Enterprise Integrations
Phase 5 is the integration surface that makes the runtime-governance story usable in real stacks without custom glue everywhere.
GA Integration Story
The GA integration surface covers:
- OpenAI
- Anthropic
- Azure OpenAI
- LangChain
- LangGraph
- OTLP export
- OpenInference-compatible span translation
- JSONL export for audit workflows
- SIEM-friendly schema mapping
Provider and Framework Paths
| Integration | Path | Primary use |
|---|---|---|
| OpenAI | spanforge.integrations.openai | SDK auto-instrumentation for traced LLM calls |
| Anthropic | spanforge.integrations.anthropic and pricing coverage | Claude integration and cross-provider cost normalization |
| Azure OpenAI | spanforge.integrations.azure_openai | Instance-level instrumentation for Azure-hosted OpenAI clients |
| LangChain | spanforge.integrations.langchain | Callback-based event capture |
| LangGraph | spanforge.integrations.langgraph | Governance-aware agent graph demo path |
Export and Interop Paths
| Path | Module | Purpose |
|---|---|---|
| OTLP | spanforge.export.otlp | coexist with existing observability backends |
| JSONL | spanforge.export.jsonl and spanforge.io | audit retention, replay, air-gapped transport |
| Splunk | spanforge.export.siem_splunk | SIEM ingestion with normalized event mapping |
| Syslog / CEF | spanforge.export.siem_syslog | legacy SIEM and SOC pipelines |
| OpenInference | spanforge.export.openinference | interoperability with OpenInference consumers |
When To Use Which Path
- Use OpenAI or Azure OpenAI instrumentation when you want token, model, and cost telemetry attached automatically to active spans.
- Use LangChain when your orchestration layer is already callback-oriented.
- Use LangGraph when you need node-level scope, RBAC, and lineage coverage in an agent workflow demo.
- Use JSONL when durability and offline portability matter more than streaming.
- Use OTLP when SpanForge should coexist with your current telemetry collector.
- Use SIEM exporters when the audience is the SOC instead of an observability team.
- Use OpenInference when you need downstream compatibility, but keep SpanForge packages as the source of signed governance evidence.
Related Docs
- API Reference:
spanforge.integrations - API Reference:
spanforge.export - User Guide: Export Backends
- Runtime Governance GA Guide
Ready to instrument your AI pipeline?