fix(batch-evaluation): pass trace ID for score creation if scope = observations

[hassiebp]

Disclaimer: Experimental PR review

Greptile Overview

Greptile Summary

Fixed batch evaluation score creation for observations by passing the required trace_id parameter. When creating scores for observations, the Langfuse API requires both observation_id and trace_id. This PR correctly extracts the trace_id from the ObservationsView item and passes it to _create_score_for_scope.

Key changes:

• Added cast import at the top of the module
• Modified _create_score_for_scope to accept optional trace_id parameter (keyword-only)
• Updated both call sites to pass trace_id when scope == "observations"
• Added test coverage for observations batch evaluation
• Uncommented test suite to enable validation

Confidence Score: 5/5

• This PR is safe to merge with no identified risks
• The fix correctly addresses a critical bug where observation scores couldn't be created due to missing trace_id parameter. The implementation is type-safe using proper casts, follows best practices with keyword-only parameters, and includes test coverage. The change is minimal, focused, and aligns with the API requirements.
• No files require special attention

Important Files Changed

Filename	Overview
langfuse/batch_evaluation.py	Added trace_id parameter to _create_score_for_scope and passes it when scope=observations, fixing score creation for observations
tests/test_batch_evaluation.py	Added test for batch evaluation on observations and uncommented test suite

Sequence Diagram

sequenceDiagram
    participant Caller
    participant BatchEvaluationRunner
    participant Evaluator
    participant LangfuseClient
    
    Caller->>BatchEvaluationRunner: run_batched_evaluation(scope="observations")
    BatchEvaluationRunner->>BatchEvaluationRunner: fetch observations
    
    loop For each observation
        BatchEvaluationRunner->>BatchEvaluationRunner: _process_batch_evaluation_item(item)
        BatchEvaluationRunner->>Evaluator: run evaluator on observation
        Evaluator-->>BatchEvaluationRunner: evaluation result
        
        Note over BatchEvaluationRunner: Extract trace_id from ObservationsView
        BatchEvaluationRunner->>BatchEvaluationRunner: cast(ObservationsView, item).trace_id
        
        BatchEvaluationRunner->>BatchEvaluationRunner: _create_score_for_scope(scope, item_id, trace_id)
        
        alt scope == "observations"
            BatchEvaluationRunner->>LangfuseClient: create_score(observation_id, trace_id, ...)
        else scope == "traces"
            BatchEvaluationRunner->>LangfuseClient: create_score(trace_id, ...)
        end
        
        LangfuseClient-->>BatchEvaluationRunner: score created
    end
    
    BatchEvaluationRunner-->>Caller: BatchEvaluationResult

Loading

[greptile-apps]

_{No files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}