Temporal + Modal

This repository contains an example of how to integrate Temporal and Modal using Sandboxes. We create a Modal App that launches Temporal Workers. These workers serve as the compute layer for running Temporal Workloads.

Usage

Step 1: set local env vars

Rename .env.example to .env. Then populate .env.example with required environment variableS:

TEMPORAL_SERVER_URL=
TEMPORAL_API_KEY=
TEMPORAL_NAMESPACE=

Then make them available in your environment with:

export $(cat .env)

Step 2: create Modal Secret

Now create a Modal Secret called temporal-secrets with those three environment variable.

Step 3: sync files

Create Volume and sync workflow, activity, and config files with:

python sync_files.py

This will store files from the following directories into a Modal Volume:

• /workflows: defines Temporal Workflows
• /activities: defines Temporal Activities (aka functions)
• /configs: groups Workflows and Activities in a config

Configs are designed to configure workers to start only with a set of workflows.

Step 4: deploy Modal app

The Modal App contains (A) a function to build images using conda to manage environments and (B) a scheduled Modal Function that builds images automatically (this is designed to observe a database and build images that match a conda environment).

modal deploy deploy.py

Step 5: run example

Run the example script with:

python run_example.py

This will

1. Build an image, using conda to install dependencies
2. Start a Temporal Worker with the image ID above
3. Trigger a Temporal Workflow, which runs in the Worker above