Backend & Data

Scheduled Agents are AI-generated Python scripts that run on a cron schedule in the cloud. Describe what you want the agent to do — GenMB generates the code, deploys it to a shared GKE worker pool, and runs it on your configured schedule. Each run is logged with output, status, duration, and trigger type.

Creating a Scheduled Agent takes a few minutes. Describe what you want it to do, configure the schedule, add any secrets, and deploy.

Every agent is a Python script with a single entry point: async def run(). GenMB generates this code based on your description and deploys it as-is. You can open the linked app in the code editor to view or modify it at any time.

import httpx, os

async def run():
    api_key = os.environ["MY_API_KEY"]
    async with httpx.AsyncClient() as client:
        resp = await client.get(
            "https://api.example.com/data",
            headers={"Authorization": f"Bearer {api_key}"}
        )
        data = resp.json()
    return {"records": len(data)}

import httpx, os

async def run():
    ai_url = os.environ["GENMB_AI_URL"]
    async with httpx.AsyncClient() as client:
        resp = await client.post(ai_url, json={
            "prompt": "Summarize this in 3 bullet points: ...",
            "max_tokens": 300
        })
    return {"summary": resp.json()["text"]}

from google.cloud import firestore
import os

async def run():
    db = firestore.Client()
    state_ref = db.collection("agent_state").document(os.environ["AGENT_ID"])
    state = state_ref.get().to_dict() or {}
    last_processed = state.get("last_id", 0)

    # ... process new items since last_processed ...

    state_ref.set({"last_id": new_last_id}, merge=True)
    return {"processed": count}

Agents run on a cron schedule managed by Cloud Scheduler. You can also trigger a run manually at any time for testing, or configure a webhook trigger for event-driven execution.

Common Cron Expressions

Agents need credentials to call external APIs, send emails, or access third-party services. Add these as encrypted secrets — they are injected as environment variables at runtime and never stored in plaintext or logged.

Every agent execution is recorded in the Run History panel. Use it to debug failures, verify outputs, and understand execution patterns over time.

Data Insights lets you query your app's data using natural language. Ask questions in plain English and get answers with visualization suggestions — no SQL or database knowledge required.

Ask questions like "What were our top 5 products last month?" or "Show me daily signups for the past 30 days." GenMB translates your question into a database query automatically.

After each query, GenMB suggests the best visualization — bar chart, line graph, pie chart, or table. Click to add the visualization to your app. Data Insights works with your app's database (Firestore or PostgreSQL) and understands your schema to generate appropriate queries.

Data Connectors provide a secure backend proxy for connecting your app to external data sources. REST APIs, databases, and third-party services are accessed through an authenticated proxy so your credentials are never exposed.

All requests flow through GenMB's backend proxy with HMAC token authentication. Your API keys and credentials are never exposed in client-side code. Schema caching improves performance by storing API response schemas, reducing latency for repeated queries.

Use connectors to pull data from CRMs, payment processors, analytics platforms, or any REST API into your GenMB app.

Follow these guidelines to build reliable, maintainable backend automation.

Use idempotent operations whenever possible. If your agent processes records, store a "last processed ID" in Firestore and skip anything already handled. This makes reruns after failures safe and prevents duplicate actions.

Add explicit error handling and meaningful log output. A failed run with print("Starting fetch from API") and print(f"Fetched {len(items)} items") is much faster to debug than a silent failure.

Scheduled Agents can use Data Connectors to pull external data, process it with AI via the built-in proxy, and store results in your app's database for Data Insights queries. This combination creates a powerful end-to-end data pipeline — external source → agent transform → database → insight visualization.

FAQs