Integrations landing
Stratix integrations — platform integrations (Stratix Premium), SDK auto-instrumentation (stratix-python), CI/CD, BYOK.
Last updated
Was this helpful?
⌘Ctrlk
Stratix integrations — platform integrations (Stratix Premium), SDK auto-instrumentation (stratix-python), CI/CD, BYOK.
Stratix integrates with your AI stack at three layers:
Platform integrations — configurable from Premium Settings → Integrations
SDK auto-instrumentation — drop-in instrumentation for LLM providers via layerlens.instrument
CI/CD and BYOK — operational integrations
Configurable from Premium → Settings → Integrations:
Langfuse — connect an existing Langfuse instance to ingest trace data into your Stratix workspace
BYOK custom models — register OpenAI-compatible endpoints (also surfaced under the Models page) — see BYOK custom models
The stratix-python SDK's layerlens.instrument package wraps LLM provider clients to emit traces automatically.
Production adapters with live integration tests:
OpenAI — chat completions, embeddings, tool-use, streaming
Anthropic — Messages API, multi-turn, cache metadata, streaming
Available via SDK extras (drop-in instrumentation):
OpenAI
pip install 'layerlens[providers-openai]'
Anthropic
pip install 'layerlens[providers-anthropic]'
Azure OpenAI
pip install 'layerlens[providers-azure-openai]'
AWS Bedrock
pip install 'layerlens[providers-bedrock]'
Google Vertex AI
pip install 'layerlens[providers-google-vertex]'
Cohere
pip install 'layerlens[providers-cohere]'
Mistral
pip install 'layerlens[providers-mistral]'
Ollama
pip install 'layerlens[providers-ollama]'
LiteLLM
pip install 'layerlens[providers-litellm]'
For canonical sample code, see samples/integrations/ — manual trace upload (openai_traced.py, anthropic_traced.py) and zero-code auto-instrumentation (openai_instrumented.py, langchain_instrumented.py).
For frameworks beyond what the canonical SDK samples cover, the samples/instrument/ directory and SDK pyproject.toml extras list the adapter surface that ships with the SDK.
For pipelines without a dedicated provider adapter, use the manual primitives:
GitHub Actions — drop-in workflow file ships in the SDK at samples/cicd/github_actions_gate.yml
Notifications are configured in Premium → Settings → Notifications. In-app and email subscriptions support:
Evaluation completion
Trace evaluation threshold crossing
GEPA optimization completion
ECU low-balance warnings
Org events (invitations, role changes)
Billing events
For routing notifications to Slack or PagerDuty, customers route their own webhooks from notification events to chat / alerting platforms. See Cookbook: Slack notification routing and PagerDuty escalation for shaped patterns.
Any OpenAI-compatible endpoint works via client.models.create_custom(...):
Last updated
Was this helpful?
Was this helpful?
from layerlens.instrument import trace, span
@trace
def run_pipeline(query: str):
with span("retrieval"):
chunks = retrieve(query)
with span("llm-call"):
return llm_call(query, chunks)