How openjiuwen-ai/dataswarm works
DataSwarm is an enterprise data-agent application for onboarding data sources and answering natural-language data questions with auditable results. The principal workflow starts with a user request, proceeds through the documented session/context/planning/execution analysis chain, and returns conclusions and artifacts. Distinct supporting capabilities include data-source onboarding, metadata and semantic/metric lookup, warehouse design/build, and report/dashboard output. The README and source index establish several links; unsampled runtime wiring is omitted rather than inferred.
Main entry points: analysis_flow.py Analysis chain · plan.py Warehouse plan · commands.py Flow commands · dw_client.py Data access client
Components 25
Analysis workflow
- Analysis chain
dataswarm/skills/dw-analysis/scripts/analysis_flow.py - Flow commands
dataswarm/skills/dw-analysis/scripts/flow/commands.py - Execution core
dataswarm/skills/dw-analysis/scripts/flow/execute_core.py - Trace observability
dataswarm/skills/dw-analysis/scripts/flow/observability.py - Data insight
dataswarm/skills/data-insight/scripts/analyzer.py
Data access
- Source onboarding
dataswarm/skills/dw-datasource-onboarding/scripts/onboard.py - Natural-language SQL
dataswarm/skills/nl2sql/scripts/execute_sql.py - SQL connection layer
dataswarm/skills/nl2sql/scripts/_db_client.py - Metric lifecycle/query
dataswarm/skills/dw-metrics/scripts/metric_query.py - Glossary and recall
dataswarm/skills/nl2sql/scripts/recall_tables.py
Warehouse engineering
- Warehouse plan
dataswarm/skills/dw-warehouse-design/scripts/plan.py - Source discovery
dataswarm/skills/dw-warehouse-design/scripts/discovery.py - Schema inference
dataswarm/skills/dw-warehouse-design/scripts/infer.py - Intent parsing
dataswarm/skills/dw-warehouse-design/scripts/intent.py - Semantic specification
dataswarm/skills/dw-warehouse-design/scripts/schema.py - Plan rendering
dataswarm/skills/dw-warehouse-design/scripts/render.py - Plan to tasks
dataswarm/skills/dw-warehouse-build/scripts/plan_to_tasks.py - Pipeline execution
dataswarm/skills/dw-warehouse-build/scripts/execute_pipeline.py
Analytical outputs
- Report resolver
dataswarm/skills/dw-report/scripts/resolver.py - Report rendering
dataswarm/skills/dw-report/scripts/render/page.py - BI dashboard
dataswarm/skills/dw-bi-dashboard/scripts/dashboard_tool.py
Shared runtime
- Data access client
dataswarm/skills/_dw_common/scripts/dw_client.py - Runtime paths
dataswarm/skills/_dw_common/scripts/dw_paths.py
Other components
- Analyst
- dwcli web API
Connections 25
- Analyst → Analysis chain: asks analysis
README.md - Analysis chain → Flow commands: routes flow
- Analysis chain → Execution core: executes analysis
- Analysis chain → Trace observability: traces steps
README.md - Analysis chain → Data insight: produces analysis
- Analysis chain → Report resolver: delivers artifacts
- Analyst → Source onboarding: onboards sources
README.md - Source onboarding → Data access client: accesses data
dataswarm/skills/dw-datasource-onboarding/scripts/onboard.py - Flow commands → dwcli web API: uses web API
README.md - Metric lifecycle/query → dwcli web API: calls metric-v2
dataswarm/skills/dw-metrics/scripts/metric_query.py - Natural-language SQL → SQL connection layer: uses connection layer
- SQL connection layer → Data access client: delegates data access
dataswarm/skills/nl2sql/scripts/_db_client.py - SQL connection layer → dwcli web API: queries service
dataswarm/skills/nl2sql/scripts/_db_client.py - Flow commands → Metric lifecycle/query: queries metrics
- Flow commands → Glossary and recall: recalls context
- Warehouse plan → Source discovery: discovers sources
dataswarm/skills/dw-warehouse-design/scripts/plan.py - Warehouse plan → Schema inference: infers schema
dataswarm/skills/dw-warehouse-design/scripts/plan.py - Warehouse plan → Intent parsing: uses intent
dataswarm/skills/dw-warehouse-design/scripts/plan.py - Warehouse plan → Semantic specification: validates specification
dataswarm/skills/dw-warehouse-design/scripts/plan.py - Warehouse plan → Plan rendering: renders plan
dataswarm/skills/dw-warehouse-design/scripts/plan.py - Plan to tasks → Pipeline execution: dispatches pipeline
- Report resolver → Data access client: reads data
dataswarm/skills/dw-report/scripts/resolver.py - Report resolver → Report rendering: prepares report
- Analyst → BI dashboard: explores dashboards
README.md - Data access client → Runtime paths: resolves runtime paths
dataswarm/skills/_dw_common/scripts/dw_client.py