Openlayer

import Tabs from '@theme/Tabs'; import TabItem from '@theme/TabItem';

Openlayer is a testing and evaluation platform for LLM applications. It provides comprehensive observability, testing, and monitoring capabilities to help you ship high-quality AI systems with confidence.

You can configure Langflow to collect tracing data about your flow executions and automatically send the data to Openlayer for analysis, monitoring, and evaluation.

Prerequisites

An Openlayer account
A running Langflow server with a flow that you want to trace
An Openlayer inference pipeline

:::tip If you need a flow to test the Openlayer integration, see the Langflow quickstart. :::

Set Openlayer credentials as environment variables

Get your Openlayer API key from your Openlayer account.
Create an inference pipeline in Openlayer and copy the pipeline ID.
Set your Openlayer credentials as environment variables in the same environment where you run Langflow.

In the following examples, replace YOUR_API_KEY and YOUR_PIPELINE_ID with your actual Openlayer credentials.
<Tabs> <TabItem value="linux-macos" label="Linux or macOS" default>
These commands set the environment variables in a Linux or macOS terminal session:
bash
```
export OPENLAYER_API_KEY="YOUR_API_KEY"
export OPENLAYER_INFERENCE_PIPELINE_ID="YOUR_PIPELINE_ID"
```
</TabItem> <TabItem value="windows" label="Windows">
These commands set the environment variables in a Windows command prompt session:
cmd
```
set OPENLAYER_API_KEY=YOUR_API_KEY
set OPENLAYER_INFERENCE_PIPELINE_ID=YOUR_PIPELINE_ID
```
</TabItem> </Tabs>

Start Langflow and view traces in Openlayer

Start Langflow in the same environment where you set the Openlayer environment variables:
bash
```
uv run langflow run
```
Run a flow in Langflow.

Langflow automatically collects and sends tracing data about the flow execution to Openlayer, including:
- Component inputs and outputs
- Execution timing and latency
- LLM calls and nested operations
- User and session context
View the collected data in your Openlayer dashboard.

Each flow execution appears as a trace with a hierarchical view of all components and their nested operations.

Advanced configuration

Flow-specific pipelines

You can configure different Openlayer inference pipelines for different flows using flow-specific environment variables:

bash

export OPENLAYER_PIPELINE_MY_FLOW_NAME="pipeline-id-1"
export OPENLAYER_PIPELINE_ANOTHER_FLOW="pipeline-id-2"

The flow name is converted to uppercase and non-alphanumeric characters are replaced with underscores. For example, "My Flow-Name" becomes OPENLAYER_PIPELINE_MY_FLOW_NAME.

JSON mapping

Alternatively, you can use a JSON mapping to configure multiple flows at once:

bash

export OPENLAYER_LANGFLOW_MAPPING='{"Flow Name 1":"pipeline-id-1","Flow Name 2":"pipeline-id-2"}'

</TabItem> <TabItem value="windows" label="Windows">

cmd

set OPENLAYER_LANGFLOW_MAPPING={"Flow Name 1":"pipeline-id-1","Flow Name 2":"pipeline-id-2"}

</TabItem> </Tabs>

Configuration priority

Openlayer configuration is resolved in the following order (highest priority first):

Flow-specific environment variable: OPENLAYER_PIPELINE_<FLOW_NAME>
JSON mapping: OPENLAYER_LANGFLOW_MAPPING
Default environment variable: OPENLAYER_INFERENCE_PIPELINE_ID

This allows you to set a default pipeline for all flows and override it for specific flows as needed.

Disable Openlayer tracing

To disable the Openlayer integration, remove the OPENLAYER_API_KEY environment variable, and then restart Langflow.

Features

The Openlayer integration automatically captures:

Component hierarchy: All flow components with parent-child relationships
LangChain callbacks: Nested LLM calls and tool executions appear within their parent components
Timing metrics: Start time, end time, and latency for each component
Inputs and outputs: Component inputs and outputs with automatic type conversion
User context: User ID and session ID propagation for better analytics
Error tracking: Errors and logs captured in component metadata