langfuse-observation-view

name	langfuse-observation-view
description	View Langfuse observation (Generation/Span) details. Use when checking specific LLM call input/output, debugging issues, or analyzing costs.
license	MIT
compatibility	Node.js 18+ (native fetch required)
metadata	{"author":"neuradex","category":"observability"}
allowed-tools	["Bash(npx tsx *scripts/langfuse-observation-view.ts*)"]

Langfuse Observation View

Display detailed information for a specific Observation (Generation/Span).

Setup

Set the following environment variables before use:

Variable	Required	Description
LANGFUSE_PUBLIC_KEY	Yes	Langfuse public key
LANGFUSE_SECRET_KEY	Yes	Langfuse secret key
LANGFUSE_HOST or LANGFUSE_BASE_URL	No	Langfuse host URL (default: https://us.cloud.langfuse.com)

When to Use

• Checking specific LLM call details
• Viewing input/output
• Checking costs and token usage
• Viewing prompts used

Commands

1. Get Observation Details

npx tsx scripts/langfuse-observation-view.ts <observationId>

2. Show Full Input/Output

npx tsx scripts/langfuse-observation-view.ts <observationId> --full

3. JSON Output

npx tsx scripts/langfuse-observation-view.ts <observationId> --json

Output Example

Langfuse Observation Detail
===========================

ID:          obs_abc123
Trace ID:    trace_xyz789
Type:        GENERATION
Name:        chat
Start Time:  2026-01-21 10:30:45
Duration:    2.35s
Level:       DEFAULT

--- Model Info ---
Model:       qwen/qwen3-32b
Parameters:  {"temperature":0.7}

--- Usage & Cost ---
Input Tokens:   1500
Output Tokens:  350
Total Tokens:   1850
Input Cost:     $0.000435
Output Cost:    $0.000207
Total Cost:     $0.000642

--- Prompt ---
Prompt:      librarian-system (v3)
Prompt ID:   prompt_def456

--- Input ---
[{"role":"system","content":"You are..."}]

--- Output ---
Here are the search results...

Fields

Field	Description
Type	GENERATION (LLM call), SPAN (processing span), EVENT (event)
Duration	Processing time
Model	Model used
Input/Output Tokens	Token usage
Calculated Cost	Computed cost
Prompt	Langfuse prompt used