fix(volcengine-ark): clamp GLM-5 max_tokens to model output ceiling (#2428)

Ark rejects max_tokens above 128000 for GLM-5.2. Add a config-driven STRIP_RULES entry that clamps max_tokens, max_completion_tokens and max_output_tokens down to the model maxOutput before the upstream call.

Co-authored-by: Cursor <cursoragent@cursor.com>

open-sse/translator/concerns/paramSupport.js

@@ -1,3 +1,5 @@
+import { getCapabilitiesForModel } from "../../providers/capabilities.js";
+
 // Strip request params a given provider/model rejects upstream (e.g. HTTP 400).
 // Config-driven: add a rule instead of scattering `delete body.x` across executors.
 
@@ -12,6 +14,7 @@ const STRIP_RULES = [
   { provider: "github", match: (m) => /claude/i.test(m) && !/claude.*(opus|sonnet).*4\.6/i.test(m), drop: ["thinking", "reasoning_effort"] },
   // Cloudflare Workers AI: content must be plain string, rejects OpenAI content-part array (#1926)
   { provider: "cloudflare-ai", flattenContent: true },
+  { provider: "volcengine-ark", match: /glm-5/i, clampToModelMaxOutput: true },
 ];
 
 // Test a rule's match (regex or predicate) against the model id.
@@ -20,6 +23,12 @@ function matches(rule, model) {
   return typeof rule.match === "function" ? rule.match(model) : rule.match.test(model);
 }
 
+function clampNumber(body, key, ceiling) {
+  if (typeof body[key] === "number" && Number.isFinite(body[key]) && body[key] > ceiling) {
+    body[key] = ceiling;
+  }
+}
+
 // Remove unsupported params from body in place; returns body.
 export function stripUnsupportedParams(provider, model, body) {
   if (!model || !body || typeof body !== "object") return body;
@@ -39,6 +48,14 @@ export function stripUnsupportedParams(provider, model, body) {
         }
       }
     }
+    if (rule.clampToModelMaxOutput) {
+      const ceiling = getCapabilitiesForModel(provider, model).maxOutput;
+      if (Number.isFinite(ceiling) && ceiling > 0) {
+        clampNumber(body, "max_tokens", ceiling);
+        clampNumber(body, "max_completion_tokens", ceiling);
+        clampNumber(body, "max_output_tokens", ceiling);
+      }
+    }
   }
   return body;
 }

tests/unit/param-support.test.js

@@ -28,4 +28,28 @@ describe("stripUnsupportedParams", () => {
 
     expect(body).toEqual({ top_p: 1 });
   });
+
+  it("clamps VolcEngine Ark GLM max token fields to the model output ceiling", () => {
+    const body = {
+      max_tokens: 131072,
+      max_completion_tokens: 131072,
+      max_output_tokens: 131072,
+    };
+
+    stripUnsupportedParams("volcengine-ark", "GLM-5.2", body);
+
+    expect(body).toEqual({
+      max_tokens: 128000,
+      max_completion_tokens: 128000,
+      max_output_tokens: 128000,
+    });
+  });
+
+  it("keeps VolcEngine Ark GLM max tokens when already under the ceiling", () => {
+    const body = { max_tokens: 64000 };
+
+    stripUnsupportedParams("volcengine-ark", "GLM-5.2", body);
+
+    expect(body.max_tokens).toBe(64000);
+  });
 });