ApexFlow is a lightweight, code-first workflow orchestration engine. Define durable, fault-tolerant Python pipelines with automatic step checkpointing, client-side zero-knowledge encryption, and instant observability—without managing complex brokers or database clusters.
From zero to running resilient, durable production workflows in four straightforward steps.
Sign up on the ApexFlow cloud control plane in seconds. No credit card, complicated clusters, or database installations required.
Head to Project Settings and generate your secure API key (af_live_...). Your workflows and workers are instantly tenant-isolated.
Install apex-flow-sdk via pip. Wrap your Python functions with @client.workflow and execute durable steps using ctx.step(...).
Track execution timelines in real time. Inspect step checkpoints, automatic retries, worker heartbeats, and client-decrypted payloads.
from apex_flow_sdk import ApexFlowClient, WorkflowContext
# 1. Initialize client with your project API key
client = ApexFlowClient(project_key="af_live_your_key_here")
# 2. Define workflow with durable, memoized steps
@client.workflow(name="customer-onboarding-pipeline")
async def onboard_customer(ctx: WorkflowContext, payload: dict):
# Step 1: Create billing account (memoized on completion)
account = await ctx.step(
"create-stripe-customer",
lambda: {"customer_id": f"cus_{payload['email'].split('@')[0]}"},
)
# Step 2: Provision cloud workspace (with automatic retries)
workspace = await ctx.step(
"provision-workspace",
lambda: {"workspace_url": f"https://app.example.com/{account['customer_id']}"},
max_attempts=3,
)
# Step 3: Send confirmation notification
await ctx.step(
"send-welcome-email",
lambda: {"email_sent": True, "to": payload["email"]},
)
return {"status": "active", "workspace": workspace}
# 3. Start worker
if __name__ == "__main__":
import asyncio
asyncio.run(client.start_worker())Everything required to orchestrate mission-critical workflows without the burden of managing complex distributed infrastructure.
Each step is memoized upon completion. If a 5-step pipeline encounters a network glitch at Step 4, restarting it skips Steps 1–3 and resumes instantly at Step 4.
Sensitive payloads are encrypted on your worker using authenticated AES-256-GCM before transmission. The control plane holds only ciphertext, decrypted client-side in your browser.
Connect workers over standard HTTPS REST. No Kafka clusters, Redis brokers, or Cassandra databases to deploy, maintain, or keep alive.
Payloads exceeding 250KB are transparently offloaded to Cloudflare R2 or S3 with lightweight pointer descriptors dereferenced automatically by the SDK.
Fine-grained step execution delays, run-level execution timeouts, integer priority queuing, and trigger event debouncing for rapid webhook bursts.
Branch execution from any prior step checkpoint (POST .../fork?from_step=N) to diagnose bugs and replay altered inputs without re-running entire pipelines.
The code-first ergonomics of DBOS, paired with a turnkey, zero-ops cloud control plane.
| Feature | ApexFlow ⚡ | DBOS Transact | Temporal.io | Celery + Redis |
|---|---|---|---|---|
| Developer Ergonomics | Code-First (Python) | Code-First (Python/TS) | Complex SDK / DSL | Decorator-based |
| Infrastructure Setup | Zero Ops (REST API Key) | Direct PostgreSQL Conn | Heavy (Clusters + ES) | Medium (Redis / RabbitMQ) |
| Durable Step Resumption | ✅ Yes (Step Memoization) | ✅ Yes | ✅ Yes | ❌ No (Whole task fails) |
| Zero-Knowledge Encryption | ✅ AES-256-GCM Client E2EE | ❌ No | ❌ Custom Data Converter | ❌ No |
| Turnkey Web Dashboard | ✅ Included (Next.js) | ❌ No native cloud UI | ✅ Heavy Web UI | ⚠️ Basic (Flower) |
| Large Payload Offload | ✅ Automatic R2/S3 (>250KB) | ❌ Manual | ⚠️ 2MB hard ceiling | ❌ Broker bloat |
| Time-Travel Forking | ✅ POST .../fork?from_step=N | ✅ Supported | ⚠️ Replay only | ❌ No |