This guide assumes you have Python 3.9+ installed. If you don’t have an Omium account yet, sign up here. Prerequisites
Step 1: Install the SDK
Verify installation with omium --version
Step 2: Initialize Omium
You’ll be prompted to enter your API key. Config is saved to ~/.omium/config.json. import omium
omium.init(api_key="om_xxx")
Step 3: Instrument Your Framework
LangGraph
CrewAI
Manual Tracing
import omium
omium.init(api_key="om_xxx")
omium.instrument_langgraph()
# Your LangGraph code works unchanged
from langgraph.graph import StateGraph
graph = StateGraph(...)
app = graph.compile()
result = app.invoke({"input": "Hello"})
import omium
omium.init(api_key="om_xxx")
omium.instrument_crewai()
# Your CrewAI code works unchanged
from crewai import Crew, Agent, Task
crew = Crew(agents=[...], tasks=[...])
result = crew.kickoff()
import omium
omium.init(api_key="om_xxx")
@omium.trace("process_data")
def process_data(data):
return transformed_data
@omium.checkpoint("important_step")
def important_step():
return result
Step 4: View in Dashboard
After running your workflow, all traces and checkpoints appear automatically in the dashboard.
- Executions - All workflow runs with status
- Traces - Step-by-step execution details
- Checkpoints - Recovery points for failed runs
- Replay - Re-run failed executions from checkpoints
What’s Next?
