Merge branch 'main' into fix/response-model-field-incorrect-clean

Author: yossiovadia

README.md

@@ -8,7 +8,7 @@
 [![Crates.io](https://img.shields.io/crates/v/candle-semantic-router.svg)](https://crates.io/crates/candle-semantic-router)
 ![Test And Build](https://github.com/vllm-project/semantic-router/workflows/Test%20And%20Build/badge.svg)
 
-**📚 [Complete Documentation](https://vllm-semantic-router.com) | 🚀 [Quick Start](https://vllm-semantic-router.com/docs/installation) | 📣 [Blog](https://vllm-semantic-router.com/blog/) | 📖 [API Reference](https://vllm-semantic-router.com/docs/api/router/)**
+**📚 [Complete Documentation](https://vllm-semantic-router.com) | 🚀 [Quick Start](https://vllm-semantic-router.com/docs/installation) | 📣 [Blog](https://vllm-semantic-router.com/blog/) | 📖 [Publications](https://vllm-semantic-router.com/publications/)**
 
 ![code](./website/static/img/code.png)
 
@@ -64,20 +64,33 @@ Cache the semantic representation of the prompt so as to reduce the number of pr
 
 ### Distributed Tracing 🔍
 
-Comprehensive observability with OpenTelemetry distributed tracing provides fine-grained visibility into the request processing pipeline:
-
-- **Request Flow Tracing**: Track requests through classification, security checks, caching, and routing
-- **Performance Analysis**: Identify bottlenecks with detailed timing for each operation
-- **Security Monitoring**: Trace PII detection and jailbreak prevention operations
-- **Routing Decisions**: Understand why specific models were selected
-- **OpenTelemetry Standard**: Industry-standard tracing with support for Jaeger, Tempo, and other OTLP backends
+Comprehensive observability with OpenTelemetry distributed tracing provides fine-grained visibility into the request processing pipeline.
 
 ### Open WebUI Integration 💬
 
 To view the ***Chain-Of-Thought*** of the vLLM-SR's decision-making process, we have integrated with Open WebUI.
 
 ![code](./website/static/img/chat.png)
 
+## Quick Start 🚀
+
+Get up and running in seconds with our interactive setup script:
+
+```bash
+bash ./scripts/quickstart.sh
+```
+
+This command will:
+
+- 🔍 Check all prerequisites automatically
+- 📦 Install HuggingFace CLI if needed
+- 📥 Download all required AI models (~1.5GB)
+- 🐳 Start all Docker services
+- ⏳ Wait for services to become healthy
+- 🌐 Show you all the endpoints and next steps
+
+For detailed installation and configuration instructions, see the [Complete Documentation](https://vllm-semantic-router.com/docs/installation/).
+
 ## Documentation 📖
 
 For comprehensive documentation including detailed setup instructions, architecture guides, and API references, visit:

config/config-mcp-classifier-example.yaml

@@ -45,7 +45,19 @@ classifier:
 #
 # How it works:
 #   1. Router connects to MCP server at startup
-#   2. Calls 'list_categories' tool: MCP returns {"categories": ["business", "law", ...]}
+#   2. Calls 'list_categories' tool and MCP returns:
+#      {
+#        "categories": ["math", "science", "technology", "history", "general"],
+#        "category_system_prompts": {
+#          "math": "You are a mathematics expert. When answering math questions...",
+#          "science": "You are a science expert. When answering science questions...",
+#          "technology": "You are a technology expert..."
+#        },
+#        "category_descriptions": {
+#          "math": "Mathematical and computational queries",
+#          "science": "Scientific concepts and queries"
+#        }
+#      }
 #   3. For each request, calls 'classify_text' tool which returns:
 #      {
 #        "class": 3,
@@ -55,14 +67,28 @@ classifier:
 #      }
 #   4. Router uses the model and reasoning settings from MCP response
 #
+# PER-CATEGORY SYSTEM PROMPT INJECTION:
+#   - The MCP server provides SEPARATE system prompts for EACH category
+#   - Each category gets its own specialized instructions and context
+#   - The router stores these prompts and injects the appropriate one per query
+#   - Use classifier.GetCategorySystemPrompt(categoryName) to retrieve for a specific category
+#   - Examples:
+#       * Math category: "You are a mathematics expert. Show step-by-step solutions..."
+#       * Science category: "You are a science expert. Provide evidence-based answers..."
+#       * Technology category: "You are a tech expert. Include practical code examples..."
+#   - This allows domain-specific expertise per category
+#
 # BENEFITS:
 #   - MCP server makes intelligent routing decisions per query
 #   - No hardcoded routing rules needed in config
 #   - MCP can adapt routing based on query complexity, content, etc.
-#   - Centralized routing logic in MCP server
+#   - Centralized routing logic and per-category system prompts in MCP server
+#   - Category descriptions available for logging and debugging
+#   - Domain-specific LLM behavior for each category
 #
 # FALLBACK:
 #   - If MCP doesn't return model/use_reasoning, uses default_model below
+#   - If MCP doesn't return category_system_prompts, router can use default prompts
 #   - Can also add category-specific overrides here if needed
 #
 categories: []

dashboard/backend/main.go

@@ -161,6 +161,10 @@ func newReverseProxy(targetBase, stripPrefix string, forwardAuth bool) (*httputi
 		r.URL.Path = p
 		r.Host = targetURL.Host
 
+		// Set Origin header to match the target URL for Grafana embedding
+		// This is required for Grafana to accept the iframe embedding
+		r.Header.Set("Origin", targetURL.Scheme+"://"+targetURL.Host)
+
 		// Optionally forward Authorization header
 		if !forwardAuth {
 			r.Header.Del("Authorization")

dashboard/frontend/src/components/Layout.module.css

@@ -6,58 +6,127 @@
 }
 
 .sidebar {
-  width: 240px;
+  width: 260px;
   background-color: var(--color-bg-secondary);
   border-right: 1px solid var(--color-border);
   display: flex;
   flex-direction: column;
-  padding: 1rem 0.75rem;
-  gap: 1rem;
+  padding: 1.25rem 0.75rem;
+  gap: 1.5rem;
+  transition: width var(--transition-fast);
 }
 
-.brand {
+.sidebarCollapsed {
+  width: 70px;
+}
+
+.brandContainer {
   display: flex;
   align-items: center;
   gap: 0.5rem;
-  padding: 0 0.5rem;
+  padding: 0 0.25rem;
+  flex-direction: column;
+}
+
+.sidebarCollapsed .brandContainer {
+  flex-direction: column;
+  gap: 0.75rem;
+}
+
+.collapseButton {
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  width: 32px;
+  height: 32px;
+  padding: 6px;
+  background: transparent;
+  border: none;
+  border-radius: var(--radius-md);
+  cursor: pointer;
+  color: var(--color-text-secondary);
+  transition: all var(--transition-fast);
+  flex-shrink: 0;
+}
+
+.collapseButton:hover {
+  background-color: var(--color-bg-tertiary);
+  color: var(--color-text);
+}
+
+.collapseButton svg {
+  width: 20px;
+  height: 20px;
+  flex-shrink: 0;
+}
+
+.brand {
+  display: flex;
+  align-items: center;
+  gap: 0.625rem;
+  padding: 0.25rem 0.375rem;
   text-decoration: none;
   border-radius: var(--radius-md);
   transition: background-color var(--transition-fast);
   cursor: pointer;
+  flex: 1;
+  min-width: 0;
+  width: 100%;
 }
 
 .brand:hover {
   background-color: var(--color-bg-tertiary);
 }
 
+.sidebarCollapsed .brand {
+  justify-content: center;
+  padding: 0.5rem;
+}
+
 .logo {
   width: 28px;
   height: 28px;
   object-fit: contain;
+  flex-shrink: 0;
 }
 
 .brandText {
-  font-size: 0.95rem;
+  font-size: 0.9375rem;
   font-weight: 600;
   color: var(--color-text);
+  white-space: nowrap;
+  overflow: hidden;
+  text-overflow: ellipsis;
+  flex: 1;
+  min-width: 0;
 }
 
 .nav {
   display: flex;
   flex-direction: column;
-  gap: 0.25rem;
+  gap: 0.5rem;
+  flex: 1;
+  overflow-y: auto;
+  overflow-x: hidden;
 }
 
 .navLink {
   display: flex;
   align-items: center;
-  gap: 0.5rem;
-  padding: 0.5rem 0.625rem;
+  gap: 0.75rem;
+  padding: 0.75rem 0.875rem;
   border-radius: var(--radius-md);
   color: var(--color-text-secondary);
-  font-size: 0.9rem;
+  font-size: 0.9375rem;
   font-weight: 500;
   transition: all var(--transition-fast);
+  text-decoration: none;
+  border: none;
+  background: transparent;
+  cursor: pointer;
+  text-align: left;
+  width: 100%;
+  white-space: nowrap;
 }
 
 .navLink:hover {
@@ -76,14 +145,26 @@
 }
 
 .navIcon {
-  font-size: 1rem;
+  font-size: 1.125rem;
   line-height: 1;
-  width: 1.25rem;
+  width: 1.375rem;
   text-align: center;
+  flex-shrink: 0;
 }
 
 .navText {
   white-space: nowrap;
+  overflow: hidden;
+  text-overflow: ellipsis;
+}
+
+.sidebarCollapsed .navLink {
+  justify-content: center;
+  padding: 0.75rem 0.5rem;
+}
+
+.sidebarCollapsed .navIcon {
+  width: auto;
 }
 
 /* Sub Navigation (Configuration sections) - Match parent nav style */
@@ -188,4 +269,81 @@
   flex-direction: column;
   overflow: hidden;
   min-height: 0;
+}
+
+.header {
+  background-color: var(--color-bg-secondary);
+  border-bottom: 1px solid var(--color-border);
+  padding: 0 1.5rem;
+  height: 60px;
+  display: flex;
+  align-items: center;
+  flex-shrink: 0;
+}
+
+.headerContent {
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  width: 100%;
+}
+
+.headerLeft {
+  flex: 1;
+}
+
+.headerRight {
+  display: flex;
+  align-items: center;
+  gap: 1.5rem;
+}
+
+.headerLink {
+  color: var(--color-text-secondary);
+  text-decoration: none;
+  font-size: 0.9375rem;
+  font-weight: 500;
+  transition: color var(--transition-fast);
+  white-space: nowrap;
+}
+
+.headerLink:hover {
+  color: var(--color-text);
+}
+
+.mainContent {
+  flex: 1;
+  overflow: auto;
+  min-height: 0;
+}
+
+/* Responsive adjustments */
+@media (max-width: 768px) {
+  .sidebar {
+    width: 70px;
+  }
+
+  .brandText {
+    display: none;
+  }
+
+  .navText {
+    display: none;
+  }
+
+  .collapseButton {
+    display: none;
+  }
+
+  .sidebarFooter a {
+    display: none;
+  }
+
+  .header {
+    padding: 0 1rem;
+  }
+
+  .headerRight {
+    gap: 1rem;
+  }
 }