Skip to content

CausalIQ Workflow - Development Roadmap

Last updated: March 04, 2026

This project roadmap fits into the overall ecosystem roadmap

🚧 Under development

Release 0.4.0 - Conservative Execution

Formalise action patterns and implement conservative execution.

Scope:

  • Action pattern validation:
  • Creation: output required, matrix required, input prohibited (for caches)
  • Update: input required, output prohibited, matrix prohibited

  • Aggregation: input required, output required, matrix required

  • Update action support - Actions that modify input cache entries

  • Add metadata sections to existing entries

  • Add objects to existing entries

  • Conservative execution - Skip work if results exist

  • Creation: skip if entry with matching matrix values exists in output
  • Update: skip if action metadata section exists in entry
  • Aggregation: skip if entry with matching matrix values exists in output
  • Add --mode=force option to bypass checks
  • All logic implemented in workflow executor (actions unaware)

✅ Previous Releases

  • v0.1.0 Workflow Foundations [February 2026]: Framework for plug-in actions, basic workflow and CLI support

  • v0.2.0 Knowledge Workflows [February 2026]: Include LLM graph generation in workflows and store results in Workflow caches.

  • v0.3.0 Aggregation Workflows [March 2026]: Matrix-driven aggregation processing for multi-source workflows.

🛣️ Upcoming Implementation

Release 0.5.0 - Step Output Chaining

Enable workflow steps to consume outputs from previous steps.

Scope:

  • Step output references - Template syntax {{steps.<name>.outputs.<key>}}
  • Extend _resolve_template_variables() to handle step output references
  • Track step outputs in WorkflowContext (add step_outputs: Dict[str, Any])

  • Deserialise GraphML strings back to graph objects when consumed

  • Cache restoration - Resume workflows from cached results

  • Check cache before executing step
  • Support forced re-execution flag

Release 0.6.0: Enhanced Workflow

Dry and comparison runs, runtime estimation and processing summary

Scope:

  • dry-run capability
  • standardise message format
  • support skip, would do etc messages
  • support comparison (integration test) functionality
  • processing summary
  • estimate runtime
  • progress indicators

Release 0.7.0: Discovery Integration

Structure learning algorithms integrated

Scope:

  • causaliq-discovery algorithms integrated
  • timeout supported

🚀 Possible Future Features

External Algorithm Integration (After robust test infrastructure):

  • Multi-language workflows (R bnlearn, Java Tetrad, Python causal-learn)
  • External CausalIQ package integration (discovery, analysis)
  • Matrix-driven algorithm comparisons across datasets
  • Automatic dataset download and preprocessing

Production Features:**

  • 📋 Workflow queuing - CI-style runner management
  • 📊 Monitoring dashboard - Real-time execution tracking
  • 🗺 Artifacts & caching - Persistent storage, result reuse
  • 🔒 Security & isolation - Secrets management, containers
  • 📈 Performance optimization - Resource limits, scheduling

Research Platform:

  • 🤖 LLM integration - Model averaging, hypothesis generation
  • 🌐 Web interface - Browser-based workflow designer
  • 🚀 Cloud deployment - AWS/GCP/Azure runners
  • 👥 Collaboration - Multi-researcher workflows
  • 📚 Publication workflows - Reproducible research outputs

Advanced Capabilities:

  • Workflow marketplace - Sharing and discovering research workflow templates
  • Interactive notebooks - Jupyter integration with workflow execution
  • Multi-machine execution - Distributed workflows across compute clusters
  • AI-assisted optimization - Automated hyperparameter and workflow tuning
  • Integration ecosystem - Plugins for major research tools and platforms

This roadmap leverages Git commit history for completed work, provides detailed release-based planning for upcoming functionality, and outlines future possibilities.