Add Langfuse feedback support

[loukratz-bv]

Currently, the feedback API in qtype/interpreter/feedback_api.py and telemetry.py only support Phoenix

Proposed Enhancement

Implement Langfuse feedback submission in the create_feedback_endpoint function to support users who use Langfuse as their telemetry backend.

Current State

elif telemetry.provider == "Langfuse":
    # TODO: Implement Langfuse feedback submission
    raise NotImplementedError(
        "Langfuse feedback not yet implemented"
    )

Requirements

• Research Langfuse's feedback/annotation API
• Implement feedback submission for all three feedback types (thumbs, rating, category)
• Ensure it follows the same pattern as Phoenix implementation
• Handle score conversion appropriately for each feedback type
• Add appropriate error handling and logging

Context

The Phoenix implementation uses client.annotations.add_span_annotation() with parameters: span_id, annotation_name, label, score, explanation, and annotator_kind="HUMAN".

Langfuse will need a similar implementation adapted to its SDK/API.

Related Files

• qtype/interpreter/feedback_api.py - Feedback endpoint implementation
• qtype/interpreter/telemetry.py - Langfuse telemetry setup (already implemented)