Data Flow

Purpose

This page documents how information moves through the system — from user request to final result — across the Orchestration Agent, Code Execution Agent, Research Agent, and Media Agent subsystems.

Status: Implemented (primary flows are established and operational)

Overview diagram

Loading diagram…

Development flow (Orchestration → Code Execution)

When a development task is initiated:

1. The user states a goal or request to the Orchestration Agent.
2. The Orchestration Agent verifies the project context and confirms write permissions.
3. The Orchestration Agent designs the solution — architecture, approach, constraints.
4. The Orchestration Agent writes a scoped implementation prompt for the Code Execution Agent.
5. The Code Execution Agent executes the prompt and returns generated code.
6. The Orchestration Agent reviews the output, integrates it, and tests it.
7. The user performs git commit and push manually.
8. The Orchestration Agent reports the result and updates documentation.

Research flow (Orchestration → Research Agent)

When a research task is initiated:

1. The Orchestration Agent (or user) writes a task description to the Research Agent's task queue.
2. The Research Agent's Mission Runner picks up the task file.
3. The Research Agent creates an isolated run directory for the mission.
4. The Research Agent executes research phases in DAG order, storing intermediate artifacts at each phase.
5. The final phase aggregates all artifacts into final_report.md.
6. The report is returned to the user or consumed by the Orchestration Agent.

See Research Mission DAG for phase dependency details.

Media flow (Orchestration → Media Agent)

When a media request is initiated:

1. The user makes a media request to the Orchestration Agent.
2. The Orchestration Agent delegates the request to the Media Agent.
3. The Media Agent searches for matching titles and evaluates availability.
4. The Media Agent initiates playback to the target connected device.

The Media Agent operates as an independent branch — its flow does not interact with the Research Agent or Code Execution Agent.

Result handoff

Orchestration Agent results:

• Generated code files ready for review and integration
• Documentation updates written to relevant files
• Verbal summaries or explanations

Research Agent results:

• final_report.md — structured Markdown report in the mission run directory
• Intermediate artifact files available for inspection or debugging

Media Agent results:

• Initiated playback on the connected device

A current limitation

The Research Agent and the Orchestration Agent flows are currently independent. There is no automated pipeline where Research Agent output is directly consumed as structured input to an Orchestration Agent task. This handoff is currently manual. Automated inter-agent data passing is planned.

Related pages

→ System Overview — component roles → Code Execution Agent — implementation workflow details → Research Framework — Research Agent pipeline details → Research Mission DAG — phase dependency graph → Media Agent — media automation details