integrations

Integration Samples

Most AI applications interact with LLMs through provider-specific SDKs such as OpenAI or Anthropic. These samples demonstrate how to instrument those API calls with LayerLens tracing and run post-completion evaluations -- without modifying your existing provider integration code. This enables teams to add observability and quality evaluation to production LLM calls with minimal effort.

Prerequisites

pip install layerlens --index-url https://sdk.layerlens.ai/package
export LAYERLENS_STRATIX_API_KEY=your-api-key

Each sample requires the corresponding provider SDK and API key:

Sample	Additional Dependencies
openai_traced.py	pip install openai and OPENAI_API_KEY
anthropic_traced.py	pip install anthropic and ANTHROPIC_API_KEY
openai_instrumented.py	pip install openai and OPENAI_API_KEY
langchain_instrumented.py	pip install openai langchain-openai langchain-core and OPENAI_API_KEY

Quick Start

Start with openai_traced.py if you use OpenAI, or anthropic_traced.py for Anthropic:

export OPENAI_API_KEY=your-openai-key
python openai_traced.py

Expected output: the script makes an LLM API call, automatically captures the interaction as a LayerLens trace (including prompt, completion, token usage, and latency), runs a post-completion evaluation, and prints the trace ID and evaluation scores.

Samples

File	Scenario	Description
openai_traced.py	Teams using OpenAI GPT models in production	Wraps an OpenAI chat completion call with LayerLens tracing, captures the full request/response cycle, and runs a post-completion evaluation with safety and relevance judges.
anthropic_traced.py	Teams using Anthropic Claude models in production	Wraps an Anthropic message API call with LayerLens tracing, captures the full request/response cycle, and runs a post-completion evaluation with safety and relevance judges.
openai_instrumented.py	Teams adopting auto-instrumentation for OpenAI	Uses instrument_openai() and the @trace decorator from layerlens.instrument to automatically capture all OpenAI calls as spans -- no manual trace upload needed.
langchain_instrumented.py	Teams running LangChain LCEL chains	Uses LangChainCallbackHandler from layerlens.instrument to auto-capture chain execution, LLM calls, and prompt/completion data from LangChain pipelines.

Expected Behavior

Each sample makes a real API call to the respective provider, so valid provider credentials are required. The trace is uploaded to your LayerLens workspace and the evaluation results are printed to stdout. Both samples follow the same pattern, making it straightforward to adapt the approach to additional providers.