Edge Ingress + Queue + AI Pipeline

Generic demo for digital businesses: Cloudflare validates public events, queues clean work, backend/GCP pulls safely, AI Gateway + Workers AI produces insight.

Worker-served UI · Phase 1 prototype
Architecture flow
Raw Internet Event
forms, webhooks, email replies, content events
Cloudflare Edge
WAF, bot controls, Worker validation
Cloudflare Queue
clean event buffer, absorbs spikes
Backend / GCP Pull
pulls only clean messages, controls rate
AI enrichment
Cloud Run calls Gateway, Gateway routes to model
AI Gatewayauth, logs, cost, cache, rate limits
↓
Workers AIsentiment classification model
Customer context
Scenario
1. Raw Event
Untrusted public traffic

            

          

        


        

          

            
2. Cloudflare Clean Edge

            
Worker validation + Queue buffer

          

          

            

            

            

              
0clean queued

              
0blocked at edge

            

          

        


        

          

            
3. Backend Pull + AI

            
GCP/Backend puller + AI Gateway / Workers AI

          

          

            

            

            

              
—AI latency

              
—tokens

            

            

              Cloud Run responseOptional: paste /pull-once JSON to render real AI Gateway + Workers AI output.
              
              

                
                
                Use Backend pull to copy curl, run it, paste JSON here.
              

            

          

        

      


      

        

          
Edge
1. Stop waste early
0 blocked
Bad traffic dies at Cloudflare before it can trigger origin, backend, or AI work.

          

        

        

          
Queue
2. Buffer bursts
0 queued
Valid events wait in Cloudflare Queues so backend does not scale instantly for bursts.

          

        

        

          
Pull
3. Pull at chosen pace
0 pulls
Backend/GCP controls batch size and processing rate instead of reacting to every request.

          

        

        

          
AI
4. Govern AI spend
0 AI calls
AI Gateway exposes model usage, latency, cache status, and rate-limit controls before AI spend runs away.

          

        

      


      

        
Timeline

        

      


      

        
AI security and cost-control value

        

          

            

              

                
AI Gateway

                Observe and control every AI call
                
One gateway layer for analytics, logs, request volume, token/cost visibility, latency, errors, and cache status across providers.

              

              
Obs

            

            
analyticsloggingprovider visibilitycost tracking

          

          

            

              

                
AI Gateway

                Reduce avoidable AI spend
                
Cache repeated prompts, rate-limit abuse, retry safely, and fall back between providers before usage turns into surprise cost.

              

              
Cost

            

            
cache hitsrate limitsfallbackcustom costs

          

          

            

              

                
AI Gateway

                Add security at the AI boundary
                
Authenticate gateway access, inspect prompts and responses with DLP/guardrails, and keep provider keys out of app code with BYOK.

              

              
Sec

            

            
authDLPguardrailsBYOK

          

          

            

              

                
Workers AI

                Run a Cloudflare-native model path
                
Use @cf models on Cloudflare's platform with serverless, pay-for-what-you-use neuron metering and a daily free allocation for demo-scale calls.

              

              
Model

            

            
@cf modelsserverless GPUsneuronsdaily free allocation

          

        

        
Source docs

        Workers pricing
        Queues pricing
        AI Gateway pricing
        Unified Billing
        BYOK (Store Keys)
        Gateway caching
        Gateway rate limiting
        Workers AI pricing