lmnr-ai/lmnr

What it solves

Laminar provides a centralized observability platform for AI agents, helping developers debug, monitor, and evaluate the performance of their LLM-powered applications. It addresses the complexity of tracking agentic workflows, which often involve multiple steps, tool calls, and non-deterministic outputs.

How it works

Laminar uses an OpenTelemetry-native SDK that can be integrated into a project with a single line of code to automatically trace LLM calls from providers like OpenAI, Anthropic, and Gemini, as well as frameworks like LangChain and Vercel AI SDK. It captures traces, spans, and metrics, which are then visualized in a dashboard. The system is written in Rust for high performance, featuring 20x trace compression and a custom real-time engine for viewing traces as they happen.

Who it’s for

It is designed for developers building AI agents and LLM applications who need a detailed view of their agent's internal logic, performance metrics, and evaluation datasets.

Highlights

• Tracing: Automatic instrumentation for major LLM providers and agent frameworks.
• Signals: Ability to describe agent behaviors to track in plain English and receive Slack notifications when they occur.
• Evals: An extensible SDK and CLI for running evaluations locally or in CI/CD pipelines.
• MCP/CLI Access: SQL-based querying of traces and metrics, allowing coding agents to help debug issues.
• Data Annotation: UI for fast data annotation and dataset creation for evaluations.
• Custom Dashboards: SQL-powered dashboard builder for metrics and events.