Skip to content
← Projects
SOFTWARE

Bambeso Automation: AI-Driven Dropshipping with OpenClaw

3 March 2026

Bambeso is a sustainable kitchen and home goods dropshipping brand I operate with my brother, Robbie, based out of Sydney.

Like all dropshipping businesses, the bottleneck isn't finding customers. It's the massive operational friction of sourcing products. Finding a competitor's product, verifying the AliExpress supplier, calculating margins, writing SEO-optimized copy, generating clean product images, and uploading it to Shopify takes a human about 3 hours per product.

To scale Bambeso, I built a headless Python pipeline orchestrated by a custom OpenClaw AI Agent named Mia. Mia has completely automated the backend logistics, cutting the time to launch a new product from 3 hours down to 15 minutes of human approval time.

🛠️ The Tech Stack

  • Backend / Pipeline: Python 3.11+, SQLite, Node.js (for AliExpress scraping)
  • Orchestration: OpenClaw Agent Layer ("Mia") operating via Telegram
  • Commerce Engine: Shopify Admin API (GraphQL 2026-01)
  • AI Integration: Gemini Flash (Listing/Keywords) & Nano Banana 2 (Image Generation)
  • Data Scoring: Google Trends API & Google Keyword Planner

🔄 The Autonomous Pipeline (How It Works)

The pipeline is entirely headless. Robbie never touches code. He simply talks to Mia on Telegram, and she handles the execution of the Python scripts.

The system is broken down into two distinct cycles: Discovery and Execution.

Phase 1: Automated Discovery & Scoring

Every morning at 7:00 AM (AEST), a cron job wakes the pipeline up to hunt for new products.

  1. The Scrape: A Brave Search queries 40 different competitors and scrapes their top-selling products.
  2. The Triage: The pipeline rejects any products that don't fit Bambeso's "sustainable kitchen" niche (e.g., appliances).
  3. The AI Scoring: The pipeline feeds the product keywords into the Google Trends and Google Keyword Planner APIs to score Australian search volume.
  4. The Pitch: At 8:00 AM, Mia messages Robbie on Telegram with the top 3 highest-scoring products of the day.

Phase 2: Human-in-the-Loop Approval

Robbie replies to Mia on Telegram: "approve 1". From here, the pipeline shifts from discovery to execution. Mia asks Robbie for the AliExpress supplier link and the target retail price, and then triggers the start_product.py script.

Phase 3 & 4: AI Generation

Once a product is locked in, the heavy lifting begins:

  1. Supplier Validation: The script checks the AliExpress supplier's rating, shipping time to Australia, and calculates if the profit margin is >60%.
  2. Listing Generation: Using Gemini Flash, the pipeline writes a Bambeso-branded product title, SEO description, and custom tags.
  3. Image Generation: The pipeline feeds the ugly, raw AliExpress supplier photos into the Nano Banana 2 image model as reference images. The AI generates clean, high-end lifestyle product photos that match the Bambeso aesthetic.

Mia sends these generated assets to Robbie for final approval. If Robbie wants changes, he just tells Mia ("Make the description punchier"), and she runs the generation scripts again with his notes.

💻 Code Deep Dive: Shopify GraphQL Integration

Once Robbie approves the final images and text, Mia executes the upload_to_shopify.py script.

Because we process products with multiple variants (e.g., Small, Medium, Large), we cannot use the legacy REST API. We utilize the Shopify 2026-01 GraphQL Admin API.

Shopify's modern GraphQL structure requires a complex two-step mutation flow for variants. The pipeline first creates the base product with productOptions, and then bulk-creates the variants using the REMOVE_STANDALONE_VARIANT strategy.

Here is a simplified look at the GraphQL mutation the pipeline generates to push the final approved AI listing directly to the Shopify Dev Store:

mutation createProductWithOptions($input: ProductInput!) {
  productCreate(input: $input) {
    product {
      id
      title
      status
      seo {
        title
        description
      }
    }
    userErrors {
      field
      message
    }
  }
}

Before this GraphQL call is made, the Python pipeline automatically uploads the AI-generated lifestyle images to Shopify's CDN using the stagedUploadsCreate endpoint, ensuring the final product draft is 100% complete and ready for Robbie to hit "Publish".

🚀 The Result

What used to take an entire evening of manual data entry, Photoshop, and supplier vetting now happens while Robbie drinks his morning coffee.

Mia handles the API complexity, the AI prompting, and the database tracking. Robbie just makes the executive decisions on Telegram.

OpenClaw runs dropshipping.