LangWatch Docker Images

LangWatch is composed of three main Docker images that work together to provide the complete LLM observability platform. Each image serves a specific purpose and exposes different endpoints for various functionalities.

Image Overview

1. langwatch/langwatch - Main Application

The primary LangWatch application that provides the web interface and core functionality. Purpose:
  • Web-based user interface for LangWatch
  • API endpoints for trace collection and management
  • User authentication and session management
  • Dashboard and analytics interface
  • Background workers for processing traces and evaluations
Key Features:
  • Trace collection and storage
  • Real-time monitoring dashboard
  • User management and authentication
  • API endpoints for SDK integration
  • Background job processing

2. langwatch/langwatch_nlp - NLP Backend Service

The NLP backend service that handles advanced language processing tasks and optimization studio workflows. Purpose:
  • Optimization studio workflows
  • Advanced NLP processing
  • Topic clustering and analysis
Key Features:
  • Optimization studio backend
  • Topic clustering service
  • Advanced NLP processing

3. langwatch/langevals - Evaluation Service

The dedicated evaluation service that provides built-in evaluators and evaluation processing capabilities. Purpose:
  • Built-in evaluator execution
  • Evaluation result processing
  • Model-based evaluations
  • Safety and quality checks
  • Evaluation API endpoints
Key Features:
  • Comprehensive evaluator library
  • LLM-as-a-judge evaluations
  • Safety evaluators (content safety, jailbreak detection)
  • Quality evaluators (relevance, accuracy, etc.)
  • RAG-specific evaluators

Service Communication

The three services communicate with each other through HTTP endpoints:

Running from a docker image

For detailed instructions on how to deploy LangWatch using Docker or Kubernetes, see our deployment guides: