Re-exports§
pub use drifter::scouter::PyDrifter;pub use profiler::scouter::DataProfiler;pub use crate::http::PyScouterClient;pub use crate::http::ScouterClient;pub use error::ClientError;
Modules§
Structs§
- Active
Span - ActiveSpan where all the magic happens The active Span attempts to maintain compatibility with the OpenTelemetry Span API
- Alert
- Alerts
- Attribute
- Base
Tracer - The main Tracer class
- Batch
Config - Bin
- Binned
Metric - Binned
Metric Stats - Binned
Metrics - Binned
PsiFeature Metrics - Binned
PsiMetric - Char
Stats - Console
Dispatch Config - Custom
Drift Profile - Custom
Interval - Custom
Metric - Custom
Metric Alert Condition - Custom
Metric Alert Config - Custom
Metric Drift Config - Custom
Metric Feature Queue - Custom
Metric Server Record - Data
Profile - Distinct
- Doane
- Drift
Alert Request - Drift
Request - Equal
Width Binning - Evaluation
Config - Export
Config - Feature
Map - Feature
Profile - Features
- Freedman
Diaconis - GetProfile
Request - Grpc
Config - Grpc
Span Exporter - Histogram
- Python class for a feature histogram
- Http
Config - Http
Span Exporter - Kafka
Config - LLMAlert
Config - LLMDrift
Config - LLMDrift
Map - LLMDrift
Metric - LLMDrift
Profile - LLMDrift
Record Pagination Request - LLMDrift
Server Record - LLMEval
Metric - LLMEval
Record - LLMEval
Results - Enhanced results collection that captures both successes and failures
- LLMEval
Task Result - Struct for collecting results from LLM evaluation tasks.
- LLMMetric
Alert Condition - LLMMetric
Record - LLMRecord
- LLMRecord
Queue - Latency
Metrics - Manual
- Metric
- Metrics
- Mock
Config - NumProfiler
- Numeric
Stats - Observability
Metrics - Observer
- OpsGenie
Dispatch Config - Otel
Http Config - Pagination
Cursor - Pagination
Response - Profile
Request - Profile
Status Request - PsiAlert
Config - PsiChi
Square Threshold - PsiDrift
Config - PsiDrift
Map - PsiDrift
Profile - PsiFeature
Drift Profile - PsiFeature
Queue - PsiFixed
Threshold - PsiMonitor
- PsiNormal
Threshold - PsiServer
Record - Quantile
Binning - Quantiles
- Python class for quantiles
- Queue
Bus - QueueBus is an mpsc bus that allows for publishing events to subscribers. It leverage an unbounded channel Primary way to publish non-blocking events to background queues with ScouterQueue
- RabbitMQ
Config - Redis
Config - Registered
Profile Response - Rice
- Route
Metrics - Scott
- Scouter
Queue - Scouter
Response - Scouter
Server Error - Common struct for returning errors from scouter server (axum response)
- Server
Records - Slack
Dispatch Config - Span
Event - Span
Link - SpcAlert
- SpcAlert
Config - SpcAlert
Rule - SpcDrift
Config - Python class for a monitoring configuration
- SpcDrift
Feature - SpcDrift
Features - SpcDrift
Map - Python class for a Drift map of features with calculated drift
- SpcDrift
Profile - SpcFeature
Alert - SpcFeature
Alerts - SpcFeature
Drift - Python class for a feature drift
- SpcFeature
Drift Profile - Python class for a monitoring profile
- SpcFeature
Queue - SpcMonitor
- SpcServer
Record - Square
Root - Stdout
Span Exporter - String
Profiler - String
Stats - Sturges
- TagRecord
- Tags
Response - Task
State - Terrell
Scott - Test
Span Exporter - Trace
Baggage Record - Trace
Baggage Response - Trace
Filters - Trace
List Item - Trace
Metric Bucket - Trace
Metrics Request - Trace
Metrics Response - Trace
Pagination Response - Trace
Record - Trace
Span - Trace
Span Record - Trace
Spans Response - Update
Alert Response - Update
Alert Status - Version
Request - Word
Stats
Enums§
- Alert
Dispatch Type - Alert
Threshold - Alert
Zone - Common
Crons - Compression
Type - Contract
Error - Data
Profile Error - Data
Type - Drift
Error - Drift
Profile - Drift
Type - Entity
Type - Evaluation
Error - Event
Error - Feature
- Function
Type - Otel
Protocol - Profile
Error - PyEvent
Error - Record
Error - Record
Type - Server
Record - Span
Kind - SpcAlert
Type - Time
Interval - Type
Error - Util
Error
Traits§
Functions§
- async_
evaluate_ llm - Main orchestration function that decides which execution path to take
- compute_
feature_ correlations - create_
feature_ map - evaluate_
llm - Function for evaluating LLM response and generating metrics. The primary use case for evaluate_llm is to take a list of data samples, which often contain inputs and outputs from LLM systems and evaluate them against user-defined metrics in a LLM as a judge pipeline. The user is expected provide a list of dict objects and a list of LLMEval metrics. These eval metrics will be used to create a workflow, which is then executed in an async context. All eval scores are extracted and returned to the user.
- flush_
tracer - Helper function to force flush the tracer provider
- generate_
alerts - Generate alerts for each feature in the drift array
- get_
function_ type - Function to determine if a Python function is async, async generator, or generator This is a helper util function used in tracing decorators
- init_
tracer - Global initialization function for the tracer. This sets up the tracer provider with the specified service name, endpoint, and sampling ratio. If no endpoint is provided, spans will be exported to stdout for debugging purposes.
- shutdown_
tracer - workflow_
from_ eval_ metrics - Builds a workflow from a list of LLMEvalMetric objects