Spanlens: Open-source LLM observability and monitoring

Spanlens: Open-Source LLM Observability and Monitoring

Spanlens is an open-source proxy designed for developers working with large language models (LLMs). It provides essential observability and monitoring capabilities for your AI applications, acting as a central point for logging, cost tracking, and agent tracing. By integrating Spanlens, you gain deeper insights into your LLM interactions, enabling more efficient development, debugging, and cost management.

What Spanlens Does

Spanlens functions as a self-hostable proxy that sits between your application and LLM providers such as OpenAI, Anthropic, and Gemini. It intercepts and logs all requests and responses, providing a detailed audit trail of your LLM usage. This logging capability is crucial for understanding model behavior, identifying errors, and analyzing performance. Furthermore, Spanlens offers robust cost tracking features, allowing you to monitor expenditure across different LLM calls and providers, which is vital for budget control in AI development.

Key Features

• Open-Source: Full control and transparency over your LLM observability stack.
• LLM Provider Support: Compatible with OpenAI, Anthropic, and Gemini APIs.
• Request/Response Logging: Detailed capture of all LLM interactions for debugging and analysis.
• Cost Tracking: Monitor and manage expenses associated with LLM API calls.
• Agent Tracing: Visualize the execution flow of complex AI agents.
• Self-Hostable: Deployable via Docker for maximum flexibility and data privacy.

Who Spanlens is For

Spanlens is an indispensable tool for AI developers, ML engineers, and teams building applications that leverage LLMs. If you are deploying AI agents, integrating LLM APIs into your products, or managing multiple LLM services, Spanlens provides the necessary visibility to ensure smooth operation and cost-effectiveness. Its self-hostable nature makes it ideal for projects with strict data privacy requirements or those seeking to avoid vendor lock-in. Developers focused on optimizing LLM performance, debugging complex agent behaviors, and maintaining budget control will find Spanlens a critical component of their MLOps toolkit.