greenbids-tailor 0.1.2

Bring traffic shaping to your own cloud

🪡 Greenbids Tailor

Bring traffic shaping to your own cloud!

🍱 Integration

Sequence Diagram

Following the interaction diagram provided by the OpenRTB API Specification (version 2.5) (§2) here is an example of where the Greenbids Tailor product must be called.

sequenceDiagram
    participant Publisher
    activate Publisher

    box rgba(128, 128, 128, 0.33) Partner Network
        participant SSP as RTB Exchange
        participant GB as Greenbids Tailor
    end

    Publisher ->>+ SSP: 0. Ad Request

    rect rgba(30, 183, 136, 0.66)
    SSP ->>+ GB: PUT / @[Fabric, ...]
    GB -->>- SSP: 200 @[Fabric, ...]
    end

    loop for each request where fabric.prediction.shouldSend
        SSP ->>+ Bidder : 1. Bid Request
        alt 200
        Bidder -->> SSP: Bid Response
        else 204
        Bidder -->>- SSP: No Response
        end
    end

    opt fabric.prediction.<br>isExploration
        rect rgba(30, 183, 136, 0.66)
            SSP -)+ GB: POST / @[Fabric, ...]
            GB -->>- SSP: 200
        end
    end
    GB --) greenbids.ai: telemetry

    note over Publisher,SSP: Continue auction process
    deactivate SSP
    deactivate Publisher

Example

Here is a Python-like pseudocode example of integration.

def handleAdRequest(adRequest):
  # Call the Greenbids Tailor PUT endpoint to filter the list of possible bidders
  # This is a local call (localhost or local network) to ensure low latency
  fabrics = requests.put("http://tailor:8000", json=[
    {
      "featureMap": {
        "bidder": bidder.name,
        "device": adRequest.get("device", "unknown"),
        "userId": adRequest.get("user_id", "unknown"),
        "secretIngredient": "our own secret ingredient",
        "greatAnswer": 42,
      }
    }
    for bidder in BIDDER_LIST
  ]).json()

  # Do your regular calls here to send a bid requests to the selected bidders
  for fabric in fabrics:
    if not fabric["prediction"]["shouldSend"]:
        # Skip any bidder that as too few response probability
        continue

    # Create a bid request and send it to the bidder
    rsp = requests.post(bidder.url, json={...})

    # Store the outcome in the fabric
    fabric["groundTruth"]["hasResponse"] = (rsp.status_code != 204)

  # Call the Greenbids Tailor POST endpoint for exploration
  if fabrics[0]["prediction"]["isExploration"]:
    # You may use a fire-and-forget mechanism
    requests.post("http://tailor:8000", json=fabrics)

  # Continue the auction process
  ...

Routes

See API documentation online.

🚀 Deployment

Depending on your current stack, find the best way to deploy this service.

🐍 Executable

pip install greenbids-tailor
greenbids-tailor

We advise you to create a virtual environment to avoid any dependency mismatch on your system.

🐳 Docker

docker run -P -d --name greenbids-tailor ghcr.io/greenbids/tailor:latest
docker port greenbids-tailor

☸ Helm

helm upgrade --install --create-namespace --namespace greenbids tailor oci://ghcr.io/greenbids/charts/tailor

You have to authenticate your Kubernetes cluster against GitHub container registry, and pass the name of the secret through --set 'imagePullSecrets[0].name=<your-secret-name>'.