AutoSwarm-Shadow-Adapter
Zero-modification Human-in-the-Loop adapter for OpenOPC's agentic DAG runtime.
- agentic-ai
- agentic-workflow
- ai-employee
- ai-native
- ai-workforce
- dag-orchestration
- external-agent-adapter
- fastapi
- generative-ai
- hitl
- human-in-the-loop
- human-oversight
- llm
- llm-orchestration
- multi-agent
- openopc
- orchestration
- python
- react
- shadow-mode
- Stars
- 2,152
- Forks
- 2
- + today
- +4
- Created
- 2mo
Ranking data as of October 21, 2026 (UTC).
Star History
Today, hour by hour
Overview
Zero-modification Human-in-the-Loop adapter for OpenOPC's agentic DAG runtime. It ranks #186 on GitTiger, gaining +4 stars on October 21, 2026 (UTC).
The project is written in Python and has 2 forks. It was created 2mo ago.
What it is
The AutoSwarm Shadow Adapter is described as a zero-modification human-in-the-loop adapter for OpenOPC's agentic DAG runtime. The point of "zero-modification" is that the DAG runtime itself does not need patching to gain a human approval step.
Capabilities explicitly described
- Non-blocking intercept: task execution is intercepted in under 50ms, the task is parked in a separate SQLite WAL database, and the worker threads are freed at once with an `AWAITINGHUMAN` state. - Contractor and Silicon Worker Portal: a React 19 interface plus a REST API (`shadow-serve`) serving reviewers and remote workers (`shadow-worker`). - A Consolidated Corporate Brain that gathers deliverables from ancestor tasks, structured markdown briefs (`briefmd`) and any uploaded file artifacts into one package.
Topics line up with the description — `hitl`, `human-oversight`, `dag-orchestration`, `external-agent-adapter`, `fastapi`. The project is written in Python and released under MIT.
Getting started
No installation or configuration steps are present in the supplied evidence, so none are offered here. The service names `shadow-serve` and `shadow-worker` appear in the preview; consult the repository README for how they are launched and wired to a running AutoSwarm instance.
git clone https://github.com/AhmadHassan-BTed/AutoSwarm-Shadow-Adapter.git
cd AutoSwarm-Shadow-Adapter
# see README for setupFrequently Asked Questions
Does this require changes to AutoSwarm?
No. The description calls it a zero-modification adapter for OpenOPC's agentic DAG runtime.
What happens to a task while it waits for a human?
It is intercepted in under 50ms, parked in an isolated SQLite WAL database, and the execution threads are released with an `AWAITINGHUMAN` state.
Is there a UI for reviewers?
Yes: a React 19 interface and a REST API (`shadow-serve`) for reviewers and remote workers (`shadow-worker`).
What do I need to run it?
The evidence does not include setup steps. Check the repository README; the stack is Python plus a React frontend and it is MIT licensed.