Implementation:Explodinggradients Ragas RagasTracer Callbacks

Knowledge Sources	Explodinggradients_Ragas
Domains	Tracing, Callbacks
Last Updated	2026-02-10 00:00 GMT

Overview

Callback and tracing system for Ragas evaluations built on LangChain's callback infrastructure, enabling hierarchical trace collection.

Description

This module provides RagasTracer (a LangChain BaseCallbackHandler) that traces chain execution, recording parent-child relationships and run metadata. ChainRun and MetricTrace models store trace data. The new_group function creates callback manager groups for evaluation chains. parse_run_traces converts hierarchical traces into flat metric trace lists.

Usage

Use RagasTracer as a callback handler when running evaluations to collect execution traces for debugging or observability.

Code Reference

Source Location

• Repository: Explodinggradients_Ragas
• File: src/ragas/callbacks.py
• Lines: 19-173

Signature

def new_group(
    name: str,
    inputs: Dict[str, Any],
    callbacks: Callbacks,
    tags: Optional[List[str]] = None,
    metadata: Optional[Dict[str, Any]] = None,
) -> Tuple[CallbackManagerForChainRun, CallbackManagerForChainGroup]:
    ...

@dataclass
class RagasTracer(BaseCallbackHandler):
    traces: Dict[str, ChainRun] = field(default_factory=dict)
    def on_chain_start(self, serialized, inputs, *, run_id, parent_run_id=None, tags=None, metadata=None, **kwargs):
        ...
    def on_chain_end(self, outputs, *, run_id, **kwargs):
        ...
    def to_jsons(self) -> str:
        ...

def parse_run_traces(traces: Dict[str, ChainRun], parent_run_id: Optional[str] = None) -> List[Dict[str, Any]]:
    ...

Import

from ragas.callbacks import RagasTracer, new_group, parse_run_traces

I/O Contract

Inputs

Name	Type	Required	Description
name	str	Yes	Name for the evaluation chain group
inputs	Dict[str, Any]	Yes	Initial inputs for the chain
callbacks	Callbacks	Yes	LangChain callback configuration

Outputs

Name	Type	Description
new_group returns	Tuple[CallbackManagerForChainRun, CallbackManagerForChainGroup]	Callback managers for the chain
to_jsons returns	str	JSON-serialized traces
parse_run_traces returns	List[Dict[str, Any]]	Flat list of metric traces

Usage Examples

from ragas.callbacks import RagasTracer

# Create tracer and use it during evaluation
tracer = RagasTracer()
# Pass tracer as a callback to evaluation functions
# After evaluation, inspect traces:
print(tracer.to_jsons())