Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 3 additions & 1 deletion langfuse/__init__.py
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@
EvaluatorStats,
MapperFunction,
)
from langfuse.experiment import Evaluation
from langfuse.experiment import Evaluation, RegressionError, RunnerContext

from ._client import client as _client_module
from ._client.attributes import LangfuseOtelSpanAttributes
Expand Down Expand Up @@ -63,6 +63,8 @@
"EvaluatorStats",
"BatchEvaluationResumeToken",
"BatchEvaluationResult",
"RunnerContext",
"RegressionError",
"__version__",
"is_default_export_span",
"is_langfuse_span",
Expand Down
156 changes: 156 additions & 0 deletions langfuse/experiment.py
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,9 @@
"""

import asyncio
from datetime import datetime
from typing import (
TYPE_CHECKING,
Any,
Awaitable,
Dict,
Expand All @@ -15,12 +17,17 @@
Protocol,
TypedDict,
Union,
overload,
)

from langfuse.api import DatasetItem
from langfuse.logger import langfuse_logger as logger
from langfuse.types import ExperimentScoreType

if TYPE_CHECKING:
from langfuse._client.client import Langfuse
from langfuse.batch_evaluation import CompositeEvaluatorFunction


class LocalExperimentItem(TypedDict, total=False):
"""Structure for local experiment data items (not from Langfuse datasets).
Expand Down Expand Up @@ -1049,3 +1056,152 @@
)

return langfuse_evaluator


class RunnerContext:
"""Wraps :meth:`Langfuse.run_experiment` with CI-injected defaults.

Intended for use with the ``langfuse/experiment-action`` GitHub Action
Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

RunnerContext feels like the right abstraction here, and I think the action side should lean into it as the single public integration path: construct the context, then call experiment(context) explicitly instead of also teaching users to read action-specific env vars. Since this is likely to become the primary authoring model, I’d optimize for that clean contract end-to-end.

(https://github.com/langfuse/experiment-action). The action builds a
``RunnerContext`` before invoking the user's ``experiment(context)``
function. Defaults set here (dataset, metadata tags) are applied when
the user omits them on the :meth:`run_experiment` call; users can
override any default by passing the corresponding argument explicitly.
"""

def __init__(
self,
*,
client: "Langfuse",
data: Optional[ExperimentData] = None,
dataset_version: Optional[datetime] = None,
metadata: Optional[Dict[str, str]] = None,
):
"""Build a ``RunnerContext`` populated with defaults for ``run_experiment``.

Typically called by the ``langfuse/experiment-action`` GitHub Action,
not by end users directly. Every field except ``client`` is optional:
fields left as ``None`` simply mean the corresponding argument must be
supplied on the :meth:`run_experiment` call.

Args:
client: Initialized Langfuse SDK client used to execute the
experiment. The action creates this from the
``langfuse_public_key`` / ``langfuse_secret_key`` /
``langfuse_base_url`` inputs.
data: Default dataset items to run the experiment on. Accepts
either ``List[LocalExperimentItem]`` or ``List[DatasetItem]``.

Check failure on line 1093 in langfuse/experiment.py

View check run for this annotation

Claude / Claude Code Review

RunnerContext.__init__ missing run_name — action cannot inject CI run name

RunnerContext.__init__ does not accept a run_name parameter and stores no self.run_name attribute, so the GitHub Action has no mechanism to inject a CI-meaningful run name (e.g. 'pr-42-sha-abc') as a context-level default. Fix by adding run_name: Optional[str] = None to __init__, storing self.run_name, and resolving it in run_experiment as: resolved_run_name = run_name if run_name is not None else self.run_name.
Comment thread
wochinge marked this conversation as resolved.
Injected by the action when ``dataset_name`` is configured.
If ``None``, the user must pass ``data=`` to
:meth:`run_experiment`.
dataset_version: Optional pinned dataset version. Injected by the
action when ``dataset_version`` is configured.
metadata: Default metadata attached to every experiment trace and
the dataset run. The action injects GitHub-sourced tags (SHA,
PR link, workflow run link, branch, GH user, etc.). Merged
with any ``metadata`` passed to :meth:`run_experiment`, with
user-supplied keys winning on collision.
"""
self.client = client
self.data = data
self.dataset_version = dataset_version
self.metadata = metadata

def run_experiment(
self,
*,
name: str,
run_name: Optional[str] = None,
description: Optional[str] = None,
data: Optional[ExperimentData] = None,
task: TaskFunction,
evaluators: List[EvaluatorFunction] = [],
composite_evaluator: Optional["CompositeEvaluatorFunction"] = None,
run_evaluators: List[RunEvaluatorFunction] = [],
Comment thread
wochinge marked this conversation as resolved.
max_concurrency: int = 50,
metadata: Optional[Dict[str, str]] = None,
_dataset_version: Optional[datetime] = None,
) -> ExperimentResult:
resolved_data = data if data is not None else self.data
if resolved_data is None:
raise ValueError(
"`data` must be provided either on the RunnerContext or the run_experiment call"
)

resolved_dataset_version = (
_dataset_version if _dataset_version is not None else self.dataset_version
)

merged_metadata: Optional[Dict[str, str]]
if self.metadata is None and metadata is None:
merged_metadata = None
else:
merged_metadata = {**(self.metadata or {}), **(metadata or {})}

return self.client.run_experiment(
name=name,
run_name=run_name,
description=description,
data=resolved_data,
task=task,
evaluators=evaluators,
composite_evaluator=composite_evaluator,
run_evaluators=run_evaluators,
max_concurrency=max_concurrency,
metadata=merged_metadata,
_dataset_version=resolved_dataset_version,
)


class RegressionError(Exception):
"""Raised by a user's ``experiment`` function to signal a CI gate failure.

Intended for use with the ``langfuse/experiment-action`` GitHub Action
(https://github.com/langfuse/experiment-action). The action catches this
exception and, when ``should_fail_on_error`` is enabled, fails the
workflow run and renders a callout in the PR comment using
``metric``/``value``/``threshold`` if supplied, otherwise ``str(exc)``.

Callers choose one of three forms:

- ``RegressionError(result=r)`` — minimal, generic message.
- ``RegressionError(result=r, message="...")`` — free-form message.
- ``RegressionError(result=r, metric="acc", value=0.7, threshold=0.9)`` —
structured; ``metric`` and ``value`` must be provided together so the
action can render a targeted callout without ``None`` placeholders.
"""

@overload
def __init__(self, *, result: ExperimentResult) -> None: ...
@overload
def __init__(self, *, result: ExperimentResult, message: str) -> None: ...
@overload
def __init__(
self,
*,
result: ExperimentResult,
metric: str,
value: float,
threshold: Optional[float] = None,
message: Optional[str] = None,
) -> None: ...
def __init__(
self,
*,
result: ExperimentResult,
metric: Optional[str] = None,
value: Optional[float] = None,
threshold: Optional[float] = None,
message: Optional[str] = None,
):
self.result = result
self.metric = metric
self.value = value
self.threshold = threshold
if message is not None:
formatted = message
elif metric is not None and value is not None:
formatted = f"Regression on `{metric}`: {value} (threshold {threshold})"
else:
formatted = "Experiment regression detected"
super().__init__(formatted)
Comment thread
wochinge marked this conversation as resolved.
Comment thread
wochinge marked this conversation as resolved.
Loading
Loading