Sepurux

Docs / Integrations

Install the recorder and capture traces automatically.

Debug a trace in the browser first, then use these integrations to capture traces in code and feed the same workflows into CI gates.

Install the recorderGate in CIAPI reference

Navigation

Available Integrations

All integrations are included in the sepurux package. Install once, capture traces automatically, then use CI to gate on reliability.

Install recorder
# Install the recorder for your stack
pip install sepurux
npm install @sepurux/recorder
go get github.com/sepurux/go-recorder@latest

OpenAI

Available

Wrap any openai.OpenAI client with instrument_openai. Every chat.completions.create call is recorded as an llm_call event with model, messages, output, and latency. Tool calls in the response are captured automatically.

  • llm_call with model + latency
  • tool_call for function calls in response
  • failure on API errors
pip install sepurux openai

LangChain

Available

Pass SepuruxCallbackHandler to any LangChain chain, agent, or LLM. LLM calls, tool call/result pairs, and the final agent output are all captured without changes to your chain code.

  • llm_call with model name + latency
  • tool_call and tool_result pairs
  • agent_output on finish
pip install sepurux langchain-core

OpenAI Agents SDK

Coming soon

Native tracing for the OpenAI Agents SDK — captures handoffs, tool use, and guardrail events from multi-agent pipelines.

CrewAI

Coming soon

Automatic trace recording for CrewAI agent crews — captures task assignments, tool executions, and crew-level outputs.

OpenAI

instrument_openai wraps any openai.OpenAI (or AsyncOpenAI) client. Pass it a recorder from client.trace() and every chat.completions.create call is captured automatically.

python
import openai
from sepurux import SepuruxClient
from sepurux.integrations.openai import instrument_openai

client = SepuruxClient.from_env()
openai_client = openai.OpenAI()

with client.trace(
    "customer_refund_flow",
    {"ticket_id": "t-101"},
    campaign_id=os.environ["SEPURUX_CAMPAIGN_ID"],
) as trace:
    ai = instrument_openai(openai_client, recorder=trace)
    response = ai.chat.completions.create(
        model="gpt-4o",
        messages=[{"role": "user", "content": "Classify this support ticket"}],
    )

print(trace.trace_id)
print(trace.run_id)

No code changes

Your existing chat.completions.create calls work as-is — just swap the client.

Async support

Call await ai.chat.completions.acreate(...) for AsyncOpenAI — same recording behavior.

Tool call capture

Function calls returned by the model are automatically recorded as tool_call events.

LangChain

SepuruxCallbackHandler hooks into LangChain's callback system. Pass it via config to any chain, agent, or LLM and it records everything without changes to your chain logic.

python
from sepurux.integrations.langchain import SepuruxCallbackHandler

handler = SepuruxCallbackHandler(
    "customer_refund_flow",
    {"ticket_id": "t-101"},
    campaign_id=os.environ["SEPURUX_CAMPAIGN_ID"],
)

# Pass to any LangChain chain, agent, or LLM
result = agent_executor.invoke(
    {"input": "Process refund for ticket t-101"},
    config={"callbacks": [handler]},
)

# Upload trace and start reliability run
handler.finish()
print(handler.trace_id)
print(handler.run_id)

# Or use as a context manager — finish() is called automatically
with SepuruxCallbackHandler("refund_flow", campaign_id="...") as handler:
    result = chain.invoke(inputs, config={"callbacks": [handler]})

Any runnable

Works with chains, agents, AgentExecutor, LLMs — anything that accepts callbacks.

Context manager

Use with SepuruxCallbackHandler(...) as handler: and finish() is called on exit.

Full event capture

LLM calls with latency, tool call/result pairs, and final agent output — all recorded.

Coming Soon

These integrations are in development. Star the repo or sign up for updates to be notified when they ship.

OpenAI Agents SDK

Capture handoffs, tool use, and guardrail events from multi-agent pipelines built on the OpenAI Agents SDK.

CrewAI

Record task assignments, tool executions, and crew-level outputs from CrewAI agent crews.

AutoGen

Trace multi-agent conversations, function calls, and code execution steps in AutoGen workflows.

Custom integration

Any framework can be instrumented manually using client.trace() and the TraceRecorder API.