Datasets:
roanbrasil
/
devops-gitops-corpus

Dataset card Data Studio Files
xet
 Community
1
text
stringlengths
0
59.1k
chart={`
%%{init: {'theme':'base', 'themeVariables': {'primaryColor': '#10b981', 'primaryTextColor': '#10b981', 'primaryBorderColor': '#10b981', 'lineColor': '#10b981', 'secondaryColor': '#ecfdf5', 'tertiaryColor': '#d1fae5', 'background': '#ffffff', 'mainBkg': '#ecfdf5', 'secondBkg': '#d1fae5', 'tertiaryBkg': '#a7f3d0'}}}%%
sequenceDiagram
participant Dev as Developer
participant App as Your App
participant VAI as Vercel AI SDK
participant OTel as OpenTelemetry
participant VE as VoltAgent Exporter
participant VP as VoltAgent Platform
participant Console as VoltOps Dashboard
Dev->>App: Write AI code with experimental_telemetry
App->>VAI: generateText() with telemetry enabled
VAI->>VAI: Execute LLM call
VAI->>OTel: Send telemetry data
alt Tool Usage
VAI->>VAI: Execute tool calls
VAI->>OTel: Send tool execution data
end
OTel->>VE: Forward telemetry data
VE->>VP: Send to VoltAgent Platform
VP->>VP: Process & store data
Dev->>Console: Open VoltOps Dashboard
Console->>VP: Request agent data
VP->>Console: Return real-time data
Console->>Dev: Display agent activity, flows, metrics
Note over Dev,Console: Real-time monitoring & debugging
`}
/>
This flow happens automatically once you add the `experimental_telemetry` flag. No additional code changes needed!
## Final Words
AI development is moving very fast. But we are still at the _stone age_ of monitoring.
The traditional APM tools are not enough for AI-specific needs. AI systems have their own share of challenges - tool calls, agent interactions, conversation flow, multi-modal data.
VoltAgent monitoring tries to fill this gap. And because it's framework-agnostic, you can use it on _your existing code_ with minimal alterations.
Test it and give feedback. We always discuss new features in our community ([Discord](https://s.voltagent.dev/discord)).
_Seeing what your AI agents do is no longer a dream. Go and see what your agents are really doing!_ ⚡
<|endoftext|>
# source: VoltAgent__voltagent/website/blog/2025-07-30-claude-code-agents/index.md type: docs
---
title: Claude Code Subagents - Taking AI Pair Programming to the Next Level
description: Learn how Claude Code's subagent system lets you delegate complex coding tasks to specialized AI assistants for better results.
slug: claude-code-subagents
image: https://cdn.voltagent.dev/2025-07-30-claude-code-agents/social.png
authors: necatiozmen
tags: [development-tools]
---
import ZoomableMermaid from '@site/src/components/blog-widgets/ZoomableMermaid';
## The Problem with One-Size-Fits-All AI Assistants
Last month I was stuck debugging a memory leak at 11 PM. Turned to my AI assistant and got a generic "check for circular references" response. Then asked it to write tests for the fix. Suddenly it forgot we were even talking about memory leaks.
This happens all the time, and here's why:
**Context Window Limitations**: Ever tried explaining a microservices architecture to ChatGPT? By the time you get to service #3, it's forgotten service #1 exists. I once spent 20 minutes re-explaining our auth flow because the AI kept suggesting I "just use JWT" when we already were.
**Jack of All Trades, Master of None**: My AI can write React hooks, debug Python, optimize SQL, and review Go code. Sounds amazing until you realize it's about as deep as a Twitter thread on each topic. Asked for advanced React performance tips? Got "use useMemo." Thanks, I guess?
**Task Complexity**: Real tasks are messy. "Refactor auth but keep the old API working and make it faster" isn't a single task - it's architecture, implementation, testing, and optimization all rolled into one. Most AIs handle this like I handle juggling - badly.
**Context Pollution**: Yesterday I was debugging WebSocket connections. Today I'm building a REST API. Guess what my AI keeps suggesting? "Have you checked your WebSocket implementation?" No, that was yesterday's problem, keep up!
## Enter Claude Code: Your Terminal-Native AI Assistant
When Anthropic announced Claude Code runs in the terminal, I rolled my eyes. "Great, another tool pretending the terminal is cool again."
Then I tried it. Holy crap, they're onto something.
![claude code general approach](https://cdn.voltagent.dev/2025-07-30-claude-code-agents/general.png)
**Direct File Editing**: No more copy-paste dance. Claude Code just... edits your files. Shows you the diff. Waits for approval. Like having a junior dev who actually knows how to use git.
```bash
# Give Claude Code a task
$ claude "Add proper error handling to all API endpoints"
# Claude Code will:
# 1. Find relevant files
# 2. Add error handling
# 3. Show you the changes
# 4. Apply them with your approval
```
**Command Execution**: Friday afternoon. Test suite failing. Instead of the usual ChatGPT back-and-forth ("run this," "what happened?", "try this"), I just said "fix the tests." Claude Code ran them, found the issue, fixed it, verified it worked. I literally made coffee while it handled everything.
```bash