Analysis Views
Domain pages read the latest completed job for the selected configuration. They are chart UIs on consolidated_results plus shortcuts into the matching markdown report.
Analysis Insights (/analysis/:jobId)
From Jobs → open a completed job. Interactive panels (distributions, correlations, module-specific charts). Full Report jumps to /reports.
Analytics (/analytics)
Operations only: job volume, completed/failed, success rate, config usage, 1–90 day window. Not Financial ARPU.
Time Series (/time-series)
For Analytic configs with time series enabled:
- Datetime column and grain (day, week, …)
- Numeric measures (sum/mean, rolling, lags)
- Optional journeys (entity + event columns)
- Event association / correlation windows
- Forecast outputs when generated
Save on this page updates the configuration; Run still goes through Jobs. Primary report: Time series (time_series_markdown).
Interactions (/recommender)
Dashboard for Interactions jobs (latest completed job for the selected config). Tabs: General, Offers / Catalog, Customers, Market intelligence, Channels, Experimentation, Marketing mix, Data Quality.
Channels and experimentation tabs read the general and offer payloads. Full narrative is in the markdown reports:
- Campaign & predictor (
recommender_general_markdown) - Offers & catalog
- Customer uptake
How to configure and what each family means: Interactions Analytic.
Personality (/personality)
Tabs aligned with families: General, Time, Customer Change, Spend, Money, Drift, Category Patterns, Calendar, Recommendations, Data Quality. Primary report: Personality overview (personality_general_markdown).
This analyzes scored or raw transaction history. Live trait scoring in the closed loop is Spend Personality.
Financial (/financial)
Unit economics, returns, value, growth, ledgers, industry pack — whatever families ran. Primary report: Financial overview (financial_general_markdown). Unmapped ROI families stay empty.
Agents (/agents)
Catalog of CRM/CVM specialists and MCP tools (list jobs, get_report, …). Test console: choose agent, optional config_id, ask in natural language. Mutating tools follow the product’s confirm rules. This is not EcoGentic customer chat and not Runtime builtin MCP (POST /mcp on the Runtime).
Local run
# ecosystem-data-agents repo
./run.sh # API :54324
./run_frontend.sh # UI :54325Python 3.14. Set MONGO_CONNECT. Docker compose publishes 54325 (nginx + API) on the ecosystem network — do not also bind ./run_frontend.sh on the same port.