.
When to Use Each Integration¶
| Integration | Use when |
|---|---|
| Container | Wrapping existing tools (ActivitySim, SUMO, BEAM) without source modification |
| DLT Loader | Cross-run SQL queries on 100K+ row datasets with schema validation |
| Config Adapters | Tracking file-based configs (YAML/HOCON hierarchies) with queryable parameters |
Container Integration¶
Execute Docker and Singularity containers with provenance tracking.
- Image digest-based caching for reproducibility
- Volume mounting and output capture
- Full guide | API reference
DLT Loader¶
Schema-validated data ingestion with DuckDB.
- Automatic provenance column injection (
consist_run_id, etc.) - Cross-run SQL queries with registered schemas
- Full guide | API reference
Config Adapters¶
Model-specific configuration discovery and tracking.
- Automatic discovery of YAML/HOCON hierarchies
- Queryable configuration tables for sensitivity analysis
- GTFS support for raw transit feed members and selected-service slices
- Full guide | ActivitySim | BEAM | GTFS
GTFS is available here because it is a standardized feed format with queryable tables and service-date identity rather than a bespoke model config tree. Start with the GTFS guide if you only need the transit story; use the BEAM guide when you want the full BEAM-specific wiring.
Building a Domain Tracker¶
For teams that run the same model repeatedly with different configurations, wrapping Consist in a domain-specific class reduces per-callsite boilerplate and centralizes identity wiring.