Debug pipelines
before your
users do.
PipelineHub captures step-by-step snapshots of your data pipelines and LangChain agents, detects silent failures automatically, and shows you exactly what changed between runs.
Open source and free to self-host. Cloud dashboard from $19/month.
Built with pipelinehub
Run diff
See exactly what changed
between any two runs.
Pick any two runs and PipelineHub shows you a step-by-step breakdown of every metric that shifted. Row counts, null percentages, token usage, tool calls. Green means improved, red means regression. No guessing what broke.
Dashboard
Everything in one place.
Runs, step snapshots, anomaly detection, SLA tracking, schema changes, and agent tool call traces. All visible in real time as your pipeline or agent executes.
Up and running in 5 minutes.
No agents to configure. No YAML. No infra to manage. PipelineHub is a library. It ships with your pipeline and your LangChain agent.
Install the library
One pip install. Zero new dependencies to manage. pipelinehub uses only stdlib for cloud sync — no httpx, no requests.
pip install pipelinehub
What pipelinehub tracks.
Every pipeline run and agent execution is recorded. Every anomaly is explained. No manual instrumentation required.
Step snapshots
Every pipeline step captures before/after snapshots — row count, null %, column list, dtypes, numeric stats. Compare any two runs side by side.
Anomaly detection
Automatically flags row count drops >50%, null percentage spikes >20pp, and schema changes between runs. No thresholds to configure.
SLA tracking
Tracks rolling average duration per step. Fires a slow_step alert when any step takes more than 3× its historical average. Catch regressions before they become incidents.
Schema change detection
Detects columns added, removed, or retyped between runs. Filters noise (int64↔float64 NaN coercion). Shows exactly which columns changed and how.
LangChain agent tracing
Track every LLM call, tool invocation, and chain step in your LangChain agents. See token usage, latency, and tool call sequences per run.
Agent cost tracking
Know exactly what each agent run costs. Track spend per step, per tool, and per model. Get alerted when costs spike beyond your baseline.
Works with your
existing pipeline.
pipelinehub is a decorator-based library. Wrap your existing step functions — no refactoring required. Supports pandas, polars (including LazyFrame), numpy, and plain Python.
- ✓pandas & polars — including LazyFrame (auto-collected before profiling)
- ✓Zero new runtime dependencies (stdlib only for cloud sync)
- ✓pip install pipelinehub — that's the only setup step
import pandas as pd
from pipelinehub import DataPipeline
pipeline = DataPipeline(name="sales-etl")
@pipeline.add_step
def extract(df):
return pd.read_csv("sales.csv")
@pipeline.add_step
def clean(df):
return df.dropna(subset=["customer_id"])
@pipeline.add_step
def enrich(df):
df["revenue"] = df["price"] * df["qty"]
return df
pipeline.execute(pd.DataFrame())
# Every step is now visible in your dashboard.Pricing that fits
how you work.
The open source library is free forever. Upgrade to cloud when you need persistent history, email alerts, agent cost tracking, or team access.
Open Source
Self-hosted. Local SQLite. No cloud required.
Everything included:
- ✓Self-hosted, local SQLite storage
- ✓Unlimited pipelines & runs
- ✓Full anomaly detection
- ✓Step-by-step snapshots
- ✓Run diff view
- ✓SLA tracking
- ✓Schema change detection
- ✓1 user
- ✓Community support
Solo ✦
For individual engineers who want the cloud dashboard.
14 days free, no credit card
Everything in Open Source, plus:
- ✓Cloud dashboard
- ✓Unlimited pipelines & runs
- ✓30-day run history
- ✓Email alerts on anomalies
- ✓1 user
Team
For data teams sharing pipelines and debugging together.
14 days free, no credit card
Everything in Solo, plus:
- ✓Shared team dashboard
- ✓Unlimited users
- ✓90-day run history
- ✓Email alerts with team routing
- ✓Pipeline-level access controls
Business
For companies with compliance and SLA requirements.
Everything in Team, plus:
- ✓Unlimited run history
- ✓SSO (SAML)
- ✓Audit logs
- ✓Priority support
- ✓SLA guarantee