The orchestration runtime for AI workflows.
pip install plexrun (coming soon — private beta)
from plexrun import workflow, step
@workflow
def research_pipeline(topic: str):
draft = summarize(topic)
return write_report(draft)
@step(model="gpt-4o", retries=3)
def summarize(topic: str) -> str:
...
@step(model="gpt-4o")
def write_report(draft: str) -> str:
...
## Features
- **DAG-based execution** — Define step dependencies, fan-out, fan-in
- **Idempotent retries** — Failed steps retry without duplicate LLM calls
- **Per-step traces** — Token usage, cost, and latency for every step
- **Serverless workers** — No fleet to manage, auto-scales with queue depth
- **YAML workflows** — Declarative pipeline definitions supported
## CLI
```bash
# Install CLI
pip install plexrun
# Initialize a new project
plexrun init my-pipeline
# Deploy workflow
plexrun deploy workflow.py
# Trigger a run
plexrun run research_pipeline --input '{"topic": "AI in healthcare"}'
# View live logs
plexrun logs <run-id>
## Status
| Component | Status |
|----------------|-------------|
| Python SDK | Alpha |
| CLI | Alpha |
| TypeScript SDK | In progress |
| Dashboard | In progress |
Currently in private beta. [Join the waitlist →](https://plexrun.com/#waitlist)
Links
Website: plexrun.com
Docs: plexrun.com/docs/quickstart
Contact: hello@plexrun.com