text
stringlengths
0
59.1k
'primaryBorderColor': '#059669',
'lineColor': '#6ee7b7',
'secondaryColor': '#d1fae5',
'tertiaryColor': '#ecfdf5',
'background': '#ffffff',
'mainBkg': '#10b981',
'secondBkg': '#6ee7b7',
'tertiaryBkg': '#a7f3d0',
'activationBorderColor': '#059669',
'activationBkgColor': '#d1fae5',
'sequenceNumberColor': '#ffffff',
'sectionBkgColor': '#ecfdf5',
'altSectionBkgColor': '#f0fdf4',
'gridColor': '#d1fae5',
'textColor': '#6ee7b7',
'taskBkgColor': '#10b981',
'taskTextColor': '#ffffff',
'activeTaskBkgColor': '#059669',
'activeTaskBorderColor': '#047857',
'gridColor': '#d1fae5',
'section0': '#10b981',
'section1': '#6ee7b7',
'section2': '#a7f3d0',
'section3': '#d1fae5',
'messageLine0': '#6ee7b7',
'messageLine1': '#6ee7b7',
'messageText': '#6ee7b7',
'loopTextColor': '#047857',
'noteBorderColor': '#6ee7b7',
'noteTextColor': '#6ee7b7',
'activationBorderColor': '#6ee7b7',
'signalColor': '#6ee7b7',
'signalTextColor': '#6ee7b7'
}
}}%%
sequenceDiagram
participant U as 👤 User
participant VA as 🤖 VoltAgent
participant LLM as 🧠 LLM Provider
participant LE as 📊 Langfuse Exporter
participant LF as 🔍 Langfuse Platform
participant D as 👨‍💻 Developer
Note over U,D: 🚀 LLM Observability Flow
U->>+VA: Send Query
Note right of U: "Where's my order? 📦"
VA->>+LE: Start Trace
LE->>LF: Create Root Span
Note over LF: 🆔 Trace ID: abc123
VA->>+LLM: Process Query
Note right of VA: 🎯 Intent Classification
LLM-->>-VA: Intent: order_status
VA->>+LE: Log Generation Span
LE->>LF: Record LLM Call
Note over LF: 🎯 Tokens: 150→45<br/>💰 Cost: $0.001<br/>⏱️ Latency: 0.5s
VA->>VA: Query Database
Note right of VA: 🐌 Bottleneck: 2.1s
VA->>+LE: Log Tool Span
LE->>LF: Record DB Query
Note over LF: ⏱️ Duration: 2.1s<br/>✅ Status: Success
VA->>+LLM: Format Response
LLM-->>-VA: Formatted Answer
VA->>+LE: Log Final Generation
LE->>LF: Complete Trace
Note over LF: ⏱️ Total: 4.2s<br/>💰 Cost: $0.002
VA-->>-U: "Your order #12345 is shipped! 🚚"
LF->>LF: Run Evaluations
Note over LF: ⭐ Quality: 0.95<br/>🎯 Relevance: 1.0<br/>🛡️ Toxicity: 0.0
D->>LF: View Dashboard
LF-->>D: Traces, Metrics, Alerts
Note over D: 🔍 Identify bottleneck<br/>⚡ Add caching
`} />
This diagram shows the complete flow from user query to developer insights, demonstrating how every step gets automatically traced and evaluated.
## Real Life Scenarios
### Scenario 1: Performance Debugging
**Problem:** Users report "system is slow".
**What You See in Langfuse:**
- Dashboard shows average response time is 8 seconds (should be 2 seconds)
- In traces you see external API calls taking 6 seconds
- You know which API is the problem
**Solution:** Add caching and reduce response time by 75%.