feat: remove tool display strip + extend seamless continuation for EOS fence truncation (v1.7.5)

Author: Brendan Gray

.github/copilot-instructions.md

@@ -62,6 +62,7 @@ After EVERY build, test run, or iteration where the user is about to test the ap
 - Run: `Clear-Content "C:\Users\brend\AppData\Roaming\guide-ide\logs\guide-main.log"`
 - Do this WITHOUT being asked. Every single time. No exceptions.
 - If you're about to say "ready to build" or "test it" — clear the logs first.
+- **ALSO clear immediately after a build/deployment sequence completes** — the user will test next. Do not wait to be asked.
 - The log file fills with stale entries from previous runs. Stale logs cause misdiagnosis.
 - Read log: `Get-Content "C:\Users\brend\AppData\Roaming\guide-ide\logs\guide-main.log"`
 
@@ -225,6 +226,9 @@ Do NOT stop at any step. Do NOT report success until step 10 is verified. If the
 - Do NOT run `npm run build`, `electron-builder`, or any build/package/installer command locally.
 - Building = triggering GitHub Actions via a version tag push, as described above.
 
+### GREEN LIGHT TO IMPLEMENT = GREEN LIGHT TO BUILD — NO EXCEPTIONS
+When the user approves a plan and says to proceed with implementation, that approval covers the FULL sequence: implement the changes AND run the complete 10-step build sequence defined above. Do NOT stop after writing code and wait for a second "build it" command. The build sequence is part of implementation. The task is not complete until step 10 is verified (graysoft.dev/download shows the new version and download URLs return HTTP 200).
+
 ### Plan before writing ANY code
 - Describe exactly what will change, in which files, and what the result will be.
 - Wait for explicit approval. Execute EXACTLY what was described — no more, no less.

main/agenticChat.js

@@ -1693,23 +1693,32 @@ function register(ctx) {
         // fullResponseText (fed back to the model for context) keeps the raw text.
         // displayResponseText (committed to the chat message) should only have natural language.
         // Targets: ```tool_call```, ```tool```, and ```json``` whose root object is a tool call.
+        // Tool-call fenced blocks are left in displayChunk so they appear as formatted code
+        // blocks in the chat bubble rather than being silently stripped to nothing.
         const displayChunk = responseText
-          .replace(/```(?:tool_call|tool)[^\n]*\n[\s\S]*?```/g, '')
-          .replace(/```json[^\n]*\n\s*(?:\[\s*)?\{\s*"(?:tool|name)"\s*:[\s\S]*?```/g, '')
           .replace(/\n{3,}/g, '\n\n');
         displayResponseText += displayChunk;
 
         // ── SEAMLESS CONTINUATION ──
         // If generation stopped because maxTokens was hit (not a natural EOS), and no tool
         // calls were returned, loop back and continue generating into the SAME open bubble.
         // The UI sees one uninterrupted stream throughout. Guard: max 3 continuations.
-        const _wasTruncated = (result?.stopReason === 'maxTokens' || result?.stopReason === 'max-tokens')
+        // Also trigger seamless continuation when EOS fires mid-tool-call (unclosed fenced block).
+        // This happens with small models (e.g. Qwen3-0.6B) that emit the EOS token before
+        // closing the ```json block — stopReason comes back as 'eogToken', not 'maxTokens'.
+        const _fenceIdx = responseText.search(/```(?:json|tool_call|tool)\b/);
+        const _hasUnclosedToolFence = _fenceIdx !== -1 && !responseText.slice(_fenceIdx).includes('\n```');
+        const _wasTruncated = (
+          (result?.stopReason === 'maxTokens' || result?.stopReason === 'max-tokens') ||
+          _hasUnclosedToolFence
+        )
           && nativeFunctionCalls.length === 0
           && !_timedOut
           && !isStale();
         if (_wasTruncated && continuationCount < 3) {
           continuationCount++;
-          console.log(`[AI Chat] Seamless continuation ${continuationCount}/3 — response hit maxTokens, continuing in same bubble`);
+          const _truncReason = _hasUnclosedToolFence ? 'unclosed tool fence (EOS mid-block)' : 'maxTokens';
+          console.log(`[AI Chat] Seamless continuation ${continuationCount}/3 — ${_truncReason}, continuing in same bubble`);
           iteration--; // Continuation is not a new agentic step
           currentPrompt = {
             systemContext: currentPrompt.systemContext, // Unchanged — KV cache preserved

package.json

@@ -1,6 +1,6 @@
 {
   "name": "guide-ide",
-  "version": "1.7.4",
+  "version": "1.7.5",
   "description": "guIDE - AI-Powered Offline IDE with local LLM, RAG, MCP tools, browser automation, and integrated terminal",
   "author": {
     "name": "Brendan Gray",