fix: add nvidia to PROVIDER_TOOL_LIMITS (1536) to prevent tool truncation (#6177)

NVIDIA NIM API (nvidia/* models) silently truncates the tool list to 128
(the default MAX_TOOLS_LIMIT) because nvidia is not in PROVIDER_TOOL_LIMITS.
Tools beyond index 127 are dropped, causing agents to lose access to
critical tools like task, read, or high-index MCP tools.

Verified that NVIDIA NIM API supports up to 1536 tools by direct testing.
End-to-end confirmed: 198 tools sent, model successfully called tools at
indices 193, 195, and 197 (previously dropped by truncation to 128).

Follows the same pattern as #5563 (grok-cli: 200), integrated in v3.8.43.

open-sse/services/toolLimitDetector.ts

@@ -6,6 +6,7 @@ const DEFAULT_LIMIT = MAX_TOOLS_LIMIT;
 
 const PROVIDER_TOOL_LIMITS: Record<string, number> = {
   "grok-cli": 200,
+  "nvidia": 1536,
 };
 
 const _detectedLimitsSweep = setInterval(() => {

tests/unit/tool-limit-detector.test.ts

@@ -69,6 +69,15 @@ describe("toolLimitDetector", () => {
     assert.strictEqual(getEffectiveToolLimit("grok-cli"), 200);
   });
 
+  it("should return proactive limit for nvidia (1536) without any detection", () => {
+    assert.strictEqual(getEffectiveToolLimit("nvidia"), 1536);
+  });
+
+  it("should not override nvidia proactive limit with reactive detection", () => {
+    setDetectedToolLimit("nvidia", 100);
+    assert.strictEqual(getEffectiveToolLimit("nvidia"), 1536);
+  });
+
   it("should still return default (128) for unknown providers", () => {
     assert.strictEqual(getEffectiveToolLimit("some-new-provider"), 128);
   });