You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[Feature]: Add a sub-agents section for parallel task execution #12
As a Diffex user, I need a first-class way to delegate independent or specialized work to sub-agents, because a single agent handling every task serially limits throughput and makes task offloading difficult.
Desired outcome
Diffex provides a sub-agents section that lets users create and manage delegated tasks, allows the primary agent to communicate with sub-agents, and supports parallel execution while preserving a coherent parent-task context.
Success criteria
A user can create or request sub-agent tasks from a dedicated sub-agents section.
Diffex can dispatch independent tasks to multiple sub-agents and run them in parallel.
The primary agent can send instructions to sub-agents and receive status, results, errors, and follow-up messages.
Sub-agent work is clearly associated with the parent task and can be inspected without losing context.
Users can see the task state and completion or failure status for each sub-agent.
Tests demonstrate safe task isolation, message routing, result aggregation, and bounded concurrency.
Affected area
Multiple areas
Proposed approach
Add a sub-agents section to the interaction and session UI and the underlying orchestration model.
Represent each sub-agent task with a parent task, lifecycle state, input, output, and message channel.
Let the primary agent create tasks with explicit scopes and capabilities, schedule independent tasks concurrently, and exchange messages with running sub-agents.
Aggregate results back into the parent context with clear provenance and failure handling.
Make concurrency limits, cancellation, and permission boundaries explicit and configurable.
Alternatives considered
Using separate Diffex sessions manually or doing all work in one agent context does not provide coordinated communication, shared lifecycle visibility, or reliable parallel execution.
Additional context
This should be designed as a durable orchestration capability rather than a UI-only feature.
It should support multiple sub-agent backends or providers later and preserve observability and security boundaries.
Pre-submission checks
I searched existing issues and did not find a duplicate request.
I described the need and desired outcome, not only a preferred implementation.
I did not include secrets, credentials, or private deployment details.
Activity