Build an Agentic Coding Loop with GLM

The Vercel AI SDK runs the tool-calling loop; AnyRouter is the model provider. This example uses z-ai/glm-4.7-flash; any tool-capable model works by changing the id.

1. Install:

   npm install @anyr/ai-sdk-provider ai zod

2. Define tools and run the loop. stopWhen: stepCountIs(n) bounds the steps, and steps records each tool call and result:

   import { createAnyRouter } from "@anyr/ai-sdk-provider"
   import { generateText, tool, stepCountIs } from "ai"
   import { z } from "zod"
   import { readFile } from "node:fs/promises"
   
   const anyrouter = createAnyRouter() // reads ANYROUTER_API_KEY
   
   const { text, steps } = await generateText({
     model: anyrouter("z-ai/glm-4.7-flash"),
     system: "You are a coding agent. Use the tools to inspect the repo, then answer.",
     prompt: "What does src/index.ts export, and is it covered by a test?",
     stopWhen: stepCountIs(8),
     tools: {
       read_file: tool({
         description: "Read a UTF-8 text file from the repo.",
         parameters: z.object({ path: z.string() }),
         execute: async ({ path }) => ({ contents: await readFile(path, "utf8") }),
       }),
     },
   })
   
   console.log(text, `(${steps.length} steps)`)

3. To stream progress, use streamText and read result.fullStream (text-delta, tool-call, and tool results arrive as they happen).

4. To tune reasoning, pass reasoning_effort through extraBody:

   providerOptions: { anyrouter: { extraBody: { reasoning_effort: "high" } } },