PyPI - solace-agent-mesh - Versions diffs - 0.2.3__py3-none-any.whl → 1.0.1__py3-none-any.whl - Mend

solace-agent-mesh 0.2.3py3-none-any.whl → 1.0.1py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Potentially problematic release.

This version of solace-agent-mesh might be problematic. Click here for more details.

Files changed (518) hide show

solace_agent_mesh/evaluation/report/performance_metrics_styles.css ADDED Viewed

@@ -0,0 +1,93 @@
+/* Performance Metrics Table Styles */
+.performance-summary {
+    margin-top: 2rem;
+    padding-top: 1.5rem;
+    border-top: 2px solid #e9ecef;
+}
+.performance-summary h3 {
+    color: #2c3e50;
+    margin-bottom: 1rem;
+    font-size: 1.2rem;
+    font-weight: 600;
+}
+.metrics-table {
+    width: 100%;
+    border-collapse: collapse;
+    margin-top: 1rem;
+    background: white;
+    border-radius: 8px;
+    overflow: hidden;
+    box-shadow: 0 2px 4px rgba(0,0,0,0.1);
+}
+.metrics-table thead {
+    background: #f8f9fa;
+}
+.metrics-table th {
+    padding: 1rem;
+    text-align: left;
+    font-weight: 600;
+    padding: 1rem;
+    text-align: left;
+    font-weight: 600;
+    color: #2c3e50;
+    border-bottom: 2px solid #e9ecef;
+    cursor: pointer; /* Add cursor pointer to indicate clickability */
+}
+.metrics-table td {
+    padding: 0.8rem 1rem;
+    border-bottom: 1px solid #e9ecef;
+}
+.metrics-table tbody tr:hover {
+    background: #f8f9fa;
+}
+.model-name {
+    font-weight: 600;
+    color: #2c3e50;
+}
+.metric-value {
+    color: #2c3e50;
+    font-weight: 500;
+}
+.estimated-cost {
+    color: #27ae60;
+    font-weight: 600;
+}
+@media (max-width: 768px) {
+    .metrics-table {
+        font-size: 0.9rem;
+    }
+    .metrics-table th,
+    .metrics-table td {
+        padding: 0.6rem 0.8rem;
+    }
+}
+/* Ensure the performance metrics section appears below the chart */
+.performance-metrics-section {
+    margin-top: 2rem; /* Add space above the table */
+}
+/* Styles for individual run performance metrics */
+.run-performance {
+    text-align: right; /* Align content to the right */
+    font-size: 0.8rem; /* Make text smaller */
+    font-weight: bold; /* Make text bold */
+    margin-top: 0.5rem; /* Add some space above */
+    color: #555; /* Slightly muted color */
+}
+.run-execution-time {
+    display: inline-block; /* Display inline to be on the same line if space allows */
+    margin-left: 1rem; /* Add space for execution time */
+}

solace_agent_mesh/evaluation/report/templates/footer.html ADDED Viewed

	@@ -0,0 +1,2 @@
1	+ </body>
2	+ </html>

solace_agent_mesh/evaluation/report/templates/header.html ADDED Viewed

@@ -0,0 +1,340 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <title>SAM-Evals Benchmark Report</title>
+    <script src="https://cdn.jsdelivr.net/npm/chart.js"></script>
+    <script src="https://cdn.jsdelivr.net/npm/@sgratzl/chartjs-chart-boxplot"></script>
+    <style>
+        body {
+            font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Oxygen, Ubuntu, Cantarell, sans-serif;
+            line-height: 1.6;
+            color: #333;
+            max-width: 1600px;
+            margin: 0 auto;
+            padding: 20px;
+            background-color: #f8f9fa;
+        }
+        .header {
+            background: linear-gradient(135deg, #27ae60 0%, #2ecc71 100%);
+            color: white;
+            padding: 1rem;
+            border-radius: 8px;
+            margin-bottom: 1.5rem;
+            text-align: center;
+        }
+        .header h1 {
+            margin: 0;
+            font-size: 1.8rem;
+            font-weight: 400;
+        }
+        .section {
+            background: white;
+            padding: 1.5rem;
+            border-radius: 8px;
+            margin-bottom: 1.5rem;
+            box-shadow: 0 2px 8px rgba(0,0,0,0.08);
+        }
+        .section h2 {
+            color: #2c3e50;
+            border-bottom: 3px solid #27ae60;
+            padding-bottom: 0.5rem;
+            margin-top: 0;
+            font-size: 1.4rem;
+        }
+        .info-grid {
+            display: grid;
+            grid-template-columns: repeat(auto-fit, minmax(200px, 1fr));
+            gap: 0.2rem;
+            margin-bottom: 1rem;
+        }
+        .info-item {
+            background: #f8f9fa;
+            padding: 0.8rem;
+            border-radius: 5px;
+            border-left: 4px solid #27ae60;
+        }
+        .info-item strong {
+            color: #2c3e50;
+            display: block;
+            margin-bottom: 0.3rem;
+            font-size: 0.9rem;
+        }
+        .chart-container {
+            position: relative;
+            height: 400px;
+            margin: 1.5rem 0;
+        }
+        .model-list {
+            display: flex;
+            flex-wrap: wrap;
+            gap: 0.4rem;
+            margin-top: 0.4rem;
+        }
+        .model-tag {
+            background: #27ae60;
+            color: white;
+            padding: 0.2rem 0.6rem;
+            border-radius: 12px;
+            font-size: 0.85rem;
+        }
+        .test-files {
+            background: #f8f9fa;
+            padding: 0.8rem;
+            border-radius: 5px;
+            margin-top: 0.8rem;
+        }
+        .test-files ul {
+            margin: 0.4rem 0 0 0;
+            padding-left: 1.2rem;
+        }
+        .test-files li {
+            font-size: 0.9rem;
+        }
+        .evaluation-status {
+            display: inline-block;
+            padding: 0.2rem 0.6rem;
+            border-radius: 12px;
+            font-size: 0.85rem;
+            font-weight: bold;
+        }
+        .enabled {
+            background: #d4edda;
+            color: #155724;
+        }
+        .disabled {
+            background: #f8d7da;
+            color: #721c24;
+        }
+        .breakdown-container {
+            margin-top: 1rem;
+        }
+        .category-section {
+            margin-bottom: 1.5rem;
+            border: 1px solid #e9ecef;
+            border-radius: 8px;
+            overflow: hidden;
+        }
+        .category-header {
+            background: #f8f9fa;
+            padding: 1rem;
+            cursor: pointer;
+            border-bottom: 1px solid #e9ecef;
+            display: flex;
+            justify-content: space-between;
+            align-items: center;
+            transition: background-color 0.2s;
+        }
+        .category-header:hover {
+            background: #e9ecef;
+        }
+        .category-title {
+            font-size: 1.1rem;
+            font-weight: 600;
+            color: #2c3e50;
+            margin: 0;
+        }
+        .category-toggle {
+            font-size: 1.2rem;
+            color: #2c3e50;
+            transition: transform 0.2s;
+        }
+        .category-content {
+            display: none;
+            padding: 0.8rem;
+        }
+        .category-content.active {
+            display: block;
+        }
+        .test-item {
+            background: #f8f9fa;
+            padding: 0.4rem 0.6rem;
+            margin-bottom: 0.3rem;
+            border-radius: 4px;
+            border-left: 3px solid #6c757d;
+            font-size: 0.85rem;
+        }
+        .test-header {
+            display: flex;
+            align-items: center;
+            gap: 0.5rem;
+            margin-bottom: 0.2rem;
+        }
+        .test-name {
+            font-weight: 600;
+            color: #2c3e50;
+            font-size: 0.95rem;
+        }
+        .test-description {
+            font-size: 0.8rem;
+            color: #6c757d;
+            font-style: italic;
+        }
+        .model-scores {
+            display: flex;
+            gap: 0.6rem;
+            flex-wrap: wrap;
+            margin-top: 0.3rem;
+        }
+        .model-score {
+            padding: 0.25rem 0.5rem;
+            border-radius: 3px;
+            font-size: 0.8rem;
+            color: white;
+            font-weight: 600;
+        }
+        .score-high {
+            background: #27ae60; /* Green background for scores >= 0.7 */
+        }
+        .score-medium {
+            background: #f39c12; /* Yellow/Orange background for scores 0.4-0.69 */
+        }
+        .score-low {
+            background: #e74c3c; /* Red background for scores < 0.4 */
+        }
+        /* Detailed breakdown styles */
+        .model-result {
+            padding: 0.6rem;
+            margin-bottom: 0;
+            border-radius: 6px;
+            border: 1px solid #e9ecef;
+            box-shadow: 0 1px 3px rgba(0,0,0,0.05);
+            display: flex;
+            align-items: center;
+            gap: 0.8rem;
+            flex-wrap: wrap;
+        }
+        .model-result.score-high {
+            background: #27ae60;
+            border-color: #1e8449;
+        }
+        .model-result.score-medium {
+            background: #f39c12;
+            border-color: #d68910;
+        }
+        .model-result.score-low {
+            background: #e74c3c;
+            border-color: #c0392b;
+        }
+        .model-result .model-score {
+            display: inline-block;
+            padding: 0.25rem 0.5rem;
+            border-radius: 4px;
+            font-size: 0.85rem;
+            color: white;
+            font-weight: 600;
+            margin-bottom: 0;
+            background: rgba(0,0,0,0.2);
+        }
+        .model-result .score-value {
+            background: rgba(255,255,255,0.9);
+            padding: 0.2rem 0.5rem;
+            border-radius: 3px;
+            font-weight: 600;
+            border: 1px solid rgba(255,255,255,0.3);
+            font-size: 0.8rem;
+            color: #2c3e50;
+        }
+        .model-result .avg-duration {
+            background: rgba(255,255,255,0.9);
+            padding: 0.2rem 0.5rem;
+            border-radius: 3px;
+            font-weight: 600;
+            border: 1px solid rgba(255,255,255,0.3);
+            font-size: 0.8rem;
+            color: #2c3e50;
+        }
+        .score-value {
+            color: #2c3e50 !important;
+            font-weight: 600 !important;
+        }
+        .success-rate {
+            color: #27ae60 !important;
+        }
+        .avg-duration {
+            color: #2c3e50 !important;
+            font-weight: 600 !important;
+        }
+        .run-count {
+            color: #6c757d !important;
+            font-style: italic;
+        }
+        .model-results {
+            margin-top: 0.5rem;
+            display: grid;
+            grid-template-columns: repeat(auto-fit, minmax(280px, 1fr));
+            gap: 0.8rem;
+        }
+        @media (max-width: 768px) {
+            .info-grid {
+                grid-template-columns: 1fr;
+            }
+            .header h1 {
+                font-size: 1.5rem;
+            }
+            .model-scores {
+                flex-direction: column;
+                gap: 0.3rem;
+            }
+            .test-header {
+                flex-direction: column;
+                align-items: flex-start;
+                gap: 0.2rem;
+            }
+        }
+    </style>
+</head>
+<body>
+    <div class="header">
+        <h1>SAM-Evals Benchmark Report</h1>
+    </div>

solace-agent-mesh 0.2.3__py3-none-any.whl → 1.0.1__py3-none-any.whl

Potentially problematic release.

solace-agent-mesh 0.2.3py3-none-any.whl → 1.0.1py3-none-any.whl