Sapiom deploy

Sapiom Deploy

Deploy and run code via Sapiom. ZIP your code, POST it, done.

When to Use

• User says "deploy this" / "ship it" / "put this in the cloud"
• User wants to run a script on a schedule
• User wants to process N items in parallel
• User wants a persistent environment with ports/filesystem
• User already has working code and wants it deployed

Building an agent from scratch? Use the sapiom-agent-builder skill instead — it handles the full workflow from intent to deploy.

Deploy Patterns (pick one)

Pattern	What It Does	When to Use	Reference
Scheduled job	Runs on cron	Periodic tasks (monitoring, scraping, reports)	references/jobs.md
Batch job	N parallel tasks on demand	Process a list of items in parallel	references/jobs.md
Orchestrator + workers	Scheduled job fans out to batch	Periodic task that processes N items	references/patterns.md
Sandbox	Persistent instance with ports + filesystem	Dev environments, servers, long-running processes	references/sandboxes.md

Quick Decision

• "Run X every hour" → Scheduled job
• "Process these 50 items" → Batch job
• "Every hour, check for new items, process each" → Orchestrator + workers
• "Spin up a server" / "I need a dev environment" → Sandbox

Setup (same for all patterns)

const { createFetch } = require("@sapiom/fetch");
const safeFetch = createFetch({
  apiKey: process.env.SAPIOM_API_KEY,
});
const GATEWAY = "https://blaxel.services.sapiom.ai";

Example: Deploy a Script on a Schedule

User says: "Deploy this script to run every 2 hours"

Steps:

1. Read references/jobs.md for the full API
2. ZIP index.js + package.json (files at ZIP root, not in a subdirectory)
3. POST /v1/jobs?name=my-job&schedule=0 */2 * * * with ZIP body
4. Response returns { status: "deployed" } — deploy is synchronous, no polling

Result: Deploy returns immediately with status: "building". Poll GET /v1/jobs/my-job every 10–15s until "deployed" (~60–120s). Then the job runs every 2 hours.

Example: Fan Out to Parallel Workers

User says: "Process these 100 URLs in parallel"

Steps:

1. Deploy worker code as a batch job (no schedule): POST /v1/jobs?name=my-worker with ZIP body
2. Trigger execution: POST /v1/jobs/my-worker/executions with { tasks: [{ url: "..." }, ...] }
3. Each task runs in parallel, receives its payload via blStartJob(async (args) => { ... })

Result: 100 tasks run concurrently. Check progress with GET /v1/jobs/my-worker/executions/{id}.

Troubleshooting

Deploy returns 502

Cause: Bad package.json — missing main, scripts.start, or invalid JSON. Fix: Ensure package.json has "main": "index.js" and "scripts": { "start": "node index.js" }.

Deploy returns 402

Cause: Missing or invalid SAPIOM_API_KEY. Fix: Ensure the API key is valid and passed via the @sapiom/fetch SDK.

Job deploys but fails at runtime

Cause: Code works locally but crashes in the cloud — usually missing env vars. Fix: Pass SAPIOM_API_KEY and any other env vars via the envs field in the JSON deploy body.

ZIP files not found by builder

Cause: Files nested in a subdirectory inside the ZIP (e.g., my-project/index.js instead of index.js). Fix: ZIP from inside the directory: cd my-project && zip -r ../deploy.zip .

Job name conflict

Cause: A job with that name already exists. Fix: Use PUT /v1/jobs/{name} to update, or DELETE then recreate.

References

• references/jobs.md — Jobs API (create, deploy, execute, manage)
• references/sandboxes.md — Sandboxes API (create, filesystem, processes, ports)
• references/patterns.md — Orchestrator + workers pattern with concrete example