quickdistill 0.1.4__tar.gz → 0.1.6__tar.gz

{quickdistill-0.1.4/quickdistill.egg-info → quickdistill-0.1.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: quickdistill
-Version: 0.1.4
+Version: 0.1.6
 Summary: Fast and easy toolkit for distilling AI models
 Author-email: Brett Young <bdytx5@umsystem.edu>
 License: MIT
@@ -21,12 +21,11 @@ Classifier: Programming Language :: Python :: 3.11
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Requires-Python: >=3.8
 Description-Content-Type: text/markdown
-Requires-Dist: flask>=2.0.0
-Requires-Dist: flask-cors>=3.0.0
-Requires-Dist: openai>=1.0.0
-Requires-Dist: weave>=0.50.0
-Requires-Dist: llmasajudge>=0.1.0
-Requires-Dist: datasets>=2.0.0
+Requires-Dist: flask==2.3.2
+Requires-Dist: flask-cors==4.0.0
+Requires-Dist: openai==2.14.0
+Requires-Dist: weave==0.52.14
+Requires-Dist: llmasajudge==0.1.15
 Provides-Extra: dev
 Requires-Dist: pytest>=7.0.0; extra == "dev"
 Requires-Dist: black>=22.0.0; extra == "dev"

{quickdistill-0.1.4 → quickdistill-0.1.6}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "quickdistill"
-version = "0.1.4"
+version = "0.1.6"
 description = "Fast and easy toolkit for distilling AI models"
 readme = "README.md"
 authors = [
@@ -26,12 +26,11 @@ classifiers = [
 keywords = ["ai", "ml", "distillation", "evaluation", "weave"]
 requires-python = ">=3.8"
 dependencies = [
-    "flask>=2.0.0",
-    "flask-cors>=3.0.0",
-    "openai>=1.0.0",
-    "weave>=0.50.0",
-    "llmasajudge>=0.1.0",
-    "datasets>=2.0.0",
+    "flask==2.3.2",
+    "flask-cors==4.0.0",
+    "openai==2.14.0",
+    "weave==0.52.14",
+    "llmasajudge==0.1.15",
 ]
 
 [project.optional-dependencies]

{quickdistill-0.1.4 → quickdistill-0.1.6}/quickdistill/__init__.py

@@ -8,7 +8,7 @@ This package provides tools to:
 - Export datasets for model evaluation
 """
 
-__version__ = "0.1.4"
+__version__ = "0.1.6"
 __author__ = "Brett Young"
 __email__ = "bdytx5@umsystem.edu"
 

{quickdistill-0.1.4 → quickdistill-0.1.6}/quickdistill/server.py

@@ -3,6 +3,7 @@ import json
 import openai
 import weave
 import shutil
+import threading
 from flask import Flask, request, jsonify, send_from_directory
 from flask_cors import CORS
 from llmasajudge import LLMAsAJudge
@@ -30,6 +31,9 @@ if default_project_src.exists() and not default_project_dst.exists():
 app = Flask(__name__, static_folder=str(STATIC_DIR))
 CORS(app)
 
+# Progress tracking for long-running operations
+progress_state = {}
+
 # Configuration
 PROJECT = "wandb_inference"
 
@@ -152,6 +156,7 @@ def run_inference_endpoint():
     models = data.get('models', [])
     strong_export_file = data.get('strong_export_file')
     num_examples = data.get('num_examples')
+    task_id = data.get('task_id', f"inference_{id(models)}")
 
     if not models:
         return jsonify({'error': 'No models provided'}), 400
@@ -176,8 +181,17 @@ def run_inference_endpoint():
 
     output_files = []
 
+    # Initialize progress tracking
+    total_steps = len(models) * len(traces)
+    progress_state[task_id] = {
+        'current': 0,
+        'total': total_steps,
+        'message': 'Starting inference...',
+        'status': 'running'
+    }
+
     # Run inference for each model
-    for model in models:
+    for model_idx, model in enumerate(models):
         print(f"Running model: {model}")
         results = []
 
@@ -185,6 +199,13 @@ def run_inference_endpoint():
         client = get_client_for_model(model)
 
         for i, trace in enumerate(traces):
+            step = model_idx * len(traces) + i + 1
+            progress_state[task_id] = {
+                'current': step,
+                'total': total_steps,
+                'message': f'[{model_idx+1}/{len(models)}] {model} - Example {i+1}/{len(traces)}',
+                'status': 'running'
+            }
             print(f"  Processing example {i+1}/{len(traces)}...", end=' ')
 
             # Extract messages
@@ -239,13 +260,30 @@ def run_inference_endpoint():
         output_files.append(str(output_file))
         print(f"Saved {len(results)} results to {output_file}")
 
+    # Mark progress as complete
+    progress_state[task_id] = {
+        'current': total_steps,
+        'total': total_steps,
+        'message': 'Complete!',
+        'status': 'complete'
+    }
+
     return jsonify({
         'status': 'success',
         'files': output_files,
         'total_examples': len(traces),
-        'models_run': len(models)
+        'models_run': len(models),
+        'task_id': task_id
     })
 
+@app.route('/progress/<task_id>', methods=['GET'])
+def get_progress(task_id):
+    """Get progress for a running task"""
+    if task_id in progress_state:
+        return jsonify(progress_state[task_id])
+    return jsonify({'error': 'Task not found'}), 404
+
+
 @app.route('/list_weak_models', methods=['GET'])
 def list_weak_models():
     """List available weak model result files with metadata"""
@@ -469,27 +507,38 @@ def run_evaluation_endpoint():
     data = request.json
     model_file = data.get('model_file')
     judge = data.get('judge')
+    task_id = data.get('task_id', f"eval_{id(data)}")
 
     if not model_file or not judge:
         return jsonify({'error': 'Missing model_file or judge'}), 400
 
     # Load weak model results
-    with open(model_file, 'r') as f:
-        data = json.load(f)
+    model_path = DATA_DIR / model_file
+    with open(model_path, 'r') as f:
+        file_data = json.load(f)
 
     # Handle both old format (list) and new format (dict with metadata)
-    if isinstance(data, dict) and 'results' in data:
-        metadata = data.get('metadata', {})
-        results = data['results']
+    if isinstance(file_data, dict) and 'results' in file_data:
+        metadata = file_data.get('metadata', {})
+        results = file_data['results']
         strong_export = metadata.get('strong_export_file', 'unknown')
     else:
         # Old format - just a list
-        results = data
+        results = file_data
         strong_export = 'unknown'
 
     # Extract model name from filename
     model_name = model_file.replace('weak_model_', '').replace('.json', '')
 
+    # Initialize progress tracking
+    total_steps = len(results)
+    progress_state[task_id] = {
+        'current': 0,
+        'total': total_steps,
+        'message': f'Starting evaluation: {model_name} with {judge["name"]}...',
+        'status': 'running'
+    }
+
     # Create evaluation logger
     ev = weave.EvaluationLogger(
         name=f"eval-{model_name}-{judge['name']}",
@@ -497,7 +546,13 @@ def run_evaluation_endpoint():
     )
 
     # Run evaluation
-    for example in results:
+    for idx, example in enumerate(results):
+        progress_state[task_id] = {
+            'current': idx + 1,
+            'total': total_steps,
+            'message': f'{model_name} - Example {idx+1}/{total_steps}',
+            'status': 'running'
+        }
         # Skip examples with errors (null messages/output)
         if example.get('error') or not example.get('output'):
             continue
@@ -533,12 +588,21 @@ def run_evaluation_endpoint():
     # Finish evaluation
     ev.log_summary()
 
+    # Mark progress as complete
+    progress_state[task_id] = {
+        'current': total_steps,
+        'total': total_steps,
+        'message': 'Complete!',
+        'status': 'complete'
+    }
+
     return jsonify({
         'status': 'success',
         'evaluation_name': f"eval-{model_name}-{judge['name']}",
         'examples_evaluated': len(results),
         'weave_url': ev.ui_url,
-        'strong_export': strong_export
+        'strong_export': strong_export,
+        'task_id': task_id
     })
 
 

{quickdistill-0.1.4 → quickdistill-0.1.6}/quickdistill/static/judge_manager.html

@@ -172,7 +172,6 @@
             <label for="judge-type">Judge Type</label>
             <select id="judge-type">
                 <option value="llm">LLM-as-a-Judge</option>
-                <option value="custom">Custom Function</option>
             </select>
 
             <div id="llm-options" style="display: block;">
@@ -204,13 +203,6 @@
                 <textarea id="judge-prompt"></textarea>
             </div>
 
-            <div id="custom-options" style="display: none;">
-                <label for="custom-function">Custom Function (Python)</label>
-                <textarea id="custom-function" placeholder="def custom_judge(strong_output: str, weak_output: str) -> dict:
-    # Your custom logic here
-    return {'similarity': 1.0 if strong_output == weak_output else 0.0}"></textarea>
-            </div>
-
             <button onclick="saveJudge()" id="save-btn">Save Judge</button>
             <button onclick="cancelEdit()" id="cancel-btn" style="display: none; background: #5a2a2a; margin-left: 10px;">Cancel</button>
         </div>
@@ -340,8 +332,6 @@ Respond in JSON format: {'correct': true} or {'correct': false}`
                     alert('Error: Judge prompt must include {weak_output} placeholder');
                     return;
                 }
-            } else {
-                judge.customFunction = document.getElementById('custom-function').value.trim();
             }
 
             const success = await saveJudgeToServer(judge);
@@ -358,7 +348,6 @@ Respond in JSON format: {'correct': true} or {'correct': false}`
             document.getElementById('judge-type').value = 'llm';
             document.getElementById('judge-model').value = 'gpt-5-2025-08-07';
             document.getElementById('judge-prompt').value = '';
-            document.getElementById('custom-function').value = '';
             document.getElementById('form-title').textContent = 'Create New Judge';
             document.getElementById('save-btn').textContent = 'Save Judge';
             document.getElementById('cancel-btn').style.display = 'none';
@@ -383,8 +372,6 @@ Respond in JSON format: {'correct': true} or {'correct': false}`
                 document.getElementById('judge-model').value = judge.model;
                 document.getElementById('judge-return-type').value = judge.returnType || 'scalar';
                 document.getElementById('judge-prompt').value = judge.prompt || '';
-            } else {
-                document.getElementById('custom-function').value = judge.customFunction || '';
             }
 
             document.getElementById('form-title').textContent = 'Edit Judge';
@@ -437,9 +424,8 @@ Respond in JSON format: {'correct': true} or {'correct': false}`
 
         // Toggle judge type options
         function toggleJudgeType() {
-            const type = document.getElementById('judge-type').value;
-            document.getElementById('llm-options').style.display = type === 'llm' ? 'block' : 'none';
-            document.getElementById('custom-options').style.display = type === 'custom' ? 'block' : 'none';
+            // Only LLM type is supported now
+            document.getElementById('llm-options').style.display = 'block';
         }
 
         document.getElementById('judge-type').addEventListener('change', toggleJudgeType);

{quickdistill-0.1.4 → quickdistill-0.1.6}/quickdistill/static/trace_viewer.html

@@ -375,6 +375,9 @@
 
                 <div id="inference-progress" style="display: none; margin-top: 20px; padding: 15px; background: #0f0f0f; border-radius: 4px;">
                     <div style="color: #4a9eff; margin-bottom: 10px;">Running inference...</div>
+                    <div id="inference-progress-bar" style="width: 100%; height: 6px; background: #2a2a2a; border-radius: 3px; margin-bottom: 15px; overflow: hidden;">
+                        <div id="inference-progress-fill" style="height: 100%; background: #4a9eff; width: 0%; transition: width 0.3s;"></div>
+                    </div>
                     <div id="progress-text" style="color: #888; font-family: monospace; font-size: 12px; white-space: pre-wrap;"></div>
                 </div>
             </div>
@@ -919,20 +922,46 @@
             // Show progress
             document.getElementById('inference-progress').style.display = 'block';
             const progressText = document.getElementById('progress-text');
+            const progressFill = document.getElementById('inference-progress-fill');
             progressText.textContent = `Starting inference...\n`;
-            progressText.textContent += `Strong Export: ${strongExportFile}\n`;
-            progressText.textContent += `Models: ${allModels.join(', ')}\n`;
-            progressText.textContent += `Max Examples: ${numExamples}\n\n`;
+            progressFill.style.width = '0%';
+
+            // Start inference and poll for progress
+            let taskId = null;
+            let pollInterval = null;
+
+            const pollProgress = async () => {
+                if (!taskId) return;
+                try {
+                    const resp = await fetch(`/progress/${taskId}`);
+                    if (resp.ok) {
+                        const progress = await resp.json();
+                        const percent = (progress.current / progress.total) * 100;
+                        progressFill.style.width = `${percent}%`;
+                        progressText.textContent = `${progress.message}\nProgress: ${progress.current}/${progress.total} (${percent.toFixed(1)}%)\n`;
+                    }
+                } catch (e) {
+                    console.error('Error polling progress:', e);
+                }
+            };
 
             // Call backend API
             try {
+                // Generate a task ID for polling
+                taskId = `inference_${Date.now()}`;
+
+                // Start polling immediately
+                pollInterval = setInterval(pollProgress, 300);
+
+                // Start the inference
                 const response = await fetch('/run_inference', {
                     method: 'POST',
                     headers: { 'Content-Type': 'application/json' },
                     body: JSON.stringify({
                         models: allModels,
                         strong_export_file: strongExportFile,
-                        num_examples: numExamples
+                        num_examples: numExamples,
+                        task_id: taskId
                     })
                 });
 
@@ -941,8 +970,12 @@
                 }
 
                 const result = await response.json();
-                progressText.textContent += `\n✓ Complete!\n`;
-                progressText.textContent += `Results saved to: ${result.files.join(', ')}\n`;
+
+                // Stop polling
+                if (pollInterval) clearInterval(pollInterval);
+
+                progressText.textContent = `\n✓ Complete!\nResults saved to: ${result.files.join(', ')}\n`;
+                progressFill.style.width = '100%';
 
                 setTimeout(() => {
                     document.getElementById('inference-panel').style.display = 'none';
@@ -951,8 +984,7 @@
 
             } catch (error) {
                 progressText.textContent += `\n✗ Error: ${error.message}\n`;
-                progressText.textContent += `\nNote: You need to run the backend server for inference.\n`;
-                progressText.textContent += `Run: python inference_server.py\n`;
+                if (pollInterval) clearInterval(pollInterval);
             }
         });
 
@@ -1091,21 +1123,44 @@
             const modelFiles = Array.from(selectedEvalModels);
             const results = [];
 
-            // Run evaluations sequentially
+            // Run evaluations sequentially with granular progress
             for (let i = 0; i < modelFiles.length; i++) {
                 const modelFile = modelFiles[i];
-                const progress = ((i) / modelFiles.length) * 100;
-                progressFill.style.width = `${progress}%`;
 
-                progressText.textContent += `[${i+1}/${modelFiles.length}] Evaluating ${modelFile}...\n`;
+                progressText.textContent += `[${i+1}/${modelFiles.length}] Starting ${modelFile}...\n`;
+
+                let pollInterval = null;
+                let taskId = null;
+
+                const pollProgress = async () => {
+                    if (!taskId) return;
+                    try {
+                        const resp = await fetch(`/progress/${taskId}`);
+                        if (resp.ok) {
+                            const progress = await resp.json();
+                            const percent = (progress.current / progress.total) * 100;
+                            progressFill.style.width = `${percent}%`;
+                            progressText.textContent = `[${i+1}/${modelFiles.length}] ${progress.message}\nProgress: ${progress.current}/${progress.total} (${percent.toFixed(1)}%)\n`;
+                        }
+                    } catch (e) {
+                        console.error('Error polling eval progress:', e);
+                    }
+                };
 
                 try {
+                    // Generate task ID for this evaluation
+                    taskId = `eval_${Date.now()}_${i}`;
+
+                    // Start polling
+                    pollInterval = setInterval(pollProgress, 300);
+
                     const response = await fetch('/run_evaluation', {
                         method: 'POST',
                         headers: { 'Content-Type': 'application/json' },
                         body: JSON.stringify({
                             model_file: modelFile,
-                            judge: judge
+                            judge: judge,
+                            task_id: taskId
                         })
                     });
 
@@ -1114,6 +1169,10 @@
                     }
 
                     const result = await response.json();
+
+                    // Clear polling when done
+                    if (pollInterval) clearInterval(pollInterval);
+
                     progressText.textContent += `  ✓ Complete: ${result.evaluation_name}\n`;
                     progressText.textContent += `  Examples: ${result.examples_evaluated}\n\n`;
 
@@ -1125,6 +1184,7 @@
                     });
 
                 } catch (error) {
+                    if (pollInterval) clearInterval(pollInterval);
                     progressText.textContent += `  ✗ Error: ${error.message}\n\n`;
                 }
             }

{quickdistill-0.1.4 → quickdistill-0.1.6/quickdistill.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: quickdistill
-Version: 0.1.4
+Version: 0.1.6
 Summary: Fast and easy toolkit for distilling AI models
 Author-email: Brett Young <bdytx5@umsystem.edu>
 License: MIT
@@ -21,12 +21,11 @@ Classifier: Programming Language :: Python :: 3.11
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Requires-Python: >=3.8
 Description-Content-Type: text/markdown
-Requires-Dist: flask>=2.0.0
-Requires-Dist: flask-cors>=3.0.0
-Requires-Dist: openai>=1.0.0
-Requires-Dist: weave>=0.50.0
-Requires-Dist: llmasajudge>=0.1.0
-Requires-Dist: datasets>=2.0.0
+Requires-Dist: flask==2.3.2
+Requires-Dist: flask-cors==4.0.0
+Requires-Dist: openai==2.14.0
+Requires-Dist: weave==0.52.14
+Requires-Dist: llmasajudge==0.1.15
 Provides-Extra: dev
 Requires-Dist: pytest>=7.0.0; extra == "dev"
 Requires-Dist: black>=22.0.0; extra == "dev"

quickdistill-0.1.6/quickdistill.egg-info/requires.txt

@@ -0,0 +1,11 @@
+flask==2.3.2
+flask-cors==4.0.0
+openai==2.14.0
+weave==0.52.14
+llmasajudge==0.1.15
+
+[dev]
+pytest>=7.0.0
+black>=22.0.0
+isort>=5.0.0
+flake8>=4.0.0

quickdistill-0.1.4/quickdistill.egg-info/requires.txt

@@ -1,12 +0,0 @@
-flask>=2.0.0
-flask-cors>=3.0.0
-openai>=1.0.0
-weave>=0.50.0
-llmasajudge>=0.1.0
-datasets>=2.0.0
-
-[dev]
-pytest>=7.0.0
-black>=22.0.0
-isort>=5.0.0
-flake8>=4.0.0