"""Langfuse OpenTelemetry integration module. This module implements Langfuse's core observability functionality on top of the OpenTelemetry (OTel) standard. """ import asyncio import logging import os import re import urllib.parse import warnings from datetime import datetime from hashlib import sha256 from time import time_ns from typing import ( Any, Callable, Dict, List, Literal, Optional, Type, Union, cast, overload, ) import backoff import httpx from opentelemetry import trace as otel_trace_api from opentelemetry.sdk.trace import ReadableSpan, TracerProvider from opentelemetry.sdk.trace.id_generator import RandomIdGenerator from opentelemetry.util._decorator import ( _AgnosticContextManager, _agnosticcontextmanager, ) from packaging.version import Version from typing_extensions import deprecated from langfuse._client.attributes import LangfuseOtelSpanAttributes, _serialize from langfuse._client.constants import ( LANGFUSE_SDK_EXPERIMENT_ENVIRONMENT, ObservationTypeGenerationLike, ObservationTypeLiteral, ObservationTypeLiteralNoEvent, ObservationTypeSpanLike, get_observation_types_list, ) from langfuse._client.datasets import DatasetClient from langfuse._client.environment_variables import ( LANGFUSE_BASE_URL, LANGFUSE_DEBUG, LANGFUSE_HOST, LANGFUSE_PUBLIC_KEY, LANGFUSE_RELEASE, LANGFUSE_SAMPLE_RATE, LANGFUSE_SECRET_KEY, LANGFUSE_TIMEOUT, LANGFUSE_TRACING_ENABLED, LANGFUSE_TRACING_ENVIRONMENT, ) from langfuse._client.propagation import ( PropagatedExperimentAttributes, _propagate_attributes, ) from langfuse._client.resource_manager import LangfuseResourceManager from langfuse._client.span import ( LangfuseAgent, LangfuseChain, LangfuseEmbedding, LangfuseEvaluator, LangfuseEvent, LangfuseGeneration, LangfuseGuardrail, LangfuseRetriever, LangfuseSpan, LangfuseTool, ) from langfuse._client.utils import get_sha256_hash_hex, run_async_safely from langfuse._utils import _get_timestamp from langfuse._utils.environment import get_common_release_envs from langfuse._utils.parse_error import handle_fern_exception from langfuse._utils.prompt_cache import PromptCache from langfuse.api import ( CreateChatPromptRequest, CreateChatPromptType, CreateTextPromptRequest, Dataset, DatasetItem, DatasetRunWithItems, DatasetStatus, DeleteDatasetRunResponse, Error, MapValue, NotFoundError, PaginatedDatasetRuns, Prompt_Chat, Prompt_Text, ScoreBody, TraceBody, ) from langfuse.batch_evaluation import ( BatchEvaluationResult, BatchEvaluationResumeToken, BatchEvaluationRunner, CompositeEvaluatorFunction, MapperFunction, ) from langfuse.experiment import ( Evaluation, EvaluatorFunction, ExperimentData, ExperimentItem, ExperimentItemResult, ExperimentResult, RunEvaluatorFunction, TaskFunction, _run_evaluator, _run_task, ) from langfuse.logger import langfuse_logger from langfuse.media import LangfuseMedia from langfuse.model import ( ChatMessageDict, ChatMessageWithPlaceholdersDict, ChatPromptClient, PromptClient, TextPromptClient, ) from langfuse.types import MaskFunction, ScoreDataType, SpanLevel, TraceContext class Langfuse: """Main client for Langfuse tracing and platform features. This class provides an interface for creating and managing traces, spans, and generations in Langfuse as well as interacting with the Langfuse API. The client features a thread-safe singleton pattern for each unique public API key, ensuring consistent trace context propagation across your application. It implements efficient batching of spans with configurable flush settings and includes background thread management for media uploads and score ingestion. Configuration is flexible through either direct parameters or environment variables, with graceful fallbacks and runtime configuration updates. Attributes: api: Synchronous API client for Langfuse backend communication async_api: Asynchronous API client for Langfuse backend communication _otel_tracer: Internal LangfuseTracer instance managing OpenTelemetry components Parameters: public_key (Optional[str]): Your Langfuse public API key. Can also be set via LANGFUSE_PUBLIC_KEY environment variable. secret_key (Optional[str]): Your Langfuse secret API key. Can also be set via LANGFUSE_SECRET_KEY environment variable. base_url (Optional[str]): The Langfuse API base URL. Defaults to "https://cloud.langfuse.com". Can also be set via LANGFUSE_BASE_URL environment variable. host (Optional[str]): Deprecated. Use base_url instead. The Langfuse API host URL. Defaults to "https://cloud.langfuse.com". timeout (Optional[int]): Timeout in seconds for API requests. Defaults to 5 seconds. httpx_client (Optional[httpx.Client]): Custom httpx client for making non-tracing HTTP requests. If not provided, a default client will be created. debug (bool): Enable debug logging. Defaults to False. Can also be set via LANGFUSE_DEBUG environment variable. tracing_enabled (Optional[bool]): Enable or disable tracing. Defaults to True. Can also be set via LANGFUSE_TRACING_ENABLED environment variable. flush_at (Optional[int]): Number of spans to batch before sending to the API. Defaults to 512. Can also be set via LANGFUSE_FLUSH_AT environment variable. flush_interval (Optional[float]): Time in seconds between batch flushes. Defaults to 5 seconds. Can also be set via LANGFUSE_FLUSH_INTERVAL environment variable. environment (Optional[str]): Environment name for tracing. Default is 'default'. Can also be set via LANGFUSE_TRACING_ENVIRONMENT environment variable. Can be any lowercase alphanumeric string with hyphens and underscores that does not start with 'langfuse'. release (Optional[str]): Release version/hash of your application. Used for grouping analytics by release. media_upload_thread_count (Optional[int]): Number of background threads for handling media uploads. Defaults to 1. Can also be set via LANGFUSE_MEDIA_UPLOAD_THREAD_COUNT environment variable. sample_rate (Optional[float]): Sampling rate for traces (0.0 to 1.0). Defaults to 1.0 (100% of traces are sampled). Can also be set via LANGFUSE_SAMPLE_RATE environment variable. mask (Optional[MaskFunction]): Function to mask sensitive data in traces before sending to the API. blocked_instrumentation_scopes (Optional[List[str]]): Deprecated. Use `should_export_span` instead. Equivalent behavior: ```python from langfuse.span_filter import is_default_export_span blocked = {"sqlite", "requests"} should_export_span = lambda span: ( is_default_export_span(span) and ( span.instrumentation_scope is None or span.instrumentation_scope.name not in blocked ) ) ``` should_export_span (Optional[Callable[[ReadableSpan], bool]]): Callback to decide whether to export a span. If omitted, Langfuse uses the default filter (Langfuse SDK spans, spans with `gen_ai.*` attributes, and known LLM instrumentation scopes). additional_headers (Optional[Dict[str, str]]): Additional headers to include in all API requests and OTLPSpanExporter requests. These headers will be merged with default headers. Note: If httpx_client is provided, additional_headers must be set directly on your custom httpx_client as well. tracer_provider(Optional[TracerProvider]): OpenTelemetry TracerProvider to use for Langfuse. This can be useful to set to have disconnected tracing between Langfuse and other OpenTelemetry-span emitting libraries. Note: To track active spans, the context is still shared between TracerProviders. This may lead to broken trace trees. Example: ```python from langfuse.otel import Langfuse # Initialize the client (reads from env vars if not provided) langfuse = Langfuse( public_key="your-public-key", secret_key="your-secret-key", host="https://cloud.langfuse.com", # Optional, default shown ) # Create a trace span with langfuse.start_as_current_observation(name="process-query") as span: # Your application code here # Create a nested generation span for an LLM call with span.start_as_current_generation( name="generate-response", model="gpt-4", input={"query": "Tell me about AI"}, model_parameters={"temperature": 0.7, "max_tokens": 500} ) as generation: # Generate response here response = "AI is a field of computer science..." generation.update( output=response, usage_details={"prompt_tokens": 10, "completion_tokens": 50}, cost_details={"total_cost": 0.0023} ) # Score the generation (supports NUMERIC, BOOLEAN, CATEGORICAL) generation.score(name="relevance", value=0.95, data_type="NUMERIC") ``` """ _resources: Optional[LangfuseResourceManager] = None _mask: Optional[MaskFunction] = None _otel_tracer: otel_trace_api.Tracer def __init__( self, *, public_key: Optional[str] = None, secret_key: Optional[str] = None, base_url: Optional[str] = None, host: Optional[str] = None, timeout: Optional[int] = None, httpx_client: Optional[httpx.Client] = None, debug: bool = False, tracing_enabled: Optional[bool] = True, flush_at: Optional[int] = None, flush_interval: Optional[float] = None, environment: Optional[str] = None, release: Optional[str] = None, media_upload_thread_count: Optional[int] = None, sample_rate: Optional[float] = None, mask: Optional[MaskFunction] = None, blocked_instrumentation_scopes: Optional[List[str]] = None, should_export_span: Optional[Callable[[ReadableSpan], bool]] = None, additional_headers: Optional[Dict[str, str]] = None, tracer_provider: Optional[TracerProvider] = None, ): self._base_url = ( base_url or os.environ.get(LANGFUSE_BASE_URL) or host or os.environ.get(LANGFUSE_HOST, "https://cloud.langfuse.com") ) self._environment = environment or cast( str, os.environ.get(LANGFUSE_TRACING_ENVIRONMENT) ) self._release = ( release or os.environ.get(LANGFUSE_RELEASE, None) or get_common_release_envs() ) self._project_id: Optional[str] = None sample_rate = sample_rate or float(os.environ.get(LANGFUSE_SAMPLE_RATE, 1.0)) if not 0.0 <= sample_rate <= 1.0: raise ValueError( f"Sample rate must be between 0.0 and 1.0, got {sample_rate}" ) timeout = timeout or int(os.environ.get(LANGFUSE_TIMEOUT, 5)) self._tracing_enabled = ( tracing_enabled and os.environ.get(LANGFUSE_TRACING_ENABLED, "true").lower() != "false" ) if not self._tracing_enabled: langfuse_logger.info( "Configuration: Langfuse tracing is explicitly disabled. No data will be sent to the Langfuse API." ) debug = ( debug if debug else (os.getenv(LANGFUSE_DEBUG, "false").lower() == "true") ) if debug: logging.basicConfig( format="%(asctime)s - %(name)s - %(levelname)s - %(message)s" ) langfuse_logger.setLevel(logging.DEBUG) public_key = public_key or os.environ.get(LANGFUSE_PUBLIC_KEY) if public_key is None: langfuse_logger.warning( "Authentication error: Langfuse client initialized without public_key. Client will be disabled. " "Provide a public_key parameter or set LANGFUSE_PUBLIC_KEY environment variable. " ) self._otel_tracer = otel_trace_api.NoOpTracer() return secret_key = secret_key or os.environ.get(LANGFUSE_SECRET_KEY) if secret_key is None: langfuse_logger.warning( "Authentication error: Langfuse client initialized without secret_key. Client will be disabled. " "Provide a secret_key parameter or set LANGFUSE_SECRET_KEY environment variable. " ) self._otel_tracer = otel_trace_api.NoOpTracer() return if os.environ.get("OTEL_SDK_DISABLED", "false").lower() == "true": langfuse_logger.warning( "OTEL_SDK_DISABLED is set. Langfuse tracing will be disabled and no traces will appear in the UI." ) if blocked_instrumentation_scopes is not None: warnings.warn( "`blocked_instrumentation_scopes` is deprecated and will be removed in a future release. " "Use `should_export_span` instead. Example: " "from langfuse.span_filter import is_default_export_span; " 'blocked={"scope"}; should_export_span=lambda span: ' "is_default_export_span(span) and (span.instrumentation_scope is None or " "span.instrumentation_scope.name not in blocked).", DeprecationWarning, stacklevel=2, ) # Initialize api and tracer if requirements are met self._resources = LangfuseResourceManager( public_key=public_key, secret_key=secret_key, base_url=self._base_url, timeout=timeout, environment=self._environment, release=release, flush_at=flush_at, flush_interval=flush_interval, httpx_client=httpx_client, media_upload_thread_count=media_upload_thread_count, sample_rate=sample_rate, mask=mask, tracing_enabled=self._tracing_enabled, blocked_instrumentation_scopes=blocked_instrumentation_scopes, should_export_span=should_export_span, additional_headers=additional_headers, tracer_provider=tracer_provider, ) self._mask = self._resources.mask self._otel_tracer = ( self._resources.tracer if self._tracing_enabled and self._resources.tracer is not None else otel_trace_api.NoOpTracer() ) self.api = self._resources.api self.async_api = self._resources.async_api @overload def start_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["generation"], input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, completion_start_time: Optional[datetime] = None, model: Optional[str] = None, model_parameters: Optional[Dict[str, MapValue]] = None, usage_details: Optional[Dict[str, int]] = None, cost_details: Optional[Dict[str, float]] = None, prompt: Optional[PromptClient] = None, ) -> LangfuseGeneration: ... @overload def start_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["span"] = "span", input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, ) -> LangfuseSpan: ... @overload def start_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["agent"], input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, ) -> LangfuseAgent: ... @overload def start_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["tool"], input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, ) -> LangfuseTool: ... @overload def start_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["chain"], input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, ) -> LangfuseChain: ... @overload def start_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["retriever"], input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, ) -> LangfuseRetriever: ... @overload def start_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["evaluator"], input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, ) -> LangfuseEvaluator: ... @overload def start_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["embedding"], input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, completion_start_time: Optional[datetime] = None, model: Optional[str] = None, model_parameters: Optional[Dict[str, MapValue]] = None, usage_details: Optional[Dict[str, int]] = None, cost_details: Optional[Dict[str, float]] = None, prompt: Optional[PromptClient] = None, ) -> LangfuseEmbedding: ... @overload def start_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["guardrail"], input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, ) -> LangfuseGuardrail: ... def start_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: ObservationTypeLiteralNoEvent = "span", input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, completion_start_time: Optional[datetime] = None, model: Optional[str] = None, model_parameters: Optional[Dict[str, MapValue]] = None, usage_details: Optional[Dict[str, int]] = None, cost_details: Optional[Dict[str, float]] = None, prompt: Optional[PromptClient] = None, ) -> Union[ LangfuseSpan, LangfuseGeneration, LangfuseAgent, LangfuseTool, LangfuseChain, LangfuseRetriever, LangfuseEvaluator, LangfuseEmbedding, LangfuseGuardrail, ]: """Create a new observation of the specified type. This method creates a new observation but does not set it as the current span in the context. To create and use an observation within a context, use start_as_current_observation(). Args: trace_context: Optional context for connecting to an existing trace name: Name of the observation as_type: Type of observation to create (defaults to "span") input: Input data for the operation output: Output data from the operation metadata: Additional metadata to associate with the observation version: Version identifier for the code or component level: Importance level of the observation status_message: Optional status message for the observation completion_start_time: When the model started generating (for generation types) model: Name/identifier of the AI model used (for generation types) model_parameters: Parameters used for the model (for generation types) usage_details: Token usage information (for generation types) cost_details: Cost information (for generation types) prompt: Associated prompt template (for generation types) Returns: An observation object of the appropriate type that must be ended with .end() """ if trace_context: trace_id = trace_context.get("trace_id", None) parent_span_id = trace_context.get("parent_span_id", None) if trace_id: remote_parent_span = self._create_remote_parent_span( trace_id=trace_id, parent_span_id=parent_span_id ) with otel_trace_api.use_span( cast(otel_trace_api.Span, remote_parent_span) ): otel_span = self._otel_tracer.start_span(name=name) otel_span.set_attribute(LangfuseOtelSpanAttributes.AS_ROOT, True) return self._create_observation_from_otel_span( otel_span=otel_span, as_type=as_type, input=input, output=output, metadata=metadata, version=version, level=level, status_message=status_message, completion_start_time=completion_start_time, model=model, model_parameters=model_parameters, usage_details=usage_details, cost_details=cost_details, prompt=prompt, ) otel_span = self._otel_tracer.start_span(name=name) return self._create_observation_from_otel_span( otel_span=otel_span, as_type=as_type, input=input, output=output, metadata=metadata, version=version, level=level, status_message=status_message, completion_start_time=completion_start_time, model=model, model_parameters=model_parameters, usage_details=usage_details, cost_details=cost_details, prompt=prompt, ) def _create_observation_from_otel_span( self, *, otel_span: otel_trace_api.Span, as_type: ObservationTypeLiteralNoEvent, input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, completion_start_time: Optional[datetime] = None, model: Optional[str] = None, model_parameters: Optional[Dict[str, MapValue]] = None, usage_details: Optional[Dict[str, int]] = None, cost_details: Optional[Dict[str, float]] = None, prompt: Optional[PromptClient] = None, ) -> Union[ LangfuseSpan, LangfuseGeneration, LangfuseAgent, LangfuseTool, LangfuseChain, LangfuseRetriever, LangfuseEvaluator, LangfuseEmbedding, LangfuseGuardrail, ]: """Create the appropriate observation type from an OTEL span.""" if as_type in get_observation_types_list(ObservationTypeGenerationLike): observation_class = self._get_span_class(as_type) # Type ignore to prevent overloads of internal _get_span_class function, # issue is that LangfuseEvent could be returned and that classes have diff. args return observation_class( # type: ignore[return-value,call-arg] otel_span=otel_span, langfuse_client=self, environment=self._environment, release=self._release, input=input, output=output, metadata=metadata, version=version, level=level, status_message=status_message, completion_start_time=completion_start_time, model=model, model_parameters=model_parameters, usage_details=usage_details, cost_details=cost_details, prompt=prompt, ) else: # For other types (e.g. span, guardrail), create appropriate class without generation properties observation_class = self._get_span_class(as_type) # Type ignore to prevent overloads of internal _get_span_class function, # issue is that LangfuseEvent could be returned and that classes have diff. args return observation_class( # type: ignore[return-value,call-arg] otel_span=otel_span, langfuse_client=self, environment=self._environment, release=self._release, input=input, output=output, metadata=metadata, version=version, level=level, status_message=status_message, ) # span._observation_type = as_type # span._otel_span.set_attribute("langfuse.observation.type", as_type) # return span @overload def start_as_current_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["generation"], input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, completion_start_time: Optional[datetime] = None, model: Optional[str] = None, model_parameters: Optional[Dict[str, MapValue]] = None, usage_details: Optional[Dict[str, int]] = None, cost_details: Optional[Dict[str, float]] = None, prompt: Optional[PromptClient] = None, end_on_exit: Optional[bool] = None, ) -> _AgnosticContextManager[LangfuseGeneration]: ... @overload def start_as_current_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["span"] = "span", input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, end_on_exit: Optional[bool] = None, ) -> _AgnosticContextManager[LangfuseSpan]: ... @overload def start_as_current_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["agent"], input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, end_on_exit: Optional[bool] = None, ) -> _AgnosticContextManager[LangfuseAgent]: ... @overload def start_as_current_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["tool"], input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, end_on_exit: Optional[bool] = None, ) -> _AgnosticContextManager[LangfuseTool]: ... @overload def start_as_current_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["chain"], input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, end_on_exit: Optional[bool] = None, ) -> _AgnosticContextManager[LangfuseChain]: ... @overload def start_as_current_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["retriever"], input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, end_on_exit: Optional[bool] = None, ) -> _AgnosticContextManager[LangfuseRetriever]: ... @overload def start_as_current_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["evaluator"], input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, end_on_exit: Optional[bool] = None, ) -> _AgnosticContextManager[LangfuseEvaluator]: ... @overload def start_as_current_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["embedding"], input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, completion_start_time: Optional[datetime] = None, model: Optional[str] = None, model_parameters: Optional[Dict[str, MapValue]] = None, usage_details: Optional[Dict[str, int]] = None, cost_details: Optional[Dict[str, float]] = None, prompt: Optional[PromptClient] = None, end_on_exit: Optional[bool] = None, ) -> _AgnosticContextManager[LangfuseEmbedding]: ... @overload def start_as_current_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: Literal["guardrail"], input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, end_on_exit: Optional[bool] = None, ) -> _AgnosticContextManager[LangfuseGuardrail]: ... def start_as_current_observation( self, *, trace_context: Optional[TraceContext] = None, name: str, as_type: ObservationTypeLiteralNoEvent = "span", input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, completion_start_time: Optional[datetime] = None, model: Optional[str] = None, model_parameters: Optional[Dict[str, MapValue]] = None, usage_details: Optional[Dict[str, int]] = None, cost_details: Optional[Dict[str, float]] = None, prompt: Optional[PromptClient] = None, end_on_exit: Optional[bool] = None, ) -> Union[ _AgnosticContextManager[LangfuseGeneration], _AgnosticContextManager[LangfuseSpan], _AgnosticContextManager[LangfuseAgent], _AgnosticContextManager[LangfuseTool], _AgnosticContextManager[LangfuseChain], _AgnosticContextManager[LangfuseRetriever], _AgnosticContextManager[LangfuseEvaluator], _AgnosticContextManager[LangfuseEmbedding], _AgnosticContextManager[LangfuseGuardrail], ]: """Create a new observation and set it as the current span in a context manager. This method creates a new observation of the specified type and sets it as the current span within a context manager. Use this method with a 'with' statement to automatically handle the observation lifecycle within a code block. The created observation will be the child of the current span in the context. Args: trace_context: Optional context for connecting to an existing trace name: Name of the observation (e.g., function or operation name) as_type: Type of observation to create (defaults to "span") input: Input data for the operation (can be any JSON-serializable object) output: Output data from the operation (can be any JSON-serializable object) metadata: Additional metadata to associate with the observation version: Version identifier for the code or component level: Importance level of the observation (info, warning, error) status_message: Optional status message for the observation end_on_exit (default: True): Whether to end the span automatically when leaving the context manager. If False, the span must be manually ended to avoid memory leaks. The following parameters are available when as_type is: "generation" or "embedding". completion_start_time: When the model started generating the response model: Name/identifier of the AI model used (e.g., "gpt-4") model_parameters: Parameters used for the model (e.g., temperature, max_tokens) usage_details: Token usage information (e.g., prompt_tokens, completion_tokens) cost_details: Cost information for the model call prompt: Associated prompt template from Langfuse prompt management Returns: A context manager that yields the appropriate observation type based on as_type Example: ```python # Create a span with langfuse.start_as_current_observation(name="process-query", as_type="span") as span: # Do work result = process_data() span.update(output=result) # Create a child span automatically with span.start_as_current_observation(name="sub-operation") as child_span: # Do sub-operation work child_span.update(output="sub-result") # Create a tool observation with langfuse.start_as_current_observation(name="web-search", as_type="tool") as tool: # Do tool work results = search_web(query) tool.update(output=results) # Create a generation observation with langfuse.start_as_current_observation( name="answer-generation", as_type="generation", model="gpt-4" ) as generation: # Generate answer response = llm.generate(...) generation.update(output=response) ``` """ if as_type in get_observation_types_list(ObservationTypeGenerationLike): if trace_context: trace_id = trace_context.get("trace_id", None) parent_span_id = trace_context.get("parent_span_id", None) if trace_id: remote_parent_span = self._create_remote_parent_span( trace_id=trace_id, parent_span_id=parent_span_id ) return cast( Union[ _AgnosticContextManager[LangfuseGeneration], _AgnosticContextManager[LangfuseEmbedding], ], self._create_span_with_parent_context( as_type=as_type, name=name, remote_parent_span=remote_parent_span, parent=None, end_on_exit=end_on_exit, input=input, output=output, metadata=metadata, version=version, level=level, status_message=status_message, completion_start_time=completion_start_time, model=model, model_parameters=model_parameters, usage_details=usage_details, cost_details=cost_details, prompt=prompt, ), ) return cast( Union[ _AgnosticContextManager[LangfuseGeneration], _AgnosticContextManager[LangfuseEmbedding], ], self._start_as_current_otel_span_with_processed_media( as_type=as_type, name=name, end_on_exit=end_on_exit, input=input, output=output, metadata=metadata, version=version, level=level, status_message=status_message, completion_start_time=completion_start_time, model=model, model_parameters=model_parameters, usage_details=usage_details, cost_details=cost_details, prompt=prompt, ), ) if as_type in get_observation_types_list(ObservationTypeSpanLike): if trace_context: trace_id = trace_context.get("trace_id", None) parent_span_id = trace_context.get("parent_span_id", None) if trace_id: remote_parent_span = self._create_remote_parent_span( trace_id=trace_id, parent_span_id=parent_span_id ) return cast( Union[ _AgnosticContextManager[LangfuseSpan], _AgnosticContextManager[LangfuseAgent], _AgnosticContextManager[LangfuseTool], _AgnosticContextManager[LangfuseChain], _AgnosticContextManager[LangfuseRetriever], _AgnosticContextManager[LangfuseEvaluator], _AgnosticContextManager[LangfuseGuardrail], ], self._create_span_with_parent_context( as_type=as_type, name=name, remote_parent_span=remote_parent_span, parent=None, end_on_exit=end_on_exit, input=input, output=output, metadata=metadata, version=version, level=level, status_message=status_message, ), ) return cast( Union[ _AgnosticContextManager[LangfuseSpan], _AgnosticContextManager[LangfuseAgent], _AgnosticContextManager[LangfuseTool], _AgnosticContextManager[LangfuseChain], _AgnosticContextManager[LangfuseRetriever], _AgnosticContextManager[LangfuseEvaluator], _AgnosticContextManager[LangfuseGuardrail], ], self._start_as_current_otel_span_with_processed_media( as_type=as_type, name=name, end_on_exit=end_on_exit, input=input, output=output, metadata=metadata, version=version, level=level, status_message=status_message, ), ) # This should never be reached since all valid types are handled above langfuse_logger.warning( f"Unknown observation type: {as_type}, falling back to span" ) return self._start_as_current_otel_span_with_processed_media( as_type="span", name=name, end_on_exit=end_on_exit, input=input, output=output, metadata=metadata, version=version, level=level, status_message=status_message, ) def _get_span_class( self, as_type: ObservationTypeLiteral, ) -> Union[ Type[LangfuseAgent], Type[LangfuseTool], Type[LangfuseChain], Type[LangfuseRetriever], Type[LangfuseEvaluator], Type[LangfuseEmbedding], Type[LangfuseGuardrail], Type[LangfuseGeneration], Type[LangfuseEvent], Type[LangfuseSpan], ]: """Get the appropriate span class based on as_type.""" normalized_type = as_type.lower() if normalized_type == "agent": return LangfuseAgent elif normalized_type == "tool": return LangfuseTool elif normalized_type == "chain": return LangfuseChain elif normalized_type == "retriever": return LangfuseRetriever elif normalized_type == "evaluator": return LangfuseEvaluator elif normalized_type == "embedding": return LangfuseEmbedding elif normalized_type == "guardrail": return LangfuseGuardrail elif normalized_type == "generation": return LangfuseGeneration elif normalized_type == "event": return LangfuseEvent elif normalized_type == "span": return LangfuseSpan else: return LangfuseSpan @_agnosticcontextmanager def _create_span_with_parent_context( self, *, name: str, parent: Optional[otel_trace_api.Span] = None, remote_parent_span: Optional[otel_trace_api.Span] = None, as_type: ObservationTypeLiteralNoEvent, end_on_exit: Optional[bool] = None, input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, completion_start_time: Optional[datetime] = None, model: Optional[str] = None, model_parameters: Optional[Dict[str, MapValue]] = None, usage_details: Optional[Dict[str, int]] = None, cost_details: Optional[Dict[str, float]] = None, prompt: Optional[PromptClient] = None, ) -> Any: parent_span = parent or cast(otel_trace_api.Span, remote_parent_span) with otel_trace_api.use_span(parent_span): with self._start_as_current_otel_span_with_processed_media( name=name, as_type=as_type, end_on_exit=end_on_exit, input=input, output=output, metadata=metadata, version=version, level=level, status_message=status_message, completion_start_time=completion_start_time, model=model, model_parameters=model_parameters, usage_details=usage_details, cost_details=cost_details, prompt=prompt, ) as langfuse_span: if remote_parent_span is not None: langfuse_span._otel_span.set_attribute( LangfuseOtelSpanAttributes.AS_ROOT, True ) yield langfuse_span @_agnosticcontextmanager def _start_as_current_otel_span_with_processed_media( self, *, name: str, as_type: Optional[ObservationTypeLiteralNoEvent] = None, end_on_exit: Optional[bool] = None, input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, completion_start_time: Optional[datetime] = None, model: Optional[str] = None, model_parameters: Optional[Dict[str, MapValue]] = None, usage_details: Optional[Dict[str, int]] = None, cost_details: Optional[Dict[str, float]] = None, prompt: Optional[PromptClient] = None, ) -> Any: with self._otel_tracer.start_as_current_span( name=name, end_on_exit=end_on_exit if end_on_exit is not None else True, ) as otel_span: span_class = self._get_span_class( as_type or "generation" ) # default was "generation" common_args = { "otel_span": otel_span, "langfuse_client": self, "environment": self._environment, "release": self._release, "input": input, "output": output, "metadata": metadata, "version": version, "level": level, "status_message": status_message, } if span_class in [ LangfuseGeneration, LangfuseEmbedding, ]: common_args.update( { "completion_start_time": completion_start_time, "model": model, "model_parameters": model_parameters, "usage_details": usage_details, "cost_details": cost_details, "prompt": prompt, } ) # For span-like types (span, agent, tool, chain, retriever, evaluator, guardrail), no generation properties needed yield span_class(**common_args) # type: ignore[arg-type] def _get_current_otel_span(self) -> Optional[otel_trace_api.Span]: current_span = otel_trace_api.get_current_span() if current_span is otel_trace_api.INVALID_SPAN: langfuse_logger.warning( "Context error: No active span in current context. Operations that depend on an active span will be skipped. " "Ensure spans are created with start_as_current_observation() or that you're operating within an active span context." ) return None return current_span def update_current_generation( self, *, name: Optional[str] = None, input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, completion_start_time: Optional[datetime] = None, model: Optional[str] = None, model_parameters: Optional[Dict[str, MapValue]] = None, usage_details: Optional[Dict[str, int]] = None, cost_details: Optional[Dict[str, float]] = None, prompt: Optional[PromptClient] = None, ) -> None: """Update the current active generation span with new information. This method updates the current generation span in the active context with additional information. It's useful for adding output, usage stats, or other details that become available during or after model generation. Args: name: The generation name input: Updated input data for the model output: Output from the model (e.g., completions) metadata: Additional metadata to associate with the generation version: Version identifier for the model or component level: Importance level of the generation (info, warning, error) status_message: Optional status message for the generation completion_start_time: When the model started generating the response model: Name/identifier of the AI model used (e.g., "gpt-4") model_parameters: Parameters used for the model (e.g., temperature, max_tokens) usage_details: Token usage information (e.g., prompt_tokens, completion_tokens) cost_details: Cost information for the model call prompt: Associated prompt template from Langfuse prompt management Example: ```python with langfuse.start_as_current_generation(name="answer-query") as generation: # Initial setup and API call response = llm.generate(...) # Update with results that weren't available at creation time langfuse.update_current_generation( output=response.text, usage_details={ "prompt_tokens": response.usage.prompt_tokens, "completion_tokens": response.usage.completion_tokens } ) ``` """ if not self._tracing_enabled: langfuse_logger.debug( "Operation skipped: update_current_generation - Tracing is disabled or client is in no-op mode." ) return current_otel_span = self._get_current_otel_span() if current_otel_span is not None: generation = LangfuseGeneration( otel_span=current_otel_span, langfuse_client=self ) if name: current_otel_span.update_name(name) generation.update( input=input, output=output, metadata=metadata, version=version, level=level, status_message=status_message, completion_start_time=completion_start_time, model=model, model_parameters=model_parameters, usage_details=usage_details, cost_details=cost_details, prompt=prompt, ) def update_current_span( self, *, name: Optional[str] = None, input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, ) -> None: """Update the current active span with new information. This method updates the current span in the active context with additional information. It's useful for adding outputs or metadata that become available during execution. Args: name: The span name input: Updated input data for the operation output: Output data from the operation metadata: Additional metadata to associate with the span version: Version identifier for the code or component level: Importance level of the span (info, warning, error) status_message: Optional status message for the span Example: ```python with langfuse.start_as_current_observation(name="process-data") as span: # Initial processing result = process_first_part() # Update with intermediate results langfuse.update_current_span(metadata={"intermediate_result": result}) # Continue processing final_result = process_second_part(result) # Final update langfuse.update_current_span(output=final_result) ``` """ if not self._tracing_enabled: langfuse_logger.debug( "Operation skipped: update_current_span - Tracing is disabled or client is in no-op mode." ) return current_otel_span = self._get_current_otel_span() if current_otel_span is not None: span = LangfuseSpan( otel_span=current_otel_span, langfuse_client=self, environment=self._environment, release=self._release, ) if name: current_otel_span.update_name(name) span.update( input=input, output=output, metadata=metadata, version=version, level=level, status_message=status_message, ) @deprecated( "Trace-level input/output is deprecated. " "For trace attributes (user_id, session_id, tags, etc.), use propagate_attributes() instead. " "This method will be removed in a future major version." ) def set_current_trace_io( self, *, input: Optional[Any] = None, output: Optional[Any] = None, ) -> None: """Set trace-level input and output for the current span's trace. .. deprecated:: This is a legacy method for backward compatibility with Langfuse platform features that still rely on trace-level input/output (e.g., legacy LLM-as-a-judge evaluators). It will be removed in a future major version. For setting other trace attributes (user_id, session_id, metadata, tags, version), use :meth:`propagate_attributes` instead. Args: input: Input data to associate with the trace. output: Output data to associate with the trace. """ if not self._tracing_enabled: langfuse_logger.debug( "Operation skipped: set_current_trace_io - Tracing is disabled or client is in no-op mode." ) return current_otel_span = self._get_current_otel_span() if current_otel_span is not None and current_otel_span.is_recording(): existing_observation_type = current_otel_span.attributes.get( # type: ignore[attr-defined] LangfuseOtelSpanAttributes.OBSERVATION_TYPE, "span" ) # We need to preserve the class to keep the correct observation type span_class = self._get_span_class(existing_observation_type) span = span_class( otel_span=current_otel_span, langfuse_client=self, environment=self._environment, release=self._release, ) span.set_trace_io( input=input, output=output, ) def set_current_trace_as_public(self) -> None: """Make the current trace publicly accessible via its URL. When a trace is published, anyone with the trace link can view the full trace without needing to be logged in to Langfuse. This action cannot be undone programmatically - once published, the entire trace becomes public. This is a convenience method that publishes the trace from the currently active span context. Use this when you want to make a trace public from within a traced function without needing direct access to the span object. """ if not self._tracing_enabled: langfuse_logger.debug( "Operation skipped: set_current_trace_as_public - Tracing is disabled or client is in no-op mode." ) return current_otel_span = self._get_current_otel_span() if current_otel_span is not None and current_otel_span.is_recording(): existing_observation_type = current_otel_span.attributes.get( # type: ignore[attr-defined] LangfuseOtelSpanAttributes.OBSERVATION_TYPE, "span" ) # We need to preserve the class to keep the correct observation type span_class = self._get_span_class(existing_observation_type) span = span_class( otel_span=current_otel_span, langfuse_client=self, environment=self._environment, ) span.set_trace_as_public() def create_event( self, *, trace_context: Optional[TraceContext] = None, name: str, input: Optional[Any] = None, output: Optional[Any] = None, metadata: Optional[Any] = None, version: Optional[str] = None, level: Optional[SpanLevel] = None, status_message: Optional[str] = None, ) -> LangfuseEvent: """Create a new Langfuse observation of type 'EVENT'. The created Langfuse Event observation will be the child of the current span in the context. Args: trace_context: Optional context for connecting to an existing trace name: Name of the span (e.g., function or operation name) input: Input data for the operation (can be any JSON-serializable object) output: Output data from the operation (can be any JSON-serializable object) metadata: Additional metadata to associate with the span version: Version identifier for the code or component level: Importance level of the span (info, warning, error) status_message: Optional status message for the span Returns: The Langfuse Event object Example: ```python event = langfuse.create_event(name="process-event") ``` """ timestamp = time_ns() if trace_context: trace_id = trace_context.get("trace_id", None) parent_span_id = trace_context.get("parent_span_id", None) if trace_id: remote_parent_span = self._create_remote_parent_span( trace_id=trace_id, parent_span_id=parent_span_id ) with otel_trace_api.use_span( cast(otel_trace_api.Span, remote_parent_span) ): otel_span = self._otel_tracer.start_span( name=name, start_time=timestamp ) otel_span.set_attribute(LangfuseOtelSpanAttributes.AS_ROOT, True) return cast( LangfuseEvent, LangfuseEvent( otel_span=otel_span, langfuse_client=self, environment=self._environment, release=self._release, input=input, output=output, metadata=metadata, version=version, level=level, status_message=status_message, ).end(end_time=timestamp), ) otel_span = self._otel_tracer.start_span(name=name, start_time=timestamp) return cast( LangfuseEvent, LangfuseEvent( otel_span=otel_span, langfuse_client=self, environment=self._environment, release=self._release, input=input, output=output, metadata=metadata, version=version, level=level, status_message=status_message, ).end(end_time=timestamp), ) def _create_remote_parent_span( self, *, trace_id: str, parent_span_id: Optional[str] ) -> Any: if not self._is_valid_trace_id(trace_id): langfuse_logger.warning( f"Passed trace ID '{trace_id}' is not a valid 32 lowercase hex char Langfuse trace id. Ignoring trace ID." ) if parent_span_id and not self._is_valid_span_id(parent_span_id): langfuse_logger.warning( f"Passed span ID '{parent_span_id}' is not a valid 16 lowercase hex char Langfuse span id. Ignoring parent span ID." ) int_trace_id = int(trace_id, 16) int_parent_span_id = ( int(parent_span_id, 16) if parent_span_id else RandomIdGenerator().generate_span_id() ) span_context = otel_trace_api.SpanContext( trace_id=int_trace_id, span_id=int_parent_span_id, trace_flags=otel_trace_api.TraceFlags(0x01), # mark span as sampled is_remote=False, ) return otel_trace_api.NonRecordingSpan(span_context) def _is_valid_trace_id(self, trace_id: str) -> bool: pattern = r"^[0-9a-f]{32}$" return bool(re.match(pattern, trace_id)) def _is_valid_span_id(self, span_id: str) -> bool: pattern = r"^[0-9a-f]{16}$" return bool(re.match(pattern, span_id)) def _create_observation_id(self, *, seed: Optional[str] = None) -> str: """Create a unique observation ID for use with Langfuse. This method generates a unique observation ID (span ID in OpenTelemetry terms) for use with various Langfuse APIs. It can either generate a random ID or create a deterministic ID based on a seed string. Observation IDs must be 16 lowercase hexadecimal characters, representing 8 bytes. This method ensures the generated ID meets this requirement. If you need to correlate an external ID with a Langfuse observation ID, use the external ID as the seed to get a valid, deterministic observation ID. Args: seed: Optional string to use as a seed for deterministic ID generation. If provided, the same seed will always produce the same ID. If not provided, a random ID will be generated. Returns: A 16-character lowercase hexadecimal string representing the observation ID. Example: ```python # Generate a random observation ID obs_id = langfuse.create_observation_id() # Generate a deterministic ID based on a seed user_obs_id = langfuse.create_observation_id(seed="user-123-feedback") # Correlate an external item ID with a Langfuse observation ID item_id = "item-789012" correlated_obs_id = langfuse.create_observation_id(seed=item_id) # Use the ID with Langfuse APIs langfuse.create_score( name="relevance", value=0.95, trace_id=trace_id, observation_id=obs_id ) ``` """ if not seed: span_id_int = RandomIdGenerator().generate_span_id() return self._format_otel_span_id(span_id_int) return sha256(seed.encode("utf-8")).digest()[:8].hex() @staticmethod def create_trace_id(*, seed: Optional[str] = None) -> str: """Create a unique trace ID for use with Langfuse. This method generates a unique trace ID for use with various Langfuse APIs. It can either generate a random ID or create a deterministic ID based on a seed string. Trace IDs must be 32 lowercase hexadecimal characters, representing 16 bytes. This method ensures the generated ID meets this requirement. If you need to correlate an external ID with a Langfuse trace ID, use the external ID as the seed to get a valid, deterministic Langfuse trace ID. Args: seed: Optional string to use as a seed for deterministic ID generation. If provided, the same seed will always produce the same ID. If not provided, a random ID will be generated. Returns: A 32-character lowercase hexadecimal string representing the Langfuse trace ID. Example: ```python # Generate a random trace ID trace_id = langfuse.create_trace_id() # Generate a deterministic ID based on a seed session_trace_id = langfuse.create_trace_id(seed="session-456") # Correlate an external ID with a Langfuse trace ID external_id = "external-system-123456" correlated_trace_id = langfuse.create_trace_id(seed=external_id) # Use the ID with trace context with langfuse.start_as_current_observation( name="process-request", trace_context={"trace_id": trace_id} ) as span: # Operation will be part of the specific trace pass ``` """ if not seed: trace_id_int = RandomIdGenerator().generate_trace_id() return Langfuse._format_otel_trace_id(trace_id_int) return sha256(seed.encode("utf-8")).digest()[:16].hex() def _get_otel_trace_id(self, otel_span: otel_trace_api.Span) -> str: span_context = otel_span.get_span_context() return self._format_otel_trace_id(span_context.trace_id) def _get_otel_span_id(self, otel_span: otel_trace_api.Span) -> str: span_context = otel_span.get_span_context() return self._format_otel_span_id(span_context.span_id) @staticmethod def _format_otel_span_id(span_id_int: int) -> str: """Format an integer span ID to a 16-character lowercase hex string. Internal method to convert an OpenTelemetry integer span ID to the standard W3C Trace Context format (16-character lowercase hex string). Args: span_id_int: 64-bit integer representing a span ID Returns: A 16-character lowercase hexadecimal string """ return format(span_id_int, "016x") @staticmethod def _format_otel_trace_id(trace_id_int: int) -> str: """Format an integer trace ID to a 32-character lowercase hex string. Internal method to convert an OpenTelemetry integer trace ID to the standard W3C Trace Context format (32-character lowercase hex string). Args: trace_id_int: 128-bit integer representing a trace ID Returns: A 32-character lowercase hexadecimal string """ return format(trace_id_int, "032x") @overload def create_score( self, *, name: str, value: float, session_id: Optional[str] = None, dataset_run_id: Optional[str] = None, trace_id: Optional[str] = None, observation_id: Optional[str] = None, score_id: Optional[str] = None, data_type: Optional[Literal["NUMERIC", "BOOLEAN"]] = None, comment: Optional[str] = None, config_id: Optional[str] = None, metadata: Optional[Any] = None, timestamp: Optional[datetime] = None, ) -> None: ... @overload def create_score( self, *, name: str, value: str, session_id: Optional[str] = None, dataset_run_id: Optional[str] = None, trace_id: Optional[str] = None, score_id: Optional[str] = None, observation_id: Optional[str] = None, data_type: Optional[Literal["CATEGORICAL"]] = "CATEGORICAL", comment: Optional[str] = None, config_id: Optional[str] = None, metadata: Optional[Any] = None, timestamp: Optional[datetime] = None, ) -> None: ... def create_score( self, *, name: str, value: Union[float, str], session_id: Optional[str] = None, dataset_run_id: Optional[str] = None, trace_id: Optional[str] = None, observation_id: Optional[str] = None, score_id: Optional[str] = None, data_type: Optional[ScoreDataType] = None, comment: Optional[str] = None, config_id: Optional[str] = None, metadata: Optional[Any] = None, timestamp: Optional[datetime] = None, ) -> None: """Create a score for a specific trace or observation. This method creates a score for evaluating a Langfuse trace or observation. Scores can be used to track quality metrics, user feedback, or automated evaluations. Args: name: Name of the score (e.g., "relevance", "accuracy") value: Score value (can be numeric for NUMERIC/BOOLEAN types or string for CATEGORICAL) session_id: ID of the Langfuse session to associate the score with dataset_run_id: ID of the Langfuse dataset run to associate the score with trace_id: ID of the Langfuse trace to associate the score with observation_id: Optional ID of the specific observation to score. Trace ID must be provided too. score_id: Optional custom ID for the score (auto-generated if not provided) data_type: Type of score (NUMERIC, BOOLEAN, or CATEGORICAL) comment: Optional comment or explanation for the score config_id: Optional ID of a score config defined in Langfuse metadata: Optional metadata to be attached to the score timestamp: Optional timestamp for the score (defaults to current UTC time) Example: ```python # Create a numeric score for accuracy langfuse.create_score( name="accuracy", value=0.92, trace_id="abcdef1234567890abcdef1234567890", data_type="NUMERIC", comment="High accuracy with minor irrelevant details" ) # Create a categorical score for sentiment langfuse.create_score( name="sentiment", value="positive", trace_id="abcdef1234567890abcdef1234567890", observation_id="abcdef1234567890", data_type="CATEGORICAL" ) ``` """ if not self._tracing_enabled: return score_id = score_id or self._create_observation_id() try: new_body = ScoreBody( id=score_id, session_id=session_id, datasetRunId=dataset_run_id, traceId=trace_id, observationId=observation_id, name=name, value=value, dataType=data_type, # type: ignore comment=comment, configId=config_id, environment=self._environment, metadata=metadata, ) event = { "id": self.create_trace_id(), "type": "score-create", "timestamp": timestamp or _get_timestamp(), "body": new_body, } if self._resources is not None: # Force the score to be in sample if it was for a legacy trace ID, i.e. non-32 hexchar force_sample = ( not self._is_valid_trace_id(trace_id) if trace_id else True ) self._resources.add_score_task( event, force_sample=force_sample, ) except Exception as e: langfuse_logger.exception( f"Error creating score: Failed to process score event for trace_id={trace_id}, name={name}. Error: {e}" ) def _create_trace_tags_via_ingestion( self, *, trace_id: str, tags: List[str], ) -> None: """Private helper to enqueue trace tag updates via ingestion API events.""" if not self._tracing_enabled: return if len(tags) == 0: return try: new_body = TraceBody( id=trace_id, tags=tags, ) event = { "id": self.create_trace_id(), "type": "trace-create", "timestamp": _get_timestamp(), "body": new_body, } if self._resources is not None: self._resources.add_trace_task(event) except Exception as e: langfuse_logger.exception( f"Error updating trace tags: Failed to process trace update event for trace_id={trace_id}. Error: {e}" ) @overload def score_current_span( self, *, name: str, value: float, score_id: Optional[str] = None, data_type: Optional[Literal["NUMERIC", "BOOLEAN"]] = None, comment: Optional[str] = None, config_id: Optional[str] = None, metadata: Optional[Any] = None, ) -> None: ... @overload def score_current_span( self, *, name: str, value: str, score_id: Optional[str] = None, data_type: Optional[Literal["CATEGORICAL"]] = "CATEGORICAL", comment: Optional[str] = None, config_id: Optional[str] = None, metadata: Optional[Any] = None, ) -> None: ... def score_current_span( self, *, name: str, value: Union[float, str], score_id: Optional[str] = None, data_type: Optional[ScoreDataType] = None, comment: Optional[str] = None, config_id: Optional[str] = None, metadata: Optional[Any] = None, ) -> None: """Create a score for the current active span. This method scores the currently active span in the context. It's a convenient way to score the current operation without needing to know its trace and span IDs. Args: name: Name of the score (e.g., "relevance", "accuracy") value: Score value (can be numeric for NUMERIC/BOOLEAN types or string for CATEGORICAL) score_id: Optional custom ID for the score (auto-generated if not provided) data_type: Type of score (NUMERIC, BOOLEAN, or CATEGORICAL) comment: Optional comment or explanation for the score config_id: Optional ID of a score config defined in Langfuse metadata: Optional metadata to be attached to the score Example: ```python with langfuse.start_as_current_generation(name="answer-query") as generation: # Generate answer response = generate_answer(...) generation.update(output=response) # Score the generation langfuse.score_current_span( name="relevance", value=0.85, data_type="NUMERIC", comment="Mostly relevant but contains some tangential information", metadata={"model": "gpt-4", "prompt_version": "v2"} ) ``` """ current_span = self._get_current_otel_span() if current_span is not None: trace_id = self._get_otel_trace_id(current_span) observation_id = self._get_otel_span_id(current_span) langfuse_logger.info( f"Score: Creating score name='{name}' value={value} for current span ({observation_id}) in trace {trace_id}" ) self.create_score( trace_id=trace_id, observation_id=observation_id, name=name, value=cast(str, value), score_id=score_id, data_type=cast(Literal["CATEGORICAL"], data_type), comment=comment, config_id=config_id, metadata=metadata, ) @overload def score_current_trace( self, *, name: str, value: float, score_id: Optional[str] = None, data_type: Optional[Literal["NUMERIC", "BOOLEAN"]] = None, comment: Optional[str] = None, config_id: Optional[str] = None, metadata: Optional[Any] = None, ) -> None: ... @overload def score_current_trace( self, *, name: str, value: str, score_id: Optional[str] = None, data_type: Optional[Literal["CATEGORICAL"]] = "CATEGORICAL", comment: Optional[str] = None, config_id: Optional[str] = None, metadata: Optional[Any] = None, ) -> None: ... def score_current_trace( self, *, name: str, value: Union[float, str], score_id: Optional[str] = None, data_type: Optional[ScoreDataType] = None, comment: Optional[str] = None, config_id: Optional[str] = None, metadata: Optional[Any] = None, ) -> None: """Create a score for the current trace. This method scores the trace of the currently active span. Unlike score_current_span, this method associates the score with the entire trace rather than a specific span. It's useful for scoring overall performance or quality of the entire operation. Args: name: Name of the score (e.g., "user_satisfaction", "overall_quality") value: Score value (can be numeric for NUMERIC/BOOLEAN types or string for CATEGORICAL) score_id: Optional custom ID for the score (auto-generated if not provided) data_type: Type of score (NUMERIC, BOOLEAN, or CATEGORICAL) comment: Optional comment or explanation for the score config_id: Optional ID of a score config defined in Langfuse metadata: Optional metadata to be attached to the score Example: ```python with langfuse.start_as_current_observation(name="process-user-request") as span: # Process request result = process_complete_request() span.update(output=result) # Score the overall trace langfuse.score_current_trace( name="overall_quality", value=0.95, data_type="NUMERIC", comment="High quality end-to-end response", metadata={"evaluator": "gpt-4", "criteria": "comprehensive"} ) ``` """ current_span = self._get_current_otel_span() if current_span is not None: trace_id = self._get_otel_trace_id(current_span) langfuse_logger.info( f"Score: Creating score name='{name}' value={value} for entire trace {trace_id}" ) self.create_score( trace_id=trace_id, name=name, value=cast(str, value), score_id=score_id, data_type=cast(Literal["CATEGORICAL"], data_type), comment=comment, config_id=config_id, metadata=metadata, ) def flush(self) -> None: """Force flush all pending spans and events to the Langfuse API. This method manually flushes any pending spans, scores, and other events to the Langfuse API. It's useful in scenarios where you want to ensure all data is sent before proceeding, without waiting for the automatic flush interval. Example: ```python # Record some spans and scores with langfuse.start_as_current_observation(name="operation") as span: # Do work... pass # Ensure all data is sent to Langfuse before proceeding langfuse.flush() # Continue with other work ``` """ if self._resources is not None: self._resources.flush() def shutdown(self) -> None: """Shut down the Langfuse client and flush all pending data. This method cleanly shuts down the Langfuse client, ensuring all pending data is flushed to the API and all background threads are properly terminated. It's important to call this method when your application is shutting down to prevent data loss and resource leaks. For most applications, using the client as a context manager or relying on the automatic shutdown via atexit is sufficient. Example: ```python # Initialize Langfuse langfuse = Langfuse(public_key="...", secret_key="...") # Use Langfuse throughout your application # ... # When application is shutting down langfuse.shutdown() ``` """ if self._resources is not None: self._resources.shutdown() def get_current_trace_id(self) -> Optional[str]: """Get the trace ID of the current active span. This method retrieves the trace ID from the currently active span in the context. It can be used to get the trace ID for referencing in logs, external systems, or for creating related operations. Returns: The current trace ID as a 32-character lowercase hexadecimal string, or None if there is no active span. Example: ```python with langfuse.start_as_current_observation(name="process-request") as span: # Get the current trace ID for reference trace_id = langfuse.get_current_trace_id() # Use it for external correlation log.info(f"Processing request with trace_id: {trace_id}") # Or pass to another system external_system.process(data, trace_id=trace_id) ``` """ if not self._tracing_enabled: langfuse_logger.debug( "Operation skipped: get_current_trace_id - Tracing is disabled or client is in no-op mode." ) return None current_otel_span = self._get_current_otel_span() return self._get_otel_trace_id(current_otel_span) if current_otel_span else None def get_current_observation_id(self) -> Optional[str]: """Get the observation ID (span ID) of the current active span. This method retrieves the observation ID from the currently active span in the context. It can be used to get the observation ID for referencing in logs, external systems, or for creating scores or other related operations. Returns: The current observation ID as a 16-character lowercase hexadecimal string, or None if there is no active span. Example: ```python with langfuse.start_as_current_observation(name="process-user-query") as span: # Get the current observation ID observation_id = langfuse.get_current_observation_id() # Store it for later reference cache.set(f"query_{query_id}_observation", observation_id) # Process the query... ``` """ if not self._tracing_enabled: langfuse_logger.debug( "Operation skipped: get_current_observation_id - Tracing is disabled or client is in no-op mode." ) return None current_otel_span = self._get_current_otel_span() return self._get_otel_span_id(current_otel_span) if current_otel_span else None def _get_project_id(self) -> Optional[str]: """Fetch and return the current project id. Persisted across requests. Returns None if no project id is found for api keys.""" if not self._project_id: proj = self.api.projects.get() if not proj.data or not proj.data[0].id: return None self._project_id = proj.data[0].id return self._project_id def get_trace_url(self, *, trace_id: Optional[str] = None) -> Optional[str]: """Get the URL to view a trace in the Langfuse UI. This method generates a URL that links directly to a trace in the Langfuse UI. It's useful for providing links in logs, notifications, or debugging tools. Args: trace_id: Optional trace ID to generate a URL for. If not provided, the trace ID of the current active span will be used. Returns: A URL string pointing to the trace in the Langfuse UI, or None if the project ID couldn't be retrieved or no trace ID is available. Example: ```python # Get URL for the current trace with langfuse.start_as_current_observation(name="process-request") as span: trace_url = langfuse.get_trace_url() log.info(f"Processing trace: {trace_url}") # Get URL for a specific trace specific_trace_url = langfuse.get_trace_url(trace_id="1234567890abcdef1234567890abcdef") send_notification(f"Review needed for trace: {specific_trace_url}") ``` """ final_trace_id = trace_id or self.get_current_trace_id() if not final_trace_id: return None project_id = self._get_project_id() return ( f"{self._base_url}/project/{project_id}/traces/{final_trace_id}" if project_id and final_trace_id else None ) def get_dataset( self, name: str, *, fetch_items_page_size: Optional[int] = 50, version: Optional[datetime] = None, ) -> "DatasetClient": """Fetch a dataset by its name. Args: name (str): The name of the dataset to fetch. fetch_items_page_size (Optional[int]): All items of the dataset will be fetched in chunks of this size. Defaults to 50. version (Optional[datetime]): Retrieve dataset items as they existed at this specific point in time (UTC). If provided, returns the state of items at the specified UTC timestamp. If not provided, returns the latest version. Must be a timezone-aware datetime object in UTC. Returns: DatasetClient: The dataset with the given name. """ try: langfuse_logger.debug(f"Getting datasets {name}") dataset = self.api.datasets.get(dataset_name=self._url_encode(name)) dataset_items = [] page = 1 while True: new_items = self.api.dataset_items.list( dataset_name=self._url_encode(name, is_url_param=True), page=page, limit=fetch_items_page_size, version=version, ) dataset_items.extend(new_items.data) if new_items.meta.total_pages <= page: break page += 1 return DatasetClient( dataset=dataset, items=dataset_items, version=version, langfuse_client=self, ) except Error as e: handle_fern_exception(e) raise e def get_dataset_run( self, *, dataset_name: str, run_name: str ) -> DatasetRunWithItems: """Fetch a dataset run by dataset name and run name. Args: dataset_name (str): The name of the dataset. run_name (str): The name of the run. Returns: DatasetRunWithItems: The dataset run with its items. """ try: return cast( DatasetRunWithItems, self.api.datasets.get_run( dataset_name=self._url_encode(dataset_name), run_name=self._url_encode(run_name), request_options=None, ), ) except Error as e: handle_fern_exception(e) raise e def get_dataset_runs( self, *, dataset_name: str, page: Optional[int] = None, limit: Optional[int] = None, ) -> PaginatedDatasetRuns: """Fetch all runs for a dataset. Args: dataset_name (str): The name of the dataset. page (Optional[int]): Page number, starts at 1. limit (Optional[int]): Limit of items per page. Returns: PaginatedDatasetRuns: Paginated list of dataset runs. """ try: return cast( PaginatedDatasetRuns, self.api.datasets.get_runs( dataset_name=self._url_encode(dataset_name), page=page, limit=limit, request_options=None, ), ) except Error as e: handle_fern_exception(e) raise e def delete_dataset_run( self, *, dataset_name: str, run_name: str ) -> DeleteDatasetRunResponse: """Delete a dataset run and all its run items. This action is irreversible. Args: dataset_name (str): The name of the dataset. run_name (str): The name of the run. Returns: DeleteDatasetRunResponse: Confirmation of deletion. """ try: return cast( DeleteDatasetRunResponse, self.api.datasets.delete_run( dataset_name=self._url_encode(dataset_name), run_name=self._url_encode(run_name), request_options=None, ), ) except Error as e: handle_fern_exception(e) raise e def run_experiment( self, *, name: str, run_name: Optional[str] = None, description: Optional[str] = None, data: ExperimentData, task: TaskFunction, evaluators: List[EvaluatorFunction] = [], composite_evaluator: Optional[CompositeEvaluatorFunction] = None, run_evaluators: List[RunEvaluatorFunction] = [], max_concurrency: int = 50, metadata: Optional[Dict[str, str]] = None, _dataset_version: Optional[datetime] = None, ) -> ExperimentResult: """Run an experiment on a dataset with automatic tracing and evaluation. This method executes a task function on each item in the provided dataset, automatically traces all executions with Langfuse for observability, runs item-level and run-level evaluators on the outputs, and returns comprehensive results with evaluation metrics. The experiment system provides: - Automatic tracing of all task executions - Concurrent processing with configurable limits - Comprehensive error handling that isolates failures - Integration with Langfuse datasets for experiment tracking - Flexible evaluation framework supporting both sync and async evaluators Args: name: Human-readable name for the experiment. Used for identification in the Langfuse UI. run_name: Optional exact name for the experiment run. If provided, this will be used as the exact dataset run name if the `data` contains Langfuse dataset items. If not provided, this will default to the experiment name appended with an ISO timestamp. description: Optional description explaining the experiment's purpose, methodology, or expected outcomes. data: Array of data items to process. Can be either: - List of dict-like items with 'input', 'expected_output', 'metadata' keys - List of Langfuse DatasetItem objects from dataset.items task: Function that processes each data item and returns output. Must accept 'item' as keyword argument and can return sync or async results. The task function signature should be: task(*, item, **kwargs) -> Any evaluators: List of functions to evaluate each item's output individually. Each evaluator receives input, output, expected_output, and metadata. Can return single Evaluation dict or list of Evaluation dicts. composite_evaluator: Optional function that creates composite scores from item-level evaluations. Receives the same inputs as item-level evaluators (input, output, expected_output, metadata) plus the list of evaluations from item-level evaluators. Useful for weighted averages, pass/fail decisions based on multiple criteria, or custom scoring logic combining multiple metrics. run_evaluators: List of functions to evaluate the entire experiment run. Each run evaluator receives all item_results and can compute aggregate metrics. Useful for calculating averages, distributions, or cross-item comparisons. max_concurrency: Maximum number of concurrent task executions (default: 50). Controls the number of items processed simultaneously. Adjust based on API rate limits and system resources. metadata: Optional metadata dictionary to attach to all experiment traces. This metadata will be included in every trace created during the experiment. If `data` are Langfuse dataset items, the metadata will be attached to the dataset run, too. Returns: ExperimentResult containing: - run_name: The experiment run name. This is equal to the dataset run name if experiment was on Langfuse dataset. - item_results: List of results for each processed item with outputs and evaluations - run_evaluations: List of aggregate evaluation results for the entire run - dataset_run_id: ID of the dataset run (if using Langfuse datasets) - dataset_run_url: Direct URL to view results in Langfuse UI (if applicable) Raises: ValueError: If required parameters are missing or invalid Exception: If experiment setup fails (individual item failures are handled gracefully) Examples: Basic experiment with local data: ```python def summarize_text(*, item, **kwargs): return f"Summary: {item['input'][:50]}..." def length_evaluator(*, input, output, expected_output=None, **kwargs): return { "name": "output_length", "value": len(output), "comment": f"Output contains {len(output)} characters" } result = langfuse.run_experiment( name="Text Summarization Test", description="Evaluate summarization quality and length", data=[ {"input": "Long article text...", "expected_output": "Expected summary"}, {"input": "Another article...", "expected_output": "Another summary"} ], task=summarize_text, evaluators=[length_evaluator] ) print(f"Processed {len(result.item_results)} items") for item_result in result.item_results: print(f"Input: {item_result.item['input']}") print(f"Output: {item_result.output}") print(f"Evaluations: {item_result.evaluations}") ``` Advanced experiment with async task and multiple evaluators: ```python async def llm_task(*, item, **kwargs): # Simulate async LLM call response = await openai_client.chat.completions.create( model="gpt-4", messages=[{"role": "user", "content": item["input"]}] ) return response.choices[0].message.content def accuracy_evaluator(*, input, output, expected_output=None, **kwargs): if expected_output and expected_output.lower() in output.lower(): return {"name": "accuracy", "value": 1.0, "comment": "Correct answer"} return {"name": "accuracy", "value": 0.0, "comment": "Incorrect answer"} def toxicity_evaluator(*, input, output, expected_output=None, **kwargs): # Simulate toxicity check toxicity_score = check_toxicity(output) # Your toxicity checker return { "name": "toxicity", "value": toxicity_score, "comment": f"Toxicity level: {'high' if toxicity_score > 0.7 else 'low'}" } def average_accuracy(*, item_results, **kwargs): accuracies = [ eval.value for result in item_results for eval in result.evaluations if eval.name == "accuracy" ] return { "name": "average_accuracy", "value": sum(accuracies) / len(accuracies) if accuracies else 0, "comment": f"Average accuracy across {len(accuracies)} items" } result = langfuse.run_experiment( name="LLM Safety and Accuracy Test", description="Evaluate model accuracy and safety across diverse prompts", data=test_dataset, # Your dataset items task=llm_task, evaluators=[accuracy_evaluator, toxicity_evaluator], run_evaluators=[average_accuracy], max_concurrency=5, # Limit concurrent API calls metadata={"model": "gpt-4", "temperature": 0.7} ) ``` Using with Langfuse datasets: ```python # Get dataset from Langfuse dataset = langfuse.get_dataset("my-eval-dataset") result = dataset.run_experiment( name="Production Model Evaluation", description="Monthly evaluation of production model performance", task=my_production_task, evaluators=[accuracy_evaluator, latency_evaluator] ) # Results automatically linked to dataset in Langfuse UI print(f"View results: {result['dataset_run_url']}") ``` Note: - Task and evaluator functions can be either synchronous or asynchronous - Individual item failures are logged but don't stop the experiment - All executions are automatically traced and visible in Langfuse UI - When using Langfuse datasets, results are automatically linked for easy comparison - This method works in both sync and async contexts (Jupyter notebooks, web apps, etc.) - Async execution is handled automatically with smart event loop detection """ return cast( ExperimentResult, run_async_safely( self._run_experiment_async( name=name, run_name=self._create_experiment_run_name( name=name, run_name=run_name ), description=description, data=data, task=task, evaluators=evaluators or [], composite_evaluator=composite_evaluator, run_evaluators=run_evaluators or [], max_concurrency=max_concurrency, metadata=metadata, dataset_version=_dataset_version, ), ), ) async def _run_experiment_async( self, *, name: str, run_name: str, description: Optional[str], data: ExperimentData, task: TaskFunction, evaluators: List[EvaluatorFunction], composite_evaluator: Optional[CompositeEvaluatorFunction], run_evaluators: List[RunEvaluatorFunction], max_concurrency: int, metadata: Optional[Dict[str, Any]] = None, dataset_version: Optional[datetime] = None, ) -> ExperimentResult: langfuse_logger.debug( f"Starting experiment '{name}' run '{run_name}' with {len(data)} items" ) # Set up concurrency control semaphore = asyncio.Semaphore(max_concurrency) # Process all items async def process_item(item: ExperimentItem) -> ExperimentItemResult: async with semaphore: return await self._process_experiment_item( item, task, evaluators, composite_evaluator, name, run_name, description, metadata, dataset_version, ) # Run all items concurrently tasks = [process_item(item) for item in data] item_results = await asyncio.gather(*tasks, return_exceptions=True) # Filter out any exceptions and log errors valid_results: List[ExperimentItemResult] = [] for i, result in enumerate(item_results): if isinstance(result, Exception): langfuse_logger.error(f"Item {i} failed: {result}") elif isinstance(result, ExperimentItemResult): valid_results.append(result) # type: ignore # Run experiment-level evaluators run_evaluations: List[Evaluation] = [] for run_evaluator in run_evaluators: try: evaluations = await _run_evaluator( run_evaluator, item_results=valid_results ) run_evaluations.extend(evaluations) except Exception as e: langfuse_logger.error(f"Run evaluator failed: {e}") # Generate dataset run URL if applicable dataset_run_id = valid_results[0].dataset_run_id if valid_results else None dataset_run_url = None if dataset_run_id and data: try: # Check if the first item has dataset_id (for DatasetItem objects) first_item = data[0] dataset_id = None if hasattr(first_item, "dataset_id"): dataset_id = getattr(first_item, "dataset_id", None) if dataset_id: project_id = self._get_project_id() if project_id: dataset_run_url = f"{self._base_url}/project/{project_id}/datasets/{dataset_id}/runs/{dataset_run_id}" except Exception: pass # URL generation is optional # Store run-level evaluations as scores for evaluation in run_evaluations: try: if dataset_run_id: self.create_score( dataset_run_id=dataset_run_id, name=evaluation.name or "", value=evaluation.value, # type: ignore comment=evaluation.comment, metadata=evaluation.metadata, data_type=evaluation.data_type, # type: ignore config_id=evaluation.config_id, ) except Exception as e: langfuse_logger.error(f"Failed to store run evaluation: {e}") # Flush scores and traces self.flush() return ExperimentResult( name=name, run_name=run_name, description=description, item_results=valid_results, run_evaluations=run_evaluations, dataset_run_id=dataset_run_id, dataset_run_url=dataset_run_url, ) async def _process_experiment_item( self, item: ExperimentItem, task: Callable, evaluators: List[Callable], composite_evaluator: Optional[CompositeEvaluatorFunction], experiment_name: str, experiment_run_name: str, experiment_description: Optional[str], experiment_metadata: Optional[Dict[str, Any]] = None, dataset_version: Optional[datetime] = None, ) -> ExperimentItemResult: span_name = "experiment-item-run" with self.start_as_current_observation(name=span_name) as span: try: input_data = ( item.get("input") if isinstance(item, dict) else getattr(item, "input", None) ) if input_data is None: raise ValueError("Experiment Item is missing input. Skipping item.") expected_output = ( item.get("expected_output") if isinstance(item, dict) else getattr(item, "expected_output", None) ) item_metadata = ( item.get("metadata") if isinstance(item, dict) else getattr(item, "metadata", None) ) final_observation_metadata = { "experiment_name": experiment_name, "experiment_run_name": experiment_run_name, **(experiment_metadata or {}), } trace_id = span.trace_id dataset_id = None dataset_item_id = None dataset_run_id = None # Link to dataset run if this is a dataset item if hasattr(item, "id") and hasattr(item, "dataset_id"): try: # Use sync API to avoid event loop issues when run_async_safely # creates multiple event loops across different threads dataset_run_item = await asyncio.to_thread( self.api.dataset_run_items.create, run_name=experiment_run_name, run_description=experiment_description, metadata=experiment_metadata, dataset_item_id=item.id, # type: ignore trace_id=trace_id, observation_id=span.id, dataset_version=dataset_version, ) dataset_run_id = dataset_run_item.dataset_run_id except Exception as e: langfuse_logger.error(f"Failed to create dataset run item: {e}") if ( not isinstance(item, dict) and hasattr(item, "dataset_id") and hasattr(item, "id") ): dataset_id = item.dataset_id dataset_item_id = item.id final_observation_metadata.update( {"dataset_id": dataset_id, "dataset_item_id": dataset_item_id} ) if isinstance(item_metadata, dict): final_observation_metadata.update(item_metadata) experiment_id = dataset_run_id or self._create_observation_id() experiment_item_id = ( dataset_item_id or get_sha256_hash_hex(_serialize(input_data))[:16] ) span._otel_span.set_attributes( { k: v for k, v in { LangfuseOtelSpanAttributes.ENVIRONMENT: LANGFUSE_SDK_EXPERIMENT_ENVIRONMENT, LangfuseOtelSpanAttributes.EXPERIMENT_DESCRIPTION: experiment_description, LangfuseOtelSpanAttributes.EXPERIMENT_ITEM_EXPECTED_OUTPUT: _serialize( expected_output ), }.items() if v is not None } ) propagated_experiment_attributes = PropagatedExperimentAttributes( experiment_id=experiment_id, experiment_name=experiment_run_name, experiment_metadata=_serialize(experiment_metadata), experiment_dataset_id=dataset_id, experiment_item_id=experiment_item_id, experiment_item_metadata=_serialize(item_metadata), experiment_item_root_observation_id=span.id, ) with _propagate_attributes(experiment=propagated_experiment_attributes): output = await _run_task(task, item) span.update( input=input_data, output=output, metadata=final_observation_metadata, ) except Exception as e: span.update( output=f"Error: {str(e)}", level="ERROR", status_message=str(e) ) raise e # Run evaluators evaluations = [] for evaluator in evaluators: try: eval_metadata: Optional[Dict[str, Any]] = None if isinstance(item, dict): eval_metadata = item.get("metadata") elif hasattr(item, "metadata"): eval_metadata = item.metadata with _propagate_attributes( experiment=propagated_experiment_attributes ): eval_results = await _run_evaluator( evaluator, input=input_data, output=output, expected_output=expected_output, metadata=eval_metadata, ) evaluations.extend(eval_results) # Store evaluations as scores for evaluation in eval_results: self.create_score( trace_id=trace_id, observation_id=span.id, name=evaluation.name, value=evaluation.value, # type: ignore comment=evaluation.comment, metadata=evaluation.metadata, config_id=evaluation.config_id, data_type=evaluation.data_type, # type: ignore ) except Exception as e: langfuse_logger.error(f"Evaluator failed: {e}") # Run composite evaluator if provided and we have evaluations if composite_evaluator and evaluations: try: composite_eval_metadata: Optional[Dict[str, Any]] = None if isinstance(item, dict): composite_eval_metadata = item.get("metadata") elif hasattr(item, "metadata"): composite_eval_metadata = item.metadata with _propagate_attributes( experiment=propagated_experiment_attributes ): result = composite_evaluator( input=input_data, output=output, expected_output=expected_output, metadata=composite_eval_metadata, evaluations=evaluations, ) # Handle async composite evaluators if asyncio.iscoroutine(result): result = await result # Normalize to list composite_evals: List[Evaluation] = [] if isinstance(result, (dict, Evaluation)): composite_evals = [result] # type: ignore elif isinstance(result, list): composite_evals = result # type: ignore # Store composite evaluations as scores and add to evaluations list for composite_evaluation in composite_evals: self.create_score( trace_id=trace_id, observation_id=span.id, name=composite_evaluation.name, value=composite_evaluation.value, # type: ignore comment=composite_evaluation.comment, metadata=composite_evaluation.metadata, config_id=composite_evaluation.config_id, data_type=composite_evaluation.data_type, # type: ignore ) evaluations.append(composite_evaluation) except Exception as e: langfuse_logger.error(f"Composite evaluator failed: {e}") return ExperimentItemResult( item=item, output=output, evaluations=evaluations, trace_id=trace_id, dataset_run_id=dataset_run_id, ) def _create_experiment_run_name( self, *, name: Optional[str] = None, run_name: Optional[str] = None ) -> str: if run_name: return run_name iso_timestamp = _get_timestamp().isoformat().replace("+00:00", "Z") return f"{name} - {iso_timestamp}" def run_batched_evaluation( self, *, scope: Literal["traces", "observations"], mapper: MapperFunction, filter: Optional[str] = None, fetch_batch_size: int = 50, fetch_trace_fields: Optional[str] = None, max_items: Optional[int] = None, max_retries: int = 3, evaluators: List[EvaluatorFunction], composite_evaluator: Optional[CompositeEvaluatorFunction] = None, max_concurrency: int = 5, metadata: Optional[Dict[str, Any]] = None, _add_observation_scores_to_trace: bool = False, _additional_trace_tags: Optional[List[str]] = None, resume_from: Optional[BatchEvaluationResumeToken] = None, verbose: bool = False, ) -> BatchEvaluationResult: """Fetch traces or observations and run evaluations on each item. This method provides a powerful way to evaluate existing data in Langfuse at scale. It fetches items based on filters, transforms them using a mapper function, runs evaluators on each item, and creates scores that are linked back to the original entities. This is ideal for: - Running evaluations on production traces after deployment - Backtesting new evaluation metrics on historical data - Batch scoring of observations for quality monitoring - Periodic evaluation runs on recent data The method uses a streaming/pipeline approach to process items in batches, making it memory-efficient for large datasets. It includes comprehensive error handling, retry logic, and resume capability for long-running evaluations. Args: scope: The type of items to evaluate. Must be one of: - "traces": Evaluate complete traces with all their observations - "observations": Evaluate individual observations (spans, generations, events) mapper: Function that transforms API response objects into evaluator inputs. Receives a trace/observation object and returns an EvaluatorInputs instance with input, output, expected_output, and metadata fields. Can be sync or async. evaluators: List of evaluation functions to run on each item. Each evaluator receives the mapped inputs and returns Evaluation object(s). Evaluator failures are logged but don't stop the batch evaluation. filter: Optional JSON filter string for querying items (same format as Langfuse API). Examples: - '{"tags": ["production"]}' - '{"user_id": "user123", "timestamp": {"operator": ">", "value": "2024-01-01"}}' Default: None (fetches all items). fetch_batch_size: Number of items to fetch per API call and hold in memory. Larger values may be faster but use more memory. Default: 50. fetch_trace_fields: Comma-separated list of fields to include when fetching traces. Available field groups: 'core' (always included), 'io' (input, output, metadata), 'scores', 'observations', 'metrics'. If not specified, all fields are returned. Example: 'core,scores,metrics'. Note: Excluded 'observations' or 'scores' fields return empty arrays; excluded 'metrics' returns -1 for 'totalCost' and 'latency'. Only relevant if scope is 'traces'. max_items: Maximum total number of items to process. If None, processes all items matching the filter. Useful for testing or limiting evaluation runs. Default: None (process all). max_concurrency: Maximum number of items to evaluate concurrently. Controls parallelism and resource usage. Default: 5. composite_evaluator: Optional function that creates a composite score from item-level evaluations. Receives the original item and its evaluations, returns a single Evaluation. Useful for weighted averages or combined metrics. Default: None. metadata: Optional metadata dict to add to all created scores. Useful for tracking evaluation runs, versions, or other context. Default: None. max_retries: Maximum number of retry attempts for failed batch fetches. Uses exponential backoff (1s, 2s, 4s). Default: 3. verbose: If True, logs progress information to console. Useful for monitoring long-running evaluations. Default: False. resume_from: Optional resume token from a previous incomplete run. Allows continuing evaluation after interruption or failure. Default: None. Returns: BatchEvaluationResult containing: - total_items_fetched: Number of items fetched from API - total_items_processed: Number of items successfully evaluated - total_items_failed: Number of items that failed evaluation - total_scores_created: Scores created by item-level evaluators - total_composite_scores_created: Scores created by composite evaluator - total_evaluations_failed: Individual evaluator failures - evaluator_stats: Per-evaluator statistics (success rate, scores created) - resume_token: Token for resuming if incomplete (None if completed) - completed: True if all items processed - duration_seconds: Total execution time - failed_item_ids: IDs of items that failed - error_summary: Error types and counts - has_more_items: True if max_items reached but more exist Raises: ValueError: If invalid scope is provided. Examples: Basic trace evaluation: ```python from langfuse import Langfuse, EvaluatorInputs, Evaluation client = Langfuse() # Define mapper to extract fields from traces def trace_mapper(trace): return EvaluatorInputs( input=trace.input, output=trace.output, expected_output=None, metadata={"trace_id": trace.id} ) # Define evaluator def length_evaluator(*, input, output, expected_output, metadata): return Evaluation( name="output_length", value=len(output) if output else 0 ) # Run batch evaluation result = client.run_batched_evaluation( scope="traces", mapper=trace_mapper, evaluators=[length_evaluator], filter='{"tags": ["production"]}', max_items=1000, verbose=True ) print(f"Processed {result.total_items_processed} traces") print(f"Created {result.total_scores_created} scores") ``` Evaluation with composite scorer: ```python def accuracy_evaluator(*, input, output, expected_output, metadata): # ... evaluation logic return Evaluation(name="accuracy", value=0.85) def relevance_evaluator(*, input, output, expected_output, metadata): # ... evaluation logic return Evaluation(name="relevance", value=0.92) def composite_evaluator(*, item, evaluations): # Weighted average of evaluations weights = {"accuracy": 0.6, "relevance": 0.4} total = sum( e.value * weights.get(e.name, 0) for e in evaluations if isinstance(e.value, (int, float)) ) return Evaluation( name="composite_score", value=total, comment=f"Weighted average of {len(evaluations)} metrics" ) result = client.run_batched_evaluation( scope="traces", mapper=trace_mapper, evaluators=[accuracy_evaluator, relevance_evaluator], composite_evaluator=composite_evaluator, filter='{"user_id": "important_user"}', verbose=True ) ``` Handling incomplete runs with resume: ```python # Initial run that may fail or timeout result = client.run_batched_evaluation( scope="observations", mapper=obs_mapper, evaluators=[my_evaluator], max_items=10000, verbose=True ) # Check if incomplete if not result.completed and result.resume_token: print(f"Processed {result.resume_token.items_processed} items before interruption") # Resume from where it left off result = client.run_batched_evaluation( scope="observations", mapper=obs_mapper, evaluators=[my_evaluator], resume_from=result.resume_token, verbose=True ) print(f"Total items processed: {result.total_items_processed}") ``` Monitoring evaluator performance: ```python result = client.run_batched_evaluation(...) for stats in result.evaluator_stats: success_rate = stats.successful_runs / stats.total_runs print(f"{stats.name}:") print(f" Success rate: {success_rate:.1%}") print(f" Scores created: {stats.total_scores_created}") if stats.failed_runs > 0: print(f" ⚠️ Failed {stats.failed_runs} times") ``` Note: - Evaluator failures are logged but don't stop the batch evaluation - Individual item failures are tracked but don't stop processing - Fetch failures are retried with exponential backoff - All scores are automatically flushed to Langfuse at the end - The resume mechanism uses timestamp-based filtering to avoid duplicates """ runner = BatchEvaluationRunner(self) return cast( BatchEvaluationResult, run_async_safely( runner.run_async( scope=scope, mapper=mapper, evaluators=evaluators, filter=filter, fetch_batch_size=fetch_batch_size, fetch_trace_fields=fetch_trace_fields, max_items=max_items, max_concurrency=max_concurrency, composite_evaluator=composite_evaluator, metadata=metadata, _add_observation_scores_to_trace=_add_observation_scores_to_trace, _additional_trace_tags=_additional_trace_tags, max_retries=max_retries, verbose=verbose, resume_from=resume_from, ) ), ) def auth_check(self) -> bool: """Check if the provided credentials (public and secret key) are valid. Raises: Exception: If no projects were found for the provided credentials. Note: This method is blocking. It is discouraged to use it in production code. """ try: projects = self.api.projects.get() langfuse_logger.debug( f"Auth check successful, found {len(projects.data)} projects" ) if len(projects.data) == 0: raise Exception( "Auth check failed, no project found for the keys provided." ) return True except AttributeError as e: langfuse_logger.warning( f"Auth check failed: Client not properly initialized. Error: {e}" ) return False except Error as e: handle_fern_exception(e) raise e def create_dataset( self, *, name: str, description: Optional[str] = None, metadata: Optional[Any] = None, input_schema: Optional[Any] = None, expected_output_schema: Optional[Any] = None, ) -> Dataset: """Create a dataset with the given name on Langfuse. Args: name: Name of the dataset to create. description: Description of the dataset. Defaults to None. metadata: Additional metadata. Defaults to None. input_schema: JSON Schema for validating dataset item inputs. When set, all new items will be validated against this schema. expected_output_schema: JSON Schema for validating dataset item expected outputs. When set, all new items will be validated against this schema. Returns: Dataset: The created dataset as returned by the Langfuse API. """ try: langfuse_logger.debug(f"Creating datasets {name}") result = self.api.datasets.create( name=name, description=description, metadata=metadata, input_schema=input_schema, expected_output_schema=expected_output_schema, ) return cast(Dataset, result) except Error as e: handle_fern_exception(e) raise e def create_dataset_item( self, *, dataset_name: str, input: Optional[Any] = None, expected_output: Optional[Any] = None, metadata: Optional[Any] = None, source_trace_id: Optional[str] = None, source_observation_id: Optional[str] = None, status: Optional[DatasetStatus] = None, id: Optional[str] = None, ) -> DatasetItem: """Create a dataset item. Upserts if an item with id already exists. Args: dataset_name: Name of the dataset in which the dataset item should be created. input: Input data. Defaults to None. Can contain any dict, list or scalar. expected_output: Expected output data. Defaults to None. Can contain any dict, list or scalar. metadata: Additional metadata. Defaults to None. Can contain any dict, list or scalar. source_trace_id: Id of the source trace. Defaults to None. source_observation_id: Id of the source observation. Defaults to None. status: Status of the dataset item. Defaults to ACTIVE for newly created items. id: Id of the dataset item. Defaults to None. Provide your own id if you want to dedupe dataset items. Id needs to be globally unique and cannot be reused across datasets. Returns: DatasetItem: The created dataset item as returned by the Langfuse API. Example: ```python from langfuse import Langfuse langfuse = Langfuse() # Uploading items to the Langfuse dataset named "capital_cities" langfuse.create_dataset_item( dataset_name="capital_cities", input={"input": {"country": "Italy"}}, expected_output={"expected_output": "Rome"}, metadata={"foo": "bar"} ) ``` """ try: langfuse_logger.debug(f"Creating dataset item for dataset {dataset_name}") result = self.api.dataset_items.create( dataset_name=dataset_name, input=input, expected_output=expected_output, metadata=metadata, source_trace_id=source_trace_id, source_observation_id=source_observation_id, status=status, id=id, ) return cast(DatasetItem, result) except Error as e: handle_fern_exception(e) raise e def resolve_media_references( self, *, obj: Any, resolve_with: Literal["base64_data_uri"], max_depth: int = 10, content_fetch_timeout_seconds: int = 5, ) -> Any: """Replace media reference strings in an object with base64 data URIs. This method recursively traverses an object (up to max_depth) looking for media reference strings in the format "@@@langfuseMedia:...@@@". When found, it (synchronously) fetches the actual media content using the provided Langfuse client and replaces the reference string with a base64 data URI. If fetching media content fails for a reference string, a warning is logged and the reference string is left unchanged. Args: obj: The object to process. Can be a primitive value, array, or nested object. If the object has a __dict__ attribute, a dict will be returned instead of the original object type. resolve_with: The representation of the media content to replace the media reference string with. Currently only "base64_data_uri" is supported. max_depth: int: The maximum depth to traverse the object. Default is 10. content_fetch_timeout_seconds: int: The timeout in seconds for fetching media content. Default is 5. Returns: A deep copy of the input object with all media references replaced with base64 data URIs where possible. If the input object has a __dict__ attribute, a dict will be returned instead of the original object type. Example: obj = { "image": "@@@langfuseMedia:type=image/jpeg|id=123|source=bytes@@@", "nested": { "pdf": "@@@langfuseMedia:type=application/pdf|id=456|source=bytes@@@" } } result = await LangfuseMedia.resolve_media_references(obj, langfuse_client) # Result: # { # "image": "data:image/jpeg;base64,/9j/4AAQSkZJRg...", # "nested": { # "pdf": "data:application/pdf;base64,JVBERi0xLjcK..." # } # } """ return LangfuseMedia.resolve_media_references( langfuse_client=self, obj=obj, resolve_with=resolve_with, max_depth=max_depth, content_fetch_timeout_seconds=content_fetch_timeout_seconds, ) @overload def get_prompt( self, name: str, *, version: Optional[int] = None, label: Optional[str] = None, type: Literal["chat"], cache_ttl_seconds: Optional[int] = None, fallback: Optional[List[ChatMessageDict]] = None, max_retries: Optional[int] = None, fetch_timeout_seconds: Optional[int] = None, ) -> ChatPromptClient: ... @overload def get_prompt( self, name: str, *, version: Optional[int] = None, label: Optional[str] = None, type: Literal["text"] = "text", cache_ttl_seconds: Optional[int] = None, fallback: Optional[str] = None, max_retries: Optional[int] = None, fetch_timeout_seconds: Optional[int] = None, ) -> TextPromptClient: ... def get_prompt( self, name: str, *, version: Optional[int] = None, label: Optional[str] = None, type: Literal["chat", "text"] = "text", cache_ttl_seconds: Optional[int] = None, fallback: Union[Optional[List[ChatMessageDict]], Optional[str]] = None, max_retries: Optional[int] = None, fetch_timeout_seconds: Optional[int] = None, ) -> PromptClient: """Get a prompt. This method attempts to fetch the requested prompt from the local cache. If the prompt is not found in the cache or if the cached prompt has expired, it will try to fetch the prompt from the server again and update the cache. If fetching the new prompt fails, and there is an expired prompt in the cache, it will return the expired prompt as a fallback. Args: name (str): The name of the prompt to retrieve. Keyword Args: version (Optional[int]): The version of the prompt to retrieve. If no label and version is specified, the `production` label is returned. Specify either version or label, not both. label: Optional[str]: The label of the prompt to retrieve. If no label and version is specified, the `production` label is returned. Specify either version or label, not both. cache_ttl_seconds: Optional[int]: Time-to-live in seconds for caching the prompt. Must be specified as a keyword argument. If not set, defaults to 60 seconds. Disables caching if set to 0. type: Literal["chat", "text"]: The type of the prompt to retrieve. Defaults to "text". fallback: Union[Optional[List[ChatMessageDict]], Optional[str]]: The prompt string to return if fetching the prompt fails. Important on the first call where no cached prompt is available. Follows Langfuse prompt formatting with double curly braces for variables. Defaults to None. max_retries: Optional[int]: The maximum number of retries in case of API/network errors. Defaults to 2. The maximum value is 4. Retries have an exponential backoff with a maximum delay of 10 seconds. fetch_timeout_seconds: Optional[int]: The timeout in milliseconds for fetching the prompt. Defaults to the default timeout set on the SDK, which is 5 seconds per default. Returns: The prompt object retrieved from the cache or directly fetched if not cached or expired of type - TextPromptClient, if type argument is 'text'. - ChatPromptClient, if type argument is 'chat'. Raises: Exception: Propagates any exceptions raised during the fetching of a new prompt, unless there is an expired prompt in the cache, in which case it logs a warning and returns the expired prompt. """ if self._resources is None: raise Error( "SDK is not correctly initialized. Check the init logs for more details." ) if version is not None and label is not None: raise ValueError("Cannot specify both version and label at the same time.") if not name: raise ValueError("Prompt name cannot be empty.") cache_key = PromptCache.generate_cache_key(name, version=version, label=label) bounded_max_retries = self._get_bounded_max_retries( max_retries, default_max_retries=2, max_retries_upper_bound=4 ) langfuse_logger.debug(f"Getting prompt '{cache_key}'") cached_prompt = self._resources.prompt_cache.get(cache_key) if cached_prompt is None or cache_ttl_seconds == 0: langfuse_logger.debug( f"Prompt '{cache_key}' not found in cache or caching disabled." ) try: return self._fetch_prompt_and_update_cache( name, version=version, label=label, ttl_seconds=cache_ttl_seconds, max_retries=bounded_max_retries, fetch_timeout_seconds=fetch_timeout_seconds, ) except Exception as e: if fallback: langfuse_logger.warning( f"Returning fallback prompt for '{cache_key}' due to fetch error: {e}" ) fallback_client_args: Dict[str, Any] = { "name": name, "prompt": fallback, "type": type, "version": version or 0, "config": {}, "labels": [label] if label else [], "tags": [], } if type == "text": return TextPromptClient( prompt=Prompt_Text(**fallback_client_args), is_fallback=True, ) if type == "chat": return ChatPromptClient( prompt=Prompt_Chat(**fallback_client_args), is_fallback=True, ) raise e if cached_prompt.is_expired(): langfuse_logger.debug(f"Stale prompt '{cache_key}' found in cache.") try: # refresh prompt in background thread, refresh_prompt deduplicates tasks langfuse_logger.debug(f"Refreshing prompt '{cache_key}' in background.") def refresh_task() -> None: self._fetch_prompt_and_update_cache( name, version=version, label=label, ttl_seconds=cache_ttl_seconds, max_retries=bounded_max_retries, fetch_timeout_seconds=fetch_timeout_seconds, ) self._resources.prompt_cache.add_refresh_prompt_task( cache_key, refresh_task, ) langfuse_logger.debug( f"Returning stale prompt '{cache_key}' from cache." ) # return stale prompt return cached_prompt.value except Exception as e: langfuse_logger.warning( f"Error when refreshing cached prompt '{cache_key}', returning cached version. Error: {e}" ) # creation of refresh prompt task failed, return stale prompt return cached_prompt.value return cached_prompt.value def _fetch_prompt_and_update_cache( self, name: str, *, version: Optional[int] = None, label: Optional[str] = None, ttl_seconds: Optional[int] = None, max_retries: int, fetch_timeout_seconds: Optional[int], ) -> PromptClient: cache_key = PromptCache.generate_cache_key(name, version=version, label=label) langfuse_logger.debug(f"Fetching prompt '{cache_key}' from server...") try: @backoff.on_exception( backoff.constant, Exception, max_tries=max_retries + 1, logger=None ) def fetch_prompts() -> Any: return self.api.prompts.get( self._url_encode(name), version=version, label=label, request_options={ "timeout_in_seconds": fetch_timeout_seconds, } if fetch_timeout_seconds is not None else None, ) prompt_response = fetch_prompts() prompt: PromptClient if prompt_response.type == "chat": prompt = ChatPromptClient(prompt_response) else: prompt = TextPromptClient(prompt_response) if self._resources is not None: self._resources.prompt_cache.set(cache_key, prompt, ttl_seconds) return prompt except NotFoundError as not_found_error: langfuse_logger.warning( f"Prompt '{cache_key}' not found during refresh, evicting from cache." ) if self._resources is not None: self._resources.prompt_cache.delete(cache_key) raise not_found_error except Exception as e: langfuse_logger.error( f"Error while fetching prompt '{cache_key}': {str(e)}" ) raise e def _get_bounded_max_retries( self, max_retries: Optional[int], *, default_max_retries: int = 2, max_retries_upper_bound: int = 4, ) -> int: if max_retries is None: return default_max_retries bounded_max_retries = min( max(max_retries, 0), max_retries_upper_bound, ) return bounded_max_retries @overload def create_prompt( self, *, name: str, prompt: List[Union[ChatMessageDict, ChatMessageWithPlaceholdersDict]], labels: List[str] = [], tags: Optional[List[str]] = None, type: Optional[Literal["chat"]], config: Optional[Any] = None, commit_message: Optional[str] = None, ) -> ChatPromptClient: ... @overload def create_prompt( self, *, name: str, prompt: str, labels: List[str] = [], tags: Optional[List[str]] = None, type: Optional[Literal["text"]] = "text", config: Optional[Any] = None, commit_message: Optional[str] = None, ) -> TextPromptClient: ... def create_prompt( self, *, name: str, prompt: Union[ str, List[Union[ChatMessageDict, ChatMessageWithPlaceholdersDict]] ], labels: List[str] = [], tags: Optional[List[str]] = None, type: Optional[Literal["chat", "text"]] = "text", config: Optional[Any] = None, commit_message: Optional[str] = None, ) -> PromptClient: """Create a new prompt in Langfuse. Keyword Args: name : The name of the prompt to be created. prompt : The content of the prompt to be created. is_active [DEPRECATED] : A flag indicating whether the prompt is active or not. This is deprecated and will be removed in a future release. Please use the 'production' label instead. labels: The labels of the prompt. Defaults to None. To create a default-served prompt, add the 'production' label. tags: The tags of the prompt. Defaults to None. Will be applied to all versions of the prompt. config: Additional structured data to be saved with the prompt. Defaults to None. type: The type of the prompt to be created. "chat" vs. "text". Defaults to "text". commit_message: Optional string describing the change. Returns: TextPromptClient: The prompt if type argument is 'text'. ChatPromptClient: The prompt if type argument is 'chat'. """ try: langfuse_logger.debug(f"Creating prompt {name=}, {labels=}") if type == "chat": if not isinstance(prompt, list): raise ValueError( "For 'chat' type, 'prompt' must be a list of chat messages with role and content attributes." ) request: Union[CreateChatPromptRequest, CreateTextPromptRequest] = ( CreateChatPromptRequest( name=name, prompt=cast(Any, prompt), labels=labels, tags=tags, config=config or {}, commit_message=commit_message, type=CreateChatPromptType.CHAT, ) ) server_prompt = self.api.prompts.create(request=request) if self._resources is not None: self._resources.prompt_cache.invalidate(name) return ChatPromptClient(prompt=cast(Prompt_Chat, server_prompt)) if not isinstance(prompt, str): raise ValueError("For 'text' type, 'prompt' must be a string.") request = CreateTextPromptRequest( name=name, prompt=prompt, labels=labels, tags=tags, config=config or {}, commit_message=commit_message, ) server_prompt = self.api.prompts.create(request=request) if self._resources is not None: self._resources.prompt_cache.invalidate(name) return TextPromptClient(prompt=cast(Prompt_Text, server_prompt)) except Error as e: handle_fern_exception(e) raise e def update_prompt( self, *, name: str, version: int, new_labels: List[str] = [], ) -> Any: """Update an existing prompt version in Langfuse. The Langfuse SDK prompt cache is invalidated for all prompts witht he specified name. Args: name (str): The name of the prompt to update. version (int): The version number of the prompt to update. new_labels (List[str], optional): New labels to assign to the prompt version. Labels are unique across versions. The "latest" label is reserved and managed by Langfuse. Defaults to []. Returns: Prompt: The updated prompt from the Langfuse API. """ updated_prompt = self.api.prompt_version.update( name=self._url_encode(name), version=version, new_labels=new_labels, ) if self._resources is not None: self._resources.prompt_cache.invalidate(name) return updated_prompt def _url_encode(self, url: str, *, is_url_param: Optional[bool] = False) -> str: # httpx ≥ 0.28 does its own WHATWG-compliant quoting (eg. encodes bare # “%”, “?”, “#”, “|”, … in query/path parts). Re-quoting here would # double-encode, so we skip when the value is about to be sent straight # to httpx (`is_url_param=True`) and the installed version is ≥ 0.28. if is_url_param and Version(httpx.__version__) >= Version("0.28.0"): return url # urllib.parse.quote does not escape slashes "/" by default; we need to add safe="" to force escaping # we need add safe="" to force escaping of slashes # This is necessary for prompts in prompt folders return urllib.parse.quote(url, safe="") def clear_prompt_cache(self) -> None: """Clear the entire prompt cache, removing all cached prompts. This method is useful when you want to force a complete refresh of all cached prompts, for example after major updates or when you need to ensure the latest versions are fetched from the server. """ if self._resources is not None: self._resources.prompt_cache.clear()