How to Use GLM-5.2 on Cloudflare Workers AI: Model ID, Pricing, and TypeScript Setup
A technical note covering the model ID, pricing, context length, Wrangler configuration, and TypeScript implementation for calling GLM-5.2 on Cloudflare Workers AI.
Posts
Before Using Flue: Figuring Out What This Framework Actually Is
A rough summary of how Flue thinks about harnesses, agents, workflows, skills, tools, sandboxes, and persistence — before actually running anything.
What Is Flue 1.0 Beta? New Features and Quickstart Caveats from Local Testing
Hands-on notes on running the Astro team's agent framework Flue 1.0 Beta locally, covering init, build, and run behavior plus the rough edges I hit.
Pre-Implementation Notes on Vercel's Agent Framework, eve
A summary of my pre-implementation research into Vercel's eve agent framework: directory-based design, the difference between tools and skills, sandboxing, durable execution, and more.
Moving Cloudflare Pages Deploys to npm Scripts
A short operations note on adding deploy scripts to package.json so I don't have to remember wrangler commands every time I redeploy to Cloudflare Pages.
Training Hermes Agent as a Business Decision Partner
An experiment in growing Hermes Agent into a business-ready support agent for decision-making.
Natural Language Data Analysis with ClickHouse and Claude MCP — Is the Era of Writing SQL Coming to an End?
Building an environment where Claude Desktop can operate ClickHouse through natural language using the ClickHouse MCP.
Why Cloudflare KV and a Small Admin Panel Fit a Parking Lot Site
A short design note on handling frequently changing public values—like parking availability and phone numbers—using Cloudflare KV and a simple admin panel instead of environment variables.
Building a Langfuse Morning Briefing for Slack
A practical experiment that aggregates Langfuse traces in a Cloudflare Worker and sends only token, cost, and latency anomalies to Slack.
I Want to Build a Langfuse Morning Briefing That Pushes Only Anomalies to Slack
Manually checking traces every day is unsustainable. A personal GenAIOps plan to aggregate Langfuse failures, token spikes, and low-score outputs into a single Slack morning briefing.
Why I Stopped Polishing Prompts and Started Using Feedback Loops
I explain why output quality stays unstable even with careful prompt design, and how I switched to a generate-evaluate-feedback-regenerate loop. Includes the smallest manual steps to start today.
How Much of Your LLM Traffic Can Cloudflare AI Gateway Actually Log?
Using Cloudflare AI Gateway as an OpenAI-compatible endpoint, I walk through its logging, payload controls, metadata, cost estimation, and OTel integration to see how far it can serve as an entry point for LLM observability.