SuperAPI AI Directory
Partner with us中文AI newsFavorites
All toolsAI developer platformslangtrace

langtrace

Langtrace is an open-source end-to-end observability tool based on OpenTelemetry, designed for LLM applications, providing real-time tracing, evaluation, and metrics. It supports mainstream LLMs, LLM frameworks, vector databases, and more, and can be integrated via the TypeScript or Python SDK.

Open website

About this tool

Langtrace is an open-source LLM application observability software used to capture, debug, and analyze traces and metrics from applications that use LLM APIs, vector databases, and LLM frameworks.

Main capabilities:

  • Real-time monitoring: trace LLM API calls, vector operations, and framework usage.
  • Performance insights: analyze latency, cost, and usage patterns.
  • Debugging tools: trace and debug LLM application workflows.
  • Analytics: provide detailed metrics and visualizations.
  • Self-hosting option: can be deployed on your own infrastructure.
  • OpenTelemetry support: generate traces based on the OTEL standard; the project is developing semantic conventions, and the relevant definitions are maintained in the repository.

How to use: A TypeScript SDK and a Python SDK are provided.

  • Cloud service (Langtrace Cloud): after registering, create a project (projects are used to store traces and metrics generated by the application), generate an API key within the project, then install and initialize the Langtrace SDK in the application and configure it with that API key.
  • Self-hosting: three services need to be running, including a Postgres database and a Clickhouse database; by default, the application can be accessed at http://localhost:3000 . The documentation provides multiple deployment options and configurations, and the .env file can be modified. Using the -v flag can remove containers and volumes. When self-hosting the OSS client, Langtrace does not collect any telemetry data, and data does not leave the server.

Supported integrations: automatically capture traces for the following providers and frameworks, including vector databases (the specific list is not expanded in the materials).

System architecture: uses NextJS as the frontend and API; PostgresDB as metadata storage; Clickhouse DB for storing spans, metrics, logs, and traces.

Limitations and notes:

  • The specific list of providers and frameworks for integrations is not fully provided in the materials.
  • Self-hosting requires additionally deploying Postgres and Clickhouse databases.
  • The trace semantic conventions are still under development.
  • Pricing: Langtrace cloud is currently free of charge, mainly for collecting feedback; users will be notified if it is monetized in the future.
  • License: the Langtrace application (this repository) uses the AGPL 3.0 license, and the SDK uses the Apache 2.0 license.
  • Security issues can be emailed to security@scale3labs.com.

Related tools

Browse category