fix(onboard): force chat completions API for Ollama

bin/lib/onboard.js

@@ -2732,7 +2732,14 @@ async function setupNim(gpu) {
           if (!validation.ok) {
             continue;
           }
-          preferredInferenceApi = validation.api;
+          // Ollama's /v1/responses endpoint does not produce correctly
+          // formatted tool calls — force chat completions like vLLM/NIM.
+          if (validation.api !== "openai-completions") {
+            console.log(
+              "  ℹ Using chat completions API (Ollama tool calls require /v1/chat/completions)",
+            );
+          }
+          preferredInferenceApi = "openai-completions";
           break;
         }
         break;
@@ -2777,7 +2784,14 @@ async function setupNim(gpu) {
           if (!validation.ok) {
             continue;
           }
-          preferredInferenceApi = validation.api;
+          // Ollama's /v1/responses endpoint does not produce correctly
+          // formatted tool calls — force chat completions like vLLM/NIM.
+          if (validation.api !== "openai-completions") {
+            console.log(
+              "  ℹ Using chat completions API (Ollama tool calls require /v1/chat/completions)",
+            );
+          }
+          preferredInferenceApi = "openai-completions";
           break;
         }
         break;

test/onboard-selection.test.js

@@ -624,7 +624,7 @@ const { setupNim } = require(${onboardPath});
     assert.equal(result.status, 0, result.stderr);
     const payload = JSON.parse(result.stdout.trim());
     assert.equal(payload.result.provider, "ollama-local");
-    assert.equal(payload.result.preferredInferenceApi, "openai-responses");
+    assert.equal(payload.result.preferredInferenceApi, "openai-completions");
     assert.ok(
       payload.lines.some((line) => line.includes("Loading Ollama model: nemotron-3-nano:30b")),
     );