mirror of
https://github.com/langflow-ai/langflow.git
synced 2026-07-25 00:59:07 +08:00
71fbf5ff00
* version-1.9.0-docs * [autofix.ci] apply automated fixes * [autofix.ci] apply automated fixes (attempt 2/3) --------- Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
106 lines
5.5 KiB
Plaintext
106 lines
5.5 KiB
Plaintext
---
|
|
title: Traceloop
|
|
slug: /integrations-instana-traceloop
|
|
description: Instrument Langflow with the Traceloop SDK, and export traces and metrics to Instana using OpenTelemetry.
|
|
---
|
|
|
|
Traceloop SDK is a lightweight instrumentation toolkit designed for LLM applications.
|
|
It enables developers to automatically capture and export traces, metrics, and key observability signals from their LLM-powered workflows.
|
|
|
|
When combined with Instana, the exported telemetry data from Traceloop provides end-to-end visibility, allowing users to visualize traces, analyze performance bottlenecks, and ensure reliable operation of LLM-driven applications.
|
|
|
|
This guide demonstrates how to integrate the Instana observability platform with your Langflow application using the Traceloop SDK so you can monitor and analyze LLM performance.
|
|
|
|
## Prerequisites
|
|
|
|
- Create a [Traceloop API key](https://app.traceloop.com/settings/api-key)
|
|
- Create an [Instana endpoint and Instana key](https://www.ibm.com/docs/en/instana-observability/1.0.302)
|
|
- [Install Langflow](/get-started-installation)
|
|
|
|
## Configure environment variables
|
|
|
|
1. In the root folder of your Langflow application, edit your existing Langflow `.env` file or create a new one.
|
|
|
|
2. Enter the following environment variables, and then replace the placeholders with the values for your deployment or requirements:
|
|
|
|
```text
|
|
TRACELOOP_API_KEY=tl_dummy_1234567890abcdef1234567890abcdef
|
|
TRACELOOP_BASE_URL=https://otlp-magenta-saas.instana.rocks:4318
|
|
TRACELOOP_HEADERS="x-instana-key=INSTANA_KEY"
|
|
OTEL_EXPORTER_OTLP_INSECURE=false
|
|
TRACELOOP_METRICS_ENDPOINT=HOST:8000
|
|
TRACELOOP_METRICS_ENABLED=true
|
|
OTEL_METRIC_EXPORT_INTERVAL=10000
|
|
```
|
|
|
|
Set the necessary values for each environment variable:
|
|
|
|
- **`TRACELOOP_API_KEY`**: A Traceloop API key to authenticate your application with Traceloop's monitoring service.
|
|
You can get this from your Traceloop account dashboard.
|
|
|
|
This integration uses the Traceloop SDK for instrumentation, which requires a Traceloop API key to initialize properly.
|
|
If you don't have a Traceloop API key, you can proceed with the placeholder API key in the preceding example.
|
|
|
|
- **`TRACELOOP_BASE_URL`**: The Instana endpoint URL for telemetry data collection, which is your Instana backend endpoint, such as `https://otlp-magenta-saas.instana.rocks:4318`.
|
|
You can get this from your Instana configuration or by contacting your Instana administrator.
|
|
|
|
- **`TRACELOOP_HEADERS`**: Authentication headers for Instana data collection. Set this to `"x-instana-key=INSTANA_KEY"`, replacing `INSTANA_KEY` with the Instana key from your Instana setup.
|
|
|
|
- **`OTEL_EXPORTER_OTLP_INSECURE`**: Security setting for OpenTelemetry Protocol connections. Set to `false` for secure HTTPS/TLS connections. This is recommended for production Instana SaaS endpoints. Set to `true` for insecure HTTP connections during local development.
|
|
|
|
- **`TRACELOOP_METRICS_ENDPOINT`**: Separate metrics endpoint configuration, in the form of `OTEL_DC_LLM_HOST:8000`. Typically set to `host.docker.internal:8000` for Docker environments. Adjust the host and port based on your deployment setup.
|
|
|
|
- **`TRACELOOP_METRICS_ENABLED`**: Boolean to enable metrics collection. Set to `true` to activate metrics gathering.
|
|
|
|
- **`OTEL_METRIC_EXPORT_INTERVAL`**: Interval in milliseconds for metrics export. Set to `10000` for 10-second export intervals, or adjust based on your monitoring requirements.
|
|
|
|
3. Make sure the OpenTelemetry Data Collector (OTel DC) is running and correctly configured.
|
|
Open your Collector's `config.yaml` file, enter the following configuration, and then replace the placeholder values with the values from your data collector setup or requirements:
|
|
|
|
```yaml
|
|
llm.application: "LLM_DC"
|
|
instances:
|
|
- otel.agentless.mode: true
|
|
# Example endpoint: https://otlp-magenta-saas.instana.rocks:4318
|
|
otel.backend.url: "INSTANA_ENDPOINT"
|
|
otel.backend.using.http: false
|
|
callback.interval: 10
|
|
otel.service.name: "DC1"
|
|
otel.service.port: 8000
|
|
currency: "USD"
|
|
```
|
|
|
|
This configuration enables the OTel Collector to operate in agentless mode and route telemetry data to your Instana backend with proper service identification and collection intervals for effective monitoring integration.
|
|
|
|
## Start Langflow with Traceloop environment variables
|
|
|
|
Launch your Langflow application with your `.env` file:
|
|
|
|
```bash
|
|
uv run langflow run --env-file .env
|
|
```
|
|
|
|
Traceloop automatically begins monitoring and collecting telemetry data from your LLM applications.
|
|
|
|
## Verify the integration
|
|
|
|
To verify that observability is working correctly:
|
|
|
|
1. Run a flow in Langflow to generate traffic.
|
|
2. To view traces in Instana, open Instana and click **Applications**.
|
|
3. In **Services**, search for `Langflow`.
|
|
4. Click **Langflow** to view and analyze the associated calls.
|
|
|
|
|
|
|
|
5. To view metrics in Instana, open Instana and click **Infrastructure**.
|
|
6. In **Analyze Infrastructure**, click **Otel LLMonitor**.
|
|
7. To view your Metrics dashboard, click `LLM:DC1@your_machine_name.local`.
|
|
|
|
|
|
|
|
## See also
|
|
|
|
* [Traceloop documentation](https://www.traceloop.com/docs/introduction)
|
|
* [Instana setup documentation](https://www.ibm.com/docs/en/instana-observability/1.0.300?topic=started-instana-setup)
|
|
* [Otel DC setup documentation](https://www.ibm.com/docs/en/instana-observability/1.0.300?topic=started-install-otel-data-collector-llm-odcl) |