CogFlow Platform

CogFlow Platform is the new self-hosted platform for the CogFlow workflow. It is intended to replace the current JATOS-centric deployment model with a deployable system that can run on infrastructure you control while preserving a fallback path to JATOS during migration.

The platform will eventually bundle:

• Researcher-facing study management
• Builder publishing without manual token copy/paste
• Interpreter runtime integration
• Result ingestion and storage
• Asset storage and delivery
• Audit, retention, and privacy controls

Current Status

This repository is the new implementation home for the platform migration. The immediate goal is to build a vertical slice that proves the end-to-end workflow:

1. Publish a study from Builder
2. Create or update the study record automatically
3. Show the study on a researcher dashboard
4. Launch Interpreter against the Django-backed runtime
5. Persist results and reflect run state in the platform

Documentation

• docs/ARCHITECTURE.md: system overview, core domain model, security principles, and migration structure
• docs/API.md: initial API surface and contract direction for publish, run start, and result submission
• docs/DEPLOYMENT.md: local, staging, and Kubernetes deployment model
• docs/ROADMAP.md: 4-6 week delivery plan toward a deployable pilot option

Intended Repository Layout

cogflow-platform/
├── backend/
├── frontend/
│   ├── builder/
│   ├── interpreter/
│   └── portal/
├── infra/
├── docs/
└── README.md

Implementation Priorities

Priority 1

• Stand up Django, PostgreSQL, and object storage locally
• Freeze the first API contracts
• Deliver the first vertical slice

Priority 2

• Introduce privacy-safe result handling
• Add dashboard workflow and study lifecycle management
• Deploy to staging on Kubernetes

Priority 3

• Harden operations, observability, and rollback procedures
• Expand researcher portal and compliance workflows

Migration Constraints

• Keep the JATOS fallback path available during transition
• Avoid redesigning Builder and Interpreter payload formats prematurely
• Defer final encryption schema details until the Django models and access patterns are concrete

Next Step

The next implementation step is to scaffold the platform runtime and deliver the Week 1 vertical slice described in docs/ROADMAP.md.