fix: jetbrains cleanup and logging, reduce ollama api/show calls

[RomneyDa]

Description

---

Summary by cubic

Lazily fetch and cache Ollama model info to cut repeated api/show calls. Cleaned up JetBrains startup and logging to reduce noise.

• Refactors

  • Added ensureModelInfo with a cached promise; model info fetched on first use, not in constructor.
  • Safer parsing of /api/show parameters; don’t override contextLength, ignore bad stop values, fail silently.
  • Consolidated JetBrains webview startup: single initial config load with polling; removed duplicate init.
  • Simplified IPC/TCP JSON parse error logs to reduce noisy stack traces.

• Bug Fixes

  • Enable OSR context menu on macOS by removing platform check.

^{Written for commit 685acbc. Summary will update automatically on new commits.}

[github-actions]

✅ Review Complete

Code Review Summary

⚠️ Continue configuration error. Please verify that the assistant exists in Continue Hub.

---

[cubic-dev-ai]

3 issues found across 4 files

Prompt for AI agents (all 3 issues)

Understand the root cause of the following 3 issues and fix them.


<file name="core/llm/llms/Ollama.ts">

<violation number="1" location="core/llm/llms/Ollama.ts:175">
Lazy-loading the model info leaves `supportsFim()` false for the first request, so the initial autocomplete call always skips the FIM path and ignores the suffix context for FIM-capable models.</violation>
</file>

<file name="binary/src/TcpMessenger.ts">

<violation number="1" location="binary/src/TcpMessenger.ts:127">
Dropping the caught error object from the console.error call removes the stack trace and error message, making JSON parsing failures much harder to debug.</violation>
</file>

<file name="binary/src/IpcMessenger.ts">

<violation number="1" location="binary/src/IpcMessenger.ts:91">
Dropping the caught error from the log removes stack/context information, making JSON parse failures hard to diagnose. Please log the error object along with the truncated line.</violation>
</file>

_{Reply to cubic to teach it or ask questions. Re-run a review with @cubic-dev-ai review this PR}

[cubic-dev-ai]

Lazy-loading the model info leaves supportsFim() false for the first request, so the initial autocomplete call always skips the FIM path and ignores the suffix context for FIM-capable models.

Prompt for AI agents

Address the following comment on core/llm/llms/Ollama.ts at line 175:

<comment>Lazy-loading the model info leaves `supportsFim()` false for the first request, so the initial autocomplete call always skips the FIM path and ignores the suffix context for FIM-capable models.</comment>

<file context>
@@ -161,47 +161,61 @@ class Ollama extends BaseLLM implements ModelInstaller {
+   * This is called on first use rather than in the constructor to avoid
+   * making HTTP requests when models are just being instantiated for config serialization.
+   */
+  private async ensureModelInfo(): Promise&lt;void&gt; {
+    if (this.model === &quot;AUTODETECT&quot;) {
       return;
</file context>

[cubic-dev-ai]

Dropping the caught error object from the console.error call removes the stack trace and error message, making JSON parsing failures much harder to debug.

Prompt for AI agents

Address the following comment on binary/src/TcpMessenger.ts at line 127:

<comment>Dropping the caught error object from the console.error call removes the stack trace and error message, making JSON parsing failures much harder to debug.</comment>

<file context>
@@ -124,7 +124,7 @@ export class TcpMessenger&lt;
           line.substring(0, 100) + &quot;...&quot; + line.substring(line.length - 100);
       }
-      console.error(&quot;Error parsing line: &quot;, truncatedLine, e);
+      console.error(&quot;Error parsing JSON from line: &quot;, truncatedLine);
       return;
     }
</file context>

Suggested change

	console.error("Error parsing JSON from line: ", truncatedLine);
	console.error("Error parsing JSON from line: ", truncatedLine, e);

[cubic-dev-ai]

Dropping the caught error from the log removes stack/context information, making JSON parse failures hard to diagnose. Please log the error object along with the truncated line.

Prompt for AI agents

Address the following comment on binary/src/IpcMessenger.ts at line 91:

<comment>Dropping the caught error from the log removes stack/context information, making JSON parse failures hard to diagnose. Please log the error object along with the truncated line.</comment>

<file context>
@@ -88,7 +88,7 @@ class IPCMessengerBase&lt;
           line.substring(0, 100) + &quot;...&quot; + line.substring(line.length - 100);
       }
-      console.error(&quot;Error parsing line: &quot;, truncatedLine, e);
+      console.error(&quot;Error parsing JSON from line: &quot;, truncatedLine);
       return;
     }
</file context>

Suggested change

	console.error("Error parsing JSON from line: ", truncatedLine);
	console.error("Error parsing JSON from line: ", truncatedLine, e);