I PUT A PYTHON VOICE AGENT IN A CLOUDFLARE CONTAINER. TURN-TAKING WAS THE WHOLE FIGHT.
A production phone agent on Pipecat in a Cloudflare Container. A per-call turn-detection model, scale-to-zero, and every millisecond of latency I fought for.
7 posts
A production phone agent on Pipecat in a Cloudflare Container. A per-call turn-detection model, scale-to-zero, and every millisecond of latency I fought for.
I was evaluating Firecrawl. Then I found Cloudflare's Browser Rendering API. The /markdown endpoint does what I needed. One REST call. URL in, clean markdown out. Architecture, Firecrawl comparison, and the gaps I've hit in production.
Cloudflare shipped Markdown for Agents — any site on their network can now serve markdown instead of HTML through content negotiation. One header. No origin changes. Here's how it fits into an agent pipeline.
Cloudflare blew up the Worker subrequest limit — from 1,000 to a configurable max of 10 million. What changed, why it matters for Workflows, Durable Objects, and batch jobs.
Comparing Cloudflare's Moltworker container-based agent to a native Durable Objects approach — cold starts, state management, browser automation, and cost.
How I got Remotion running entirely on Cloudflare — Workers, Containers, R2, Queues — with zero AWS dependencies and zero egress fees.
How I replaced an $83/month Firecrawl subscription with Cloudflare's Browser Rendering API for effectively $0 when scraping public technical content.