adding langfuse for openai

[Vaibhav159]

Changes in This PR

Integration with Langfuse (Optional)

• Added support for Langfuse as an optional add-on for enhanced telemetry and observability.
• Created a fallback mechanism to ensure the core functionality is not blocked if Langfuse is not installed or configured.

Dummy observe Decorator

• Introduced a placeholder observe decorator for seamless use in environments without Langfuse installed.
• Ensures functions wrapped with @observe still execute normally when Langfuse is unavailable.

[CodeNameR]

interested in this

[ffesseler]

I would be surprised to see this merged as it introduces a direct dependency to a LLM platform. It would be better to add some telemetry with OpenTelemetry's standard libraries and possibly also those from openinference.

[marcklingen]

Langfuse maintainer here. Agree with @ffesseler that th best long term option would be to add native OTel instrumentation to pipecat in order to integrate well with multiple observability/evaluation projects. Happy to support this effort in case there are any questions.

[ffesseler]

@aconchillo Any guidelines on how to approach this? I would be interested in proposing an initial implementation and I would be curious to know if there are already ideas around the subject. (like what OTel instrumentation lib to use : openllmetry, openlit, openinference, ...?)

[marcklingen]

Created an issue to track an opentelemetry-based integration: #1456