/ docs

Langfuse → Slack connector setup

This connector sends Slack alerts when Langfuse traces fail, cross score thresholds, or show unusual latency — so your team sees issues fast without watching dashboards all day.

Self-hosted setup

Clone the repo or download the connector

Download the Langfuse → Slack connector package from PlugFlow, or clone the repository if you purchased the self-hosted version.

Copy .env.example to .env and fill in your keys

cp .env.example .env

Add your Langfuse API keys, Slack webhook URL, and any custom thresholds you want to use for alerts.

Run the connector

node index.js

python main.py

Use the runtime that matches the connector package you downloaded. Once started, the process will begin polling Langfuse automatically.

Verify with a test ping

Trigger a test event in Langfuse or temporarily lower your alert threshold. You should see a Slack message land in your configured channel within the next poll cycle.

Prerequisites

•Langfuse account
•Langfuse public + secret API keys
•Slack Incoming Webhook URL
•A server or runtime for the self-hosted option

Hosted setup

Want zero setup? Email us at hello@plugflow.io with your Langfuse API key and Slack webhook URL — we'll have you live in under 24 hours.

Email PlugFlow

Configuration reference

These environment variables control how the connector authenticates and when it sends alerts.

Variable	Default	Description
LANGFUSE_BASE_URL	https://cloud.langfuse.com	Your Langfuse instance URL. Use your self-hosted URL if you're not on Langfuse Cloud.
LANGFUSE_PUBLIC_KEY	—	Langfuse public API key used to authenticate requests.
LANGFUSE_SECRET_KEY	—	Langfuse secret API key used alongside the public key.
SLACK_WEBHOOK_URL	—	Incoming Slack webhook URL for the channel where alerts should be posted.
POLL_INTERVAL_SECONDS	60	How often the connector checks Langfuse for new traces, scores, and failures.
ALERT_SCORE_THRESHOLD	0.7	Minimum score threshold that triggers an alert. Lower scores will be posted to Slack.
ALERT_LATENCY_MS	5000	Latency threshold in milliseconds that triggers a slow-trace alert.
NODE_ENV	production	Runtime mode for the connector process.

FAQ

Common setup questions from teams rolling out alerts for the first time.

How often does it poll?

By default the connector polls Langfuse every 60 seconds. You can adjust this with POLL_INTERVAL_SECONDS if you want faster or slower checks.

What events trigger alerts?

The default setup watches for low evaluation scores, failed traces, and unusually high latency. You can tune thresholds to match your own alerting policy.

How do I update thresholds?

Edit your .env file and restart the connector. Update ALERT_SCORE_THRESHOLD for quality issues and ALERT_LATENCY_MS for slow responses.