# Langtrace — Open Source AI Observability Platform

> Open-source observability for LLM apps. Trace OpenAI, Anthropic, and LangChain calls with OpenTelemetry-native instrumentation and a real-time dashboard.

## Install

Save in your project root:

## Quick Use

```bash
pip install langtrace-python-sdk
```

```python
from langtrace_python_sdk import langtrace
langtrace.init(api_key="your-key")

# All OpenAI/Anthropic calls are now auto-traced
from anthropic import Anthropic
client = Anthropic()
response = client.messages.create(
    model="claude-sonnet-4-20250514",
    max_tokens=100,
    messages=[{"role": "user", "content": "Hello"}],
)
```

## What is Langtrace?

Langtrace is an open-source observability platform for LLM applications. It auto-instruments calls to OpenAI, Anthropic, LangChain, LlamaIndex, and 20+ providers using OpenTelemetry — giving you traces, latency metrics, token usage, and cost tracking in a real-time dashboard.

**Answer-Ready**: Langtrace is an open-source AI observability platform that auto-instruments LLM calls to OpenAI, Anthropic, LangChain, and 20+ providers with OpenTelemetry-native tracing, latency metrics, and cost tracking.

**Best for**: Teams running LLM apps in production who need observability. **Works with**: OpenAI, Anthropic, LangChain, LlamaIndex, Cohere, Pinecone. **Setup time**: Under 2 minutes.

## Core Features

### 1. Auto-Instrumentation
One line to trace all LLM calls:

```python
langtrace.init()  # That's it — all calls auto-traced
```

Supported providers: OpenAI, Anthropic, Google, Cohere, Mistral, Groq, LangChain, LlamaIndex, Pinecone, ChromaDB, Weaviate.

### 2. OpenTelemetry Native
Export traces to any OTel-compatible backend:

```python
langtrace.init(
    api_key="langtrace-key",
    # Or export to your own collector
    custom_remote_exporter=OTLPSpanExporter(endpoint="http://localhost:4317"),
)
```

### 3. Real-Time Dashboard
Self-hosted or cloud dashboard showing:
- Request traces with full input/output
- Latency percentiles (p50, p95, p99)
- Token usage per model
- Cost breakdown per endpoint
- Error rates and patterns

### 4. Evaluation & Testing

```python
from langtrace_python_sdk import with_langtrace_root_span

@with_langtrace_root_span("test-summarization")
def test_summary_quality():
    result = summarize("long text...")
    # Trace includes test metadata
    return result
```

### 5. Prompt Management
Track prompt versions alongside traces:

```python
from langtrace_python_sdk import inject_additional_attributes

with inject_additional_attributes({"prompt_version": "v2.3", "experiment": "temp-0.7"}):
    response = client.messages.create(...)
```

## Self-Hosting

```bash
git clone https://github.com/Scale3-Labs/langtrace
docker compose up -d
# Dashboard at http://localhost:3000
```

## FAQ

**Q: How does it compare to LangFuse?**
A: Both are open-source LLM observability tools. Langtrace is OpenTelemetry-native (standard traces), LangFuse uses custom tracing. Langtrace has broader auto-instrumentation.

**Q: Does it add latency?**
A: Traces are sent asynchronously. Overhead is < 1ms per call.

**Q: Can I use it without the cloud?**
A: Yes, fully self-hostable with Docker Compose.

## Source & Thanks

> Created by [Scale3 Labs](https://github.com/Scale3-Labs). Licensed under AGPL-3.0.
>
> [Scale3-Labs/langtrace](https://github.com/Scale3-Labs/langtrace) — 3k+ stars

<!-- ZH -->

## 快速使用

```bash
pip install langtrace-python-sdk
```

一行代码自动追踪所有 LLM 调用。

## 什么是 Langtrace？

Langtrace 是开源 AI 可观测性平台，自动追踪 OpenAI、Anthropic、LangChain 等 20+ 供应商的 LLM 调用，基于 OpenTelemetry 标准。

**一句话总结**：开源 AI 可观测性平台，OpenTelemetry 原生追踪 LLM 调用，支持延迟指标和成本追踪。

**适合人群**：生产环境运行 LLM 应用需要可观测性的团队。

## 核心功能

### 1. 自动追踪
一行代码启用，支持 20+ 供应商。

### 2. OpenTelemetry 原生
导出到任何 OTel 兼容后端。

### 3. 实时仪表盘
延迟、token 用量、成本、错误率一览。

### 4. 可自托管
Docker Compose 一键部署。

## 常见问题

**Q: 和 LangFuse 比较？**
A: Langtrace 基于 OpenTelemetry 标准，自动追踪覆盖更广。

**Q: 延迟开销？**
A: 异步发送，每次调用 < 1ms。

## 来源与致谢

> [Scale3-Labs/langtrace](https://github.com/Scale3-Labs/langtrace) — 3k+ stars, AGPL-3.0

---
Source: https://tokrepo.com/en/workflows/a53444d6-2d55-4f59-ba6f-3b672d7ec458
Author: AI Open Source