LangChain Visualizer

Visualization and debugging tool for LangChain workflows

No commits in 34 months — may not be actively maintained. See maintained alternatives →

open-sourceno-code-agent-builders agent-frameworks

Visit Website View on GitHub

738

Stars

+-0

Stars/month

Commits (90d)

Releases (6m)

Star Growth

Overview

LangChain Visualizer是一个专门为LangChain框架设计的可视化调试工具，改编自Ought's ICE visualizer。该工具提供了一个美观的Web界面，用于实时观察和分析LangChain工作流的执行过程。开发者可以查看发送给大语言模型（LLM）的完整提示文本，通过颜色编码区分硬编码内容和模板变量替换部分，追踪函数调用堆栈的执行流程。工具支持成本监控功能，可以显示每次LLM调用的费用以及整个运行过程的总成本（特别针对OpenAI的text-davinci-003模型）。该工具对于理解复杂的LangChain agent行为、优化提示模板设计以及调试多步推理流程特别有价值。它支持Jupyter notebook环境，并且可以可视化文档分块和嵌入向量的处理过程，为LangChain开发者提供了强大的调试和分析能力。

Deep Analysis

Key Differentiator

vs LangChain built-in tracing: colored prompt highlighting showing hardcoded vs templated sections — adapted from Ought's ICE visualizer for superior prompt debugging experience

⚡ Capabilities

• Visual debugging for LangChain agent workflows
• Full prompt text inspection for every LLM interaction
• Color-coded distinction between hardcoded and templated prompt sections
• LLM call cost tracking (OpenAI text-davinci-003)
• Execution flow visualization in browser UI
• Jupyter notebook compatibility

🔗 Integrations

LangChainOpenAIGoogle Search APIvcr-langchain (replay)Jupyter

✓ Best For

✓ Debugging LangChain agent prompt construction
✓ Understanding LLM call costs during development
✓ Inspecting tool execution timing and flow

✗ Not Ideal For

✗ Production monitoring or observability
✗ Non-LangChain agent frameworks
✗ Automated testing pipelines

Languages

Python

Deployment

pip installbrowser-based UI (auto-opens)Jupyter notebook integration

⚠ Known Limitations

⚠ Many LangChain features not yet supported for visualization
⚠ Requires manual integration into codebases
⚠ Cost tracking limited to specific OpenAI models
⚠ Linux requires libyaml-dev system dependency

Pros

+ 提供实时可视化界面，能够直观观察LangChain agent的完整执行过程
+ 通过颜色编码清晰区分提示中的硬编码部分和动态模板替换内容
+ 支持成本监控和函数调用栈追踪，便于性能分析和成本控制

Cons

- 仅支持LangChain框架，无法用于其他LLM框架的可视化
- 要求在Python入口文件的第一行导入，对代码结构有特定要求

Use Cases

• 调试复杂的LangChain agent行为，理解多步推理和工具调用流程
• 优化提示模板设计，分析不同模板变量对LLM响应的影响
• 监控和分析LLM API调用成本，优化应用的经济效益

Getting Started

1. 安装库：pip install langchain-visualizer 2. 在Python入口文件第一行添加import langchain_visualizer 3. 编写异步函数包装LangChain工作流，调用langchain_visualizer.visualize(your_function)启动可视化