quickdistill 0.1.7__tar.gz → 0.1.8__tar.gz

{quickdistill-0.1.7/quickdistill.egg-info → quickdistill-0.1.8}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: quickdistill
-Version: 0.1.7
+Version: 0.1.8
 Summary: Fast and easy toolkit for distilling AI models
 Author-email: Brett Young <bdytx5@umsystem.edu>
 License: MIT

{quickdistill-0.1.7 → quickdistill-0.1.8}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "quickdistill"
-version = "0.1.7"
+version = "0.1.8"
 description = "Fast and easy toolkit for distilling AI models"
 readme = "README.md"
 authors = [

{quickdistill-0.1.7 → quickdistill-0.1.8}/quickdistill/__init__.py

@@ -8,7 +8,7 @@ This package provides tools to:
 - Export datasets for model evaluation
 """
 
-__version__ = "0.1.7"
+__version__ = "0.1.8"
 __author__ = "Brett Young"
 __email__ = "bdytx5@umsystem.edu"
 

{quickdistill-0.1.7 → quickdistill-0.1.8}/quickdistill/default_judges.json

@@ -2,14 +2,14 @@
   {
     "name": "boolean_scorer",
     "type": "llm",
-    "model": "gpt-5",
+    "model": "openai/gpt-5",
     "returnType": "boolean",
     "prompt": "You are a strict evaluator comparing two AI responses (one from a strong reference model which is the ground truth, and one from a weaker model which we are testing to see how similar the responses it generates are to the strong model).\n\nStrong Model Response: {strong_output}\nWeak Model Response: {weak_output}\n\nDetermine if the weak model response is CORRECT compared to the strong model response.\nConsider a response CORRECT if it conveys the same key information and meaning, even if worded differently.\n\nRespond in JSON format: {'correct': true} or {'correct': false}"
   },
   {
     "name": "scalar_scorer",
     "type": "llm",
-    "model": "gpt-5",
+    "model": "openai/gpt-5",
     "returnType": "scalar",
     "prompt": "You are a strict evaluator comparing two AI responses (one from a strong reference model which is the ground truth, and one from a weaker model which we are testing to see how similar the responses it generates are to the strong model).\n\nStrong Model Response: {strong_output}\nWeak Model Response: {weak_output}\n\nEvaluate how similar the weak model response is to the strong model response.\nRate on a scale of 1-5 where 1=completely different and 5=nearly identical. RETURN ONLY ONE SCORE REPRESENTY THE AVERAGE SIMILARITY (EG 5-(avg_error))\n\nRespond in JSON format eg {'scores': the_score }"
   }

{quickdistill-0.1.7 → quickdistill-0.1.8}/quickdistill/static/judge_manager.html

@@ -183,12 +183,10 @@
 
             <div id="llm-options" style="display: block;">
                 <label for="judge-model">Model</label>
-                <select id="judge-model">
-                    <option value="gpt-5">gpt-5</option>
-                    <option value="gpt-4o">gpt-4o</option>
-                    <option value="gpt-4o-mini">gpt-4o-mini</option>
-                    <option value="claude-3-5-sonnet-20241022">claude-3-5-sonnet</option>
-                </select>
+                <input type="text" id="judge-model" placeholder="e.g., openai/gpt-5, anthropic/claude-3.5-sonnet" value="openai/gpt-5">
+                <p style="color: #888; font-size: 12px; margin-top: 5px; margin-bottom: 15px;">
+                    <strong>Note:</strong> Uses LiteLLM format. Examples: <code>openai/gpt-5</code>, <code>anthropic/claude-3.5-sonnet</code>, <code>openai/gpt-4o</code>
+                </p>
 
                 <label for="judge-return-type">Return Type</label>
                 <select id="judge-return-type">
@@ -393,10 +391,16 @@ Respond in JSON format: {'correct': true} or {'correct': false}`
             };
 
             if (type === 'llm') {
-                judge.model = document.getElementById('judge-model').value;
+                judge.model = document.getElementById('judge-model').value.trim();
                 judge.returnType = document.getElementById('judge-return-type').value;
                 judge.prompt = document.getElementById('judge-prompt').value.trim();
 
+                // Validate model
+                if (!judge.model) {
+                    alert('Error: Please enter a model (e.g., openai/gpt-5)');
+                    return;
+                }
+
                 // Validate required placeholders
                 if (!judge.prompt.includes('{strong_output}')) {
                     alert('Error: Judge prompt must include {strong_output} placeholder');
@@ -420,7 +424,7 @@ Respond in JSON format: {'correct': true} or {'correct': false}`
         function resetForm() {
             document.getElementById('judge-name').value = '';
             document.getElementById('judge-type').value = 'llm';
-            document.getElementById('judge-model').value = 'gpt-5-2025-08-07';
+            document.getElementById('judge-model').value = 'openai/gpt-5';
             document.getElementById('judge-prompt').value = '';
             document.getElementById('form-title').textContent = 'Create New Judge';
             document.getElementById('save-btn').textContent = 'Save Judge';

{quickdistill-0.1.7 → quickdistill-0.1.8}/quickdistill/static/trace_viewer.html

@@ -300,18 +300,25 @@
                 Select All Filtered
             </button>
 
-            <button id="export-btn" style="padding: 8px 16px; background: #4a9eff; color: white; border: none; border-radius: 4px; cursor: pointer;">
-                Export Selected to Test Set (<span id="selected-count">0</span>)
-            </button>
+            <!-- Manual Workflow Section -->
+            <div style="margin: 20px 0; padding: 15px; background: #1a2a1a; border-radius: 8px; border: 3px solid #ffffff;">
+                <div style="color: #ffffff; font-size: 14px; font-weight: 500; margin-bottom: 12px;">📋 Manual Workflow (Step-by-Step):</div>
+                <div style="display: flex; flex-wrap: wrap; gap: 10px;">
+                    <button id="export-btn" style="padding: 8px 16px; background: #4a9eff; color: white; border: none; border-radius: 4px; cursor: pointer;">
+                        Export Selected to Test Set (<span id="selected-count">0</span>)
+                    </button>
 
-            <button id="open-inference-btn" style="padding: 8px 16px; background: #7c4a9e; color: white; border: none; border-radius: 4px; cursor: pointer;">
-                Run Weak Models
-            </button>
+                    <button id="open-inference-btn" style="padding: 8px 16px; background: #7c4a9e; color: white; border: none; border-radius: 4px; cursor: pointer;">
+                        Run Weak Models
+                    </button>
 
-            <button id="open-eval-btn" style="padding: 8px 16px; background: #9e6a4a; color: white; border: none; border-radius: 4px; cursor: pointer;">
-                Run Evaluation
-            </button>
+                    <button id="open-eval-btn" style="padding: 8px 16px; background: #9e6a4a; color: white; border: none; border-radius: 4px; cursor: pointer;">
+                        Run Evaluation
+                    </button>
+                </div>
+            </div>
 
+            <!-- Utilities -->
             <a href="/judge" target="_blank" style="padding: 8px 16px; background: #4a5a9e; color: white; border: none; border-radius: 4px; text-decoration: none; display: inline-block;">
                 Manage Judges
             </a>
@@ -324,6 +331,7 @@
                 Settings
             </button>
 
+            <!-- Automatic Workflow Section -->
             <div style="margin: 20px 0; padding: 15px; background: #2a1a2a; border-radius: 8px; border: 1px solid #4a2a4a;">
                 <div style="color: #aaa; font-size: 13px; margin-bottom: 10px;">Automatic Workflow:</div>
                 <button id="open-e2e-btn" style="padding: 10px 20px; background: #7a4a9e; color: white; border: none; border-radius: 4px; cursor: pointer; font-weight: 500;">
@@ -511,9 +519,9 @@
                     <label style="display: block; color: #aaa; margin-bottom: 8px; font-size: 14px;">Judge Model:</label>
                     <input type="text" id="test-judge-model"
                         style="width: 100%; padding: 10px; background: #2a2a2a; color: #fff; border: 1px solid #3a3a3a; border-radius: 4px; font-size: 14px;"
-                        placeholder="e.g., gpt-4o, claude-3-5-sonnet-20241022">
+                        placeholder="e.g., openai/gpt-5, anthropic/claude-3.5-sonnet">
                     <div style="color: #666; font-size: 12px; margin-top: 5px;">
-                        Override the judge's model for this test
+                        Override the judge's model for this test. Uses LiteLLM format (e.g., <code style="color: #aaa;">openai/gpt-5</code>, <code style="color: #aaa;">anthropic/claude-3.5-sonnet</code>)
                     </div>
                 </div>
 

{quickdistill-0.1.7 → quickdistill-0.1.8/quickdistill.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: quickdistill
-Version: 0.1.7
+Version: 0.1.8
 Summary: Fast and easy toolkit for distilling AI models
 Author-email: Brett Young <bdytx5@umsystem.edu>
 License: MIT