Trustwise SDK

Contents

  • Installation
    • Installing the SDK
    • Requirements
  • Usage
    • Basic Setup
    • Evaluating Metrics
      • Faithfulness Metric
    • Async Usage
    • Working with JSON Output
    • Working with Python Dict Output
    • Working with Result Properties
    • Guardrails (Experimental)
  • V4 Metrics
    • Adherence
      • metrics.adherence.evaluate()
    • Answer Relevancy
      • metrics.answer_relevancy.evaluate()
    • Clarity
      • metrics.clarity.evaluate()
    • Completion
      • metrics.completion.evaluate()
    • Context Relevancy
      • metrics.context_relevancy.evaluate()
    • Faithfulness
      • metrics.faithfulness.evaluate()
    • Formality
      • metrics.formality.evaluate()
    • Helpfulness
      • metrics.helpfulness.evaluate()
    • PII Detection
      • metrics.pii.evaluate()
    • Prompt Manipulation
      • metrics.prompt_manipulation.evaluate()
    • Refusal
      • metrics.refusal.evaluate()
    • Sensitivity
      • metrics.sensitivity.evaluate()
    • Stability
      • metrics.stability.evaluate()
    • Simplicity
      • metrics.simplicity.evaluate()
    • Tone
      • metrics.tone.evaluate()
    • Toxicity
      • metrics.toxicity.evaluate()
    • Carbon
      • metrics.carbon.evaluate()
  • V3 Metrics
    • Cost
      • metrics.v3.cost.evaluate()
  • API Reference
    • SDK
      • TrustwiseSDK
        • TrustwiseSDK.__init__()
        • TrustwiseSDK.get_beta_features()
        • TrustwiseSDK.get_versions()
        • TrustwiseSDK.guardrails()
        • TrustwiseSDK.is_beta_feature()
      • TrustwiseSDKAsync
        • TrustwiseSDKAsync.__init__()
    • Configuration
      • TrustwiseConfig
        • TrustwiseConfig.api_key
        • TrustwiseConfig.__init__()
        • TrustwiseConfig.base_url
        • TrustwiseConfig._validate_url()
        • TrustwiseConfig.get_safety_url()
        • TrustwiseConfig.get_alignment_url()
        • TrustwiseConfig.get_performance_url()
        • TrustwiseConfig.get_metrics_url()
    • Types
      • Score
        • Score.value
        • Score.__float__()
        • Score.__str__()
        • Score.__repr__()
      • convert_to_score()
      • ObjectStyleScore
        • ObjectStyleScore.label
        • ObjectStyleScore.score
      • ContextChunk
        • ContextChunk.chunk_text
        • ContextChunk.chunk_id
      • ContextRelevancyRequest
        • ContextRelevancyRequest.query
        • ContextRelevancyRequest.context
        • ContextRelevancyRequest.severity
        • ContextRelevancyRequest.include_chunk_scores
        • ContextRelevancyRequest.metadata
      • PromptManipulationRequest
        • PromptManipulationRequest.text
        • PromptManipulationRequest.severity
      • PromptManipulationResponse
        • PromptManipulationResponse.score
        • PromptManipulationResponse.scores
      • AnswerRelevancyRequest
        • AnswerRelevancyRequest.query
        • AnswerRelevancyRequest.response
      • AnswerRelevancyResponse
        • AnswerRelevancyResponse.score
        • AnswerRelevancyResponse.generated_question
      • FaithfulnessRequest
        • FaithfulnessRequest.query
        • FaithfulnessRequest.response
        • FaithfulnessRequest.context
        • FaithfulnessRequest.severity
        • FaithfulnessRequest.include_citations
      • Statement
        • Statement.statement
        • Statement.label
        • Statement.probability
        • Statement.sentence_span
      • FaithfulnessResponse
        • FaithfulnessResponse.score
        • FaithfulnessResponse.statements
      • ContextRelevancyResponse
        • ContextRelevancyResponse.score
        • ContextRelevancyResponse.scores
      • ClarityRequest
        • ClarityRequest.text
      • ClarityResponse
        • ClarityResponse.score
      • FormalityRequest
        • FormalityRequest.text
      • FormalityResponse
        • FormalityResponse.score
      • SimplicityRequest
        • SimplicityRequest.text
      • SimplicityResponse
        • SimplicityResponse.score
      • ToneRequest
        • ToneRequest.text
        • ToneRequest.tones
      • ToneResponse
        • ToneResponse.scores
      • ToxicityRequest
        • ToxicityRequest.text
        • ToxicityRequest.severity
      • ToxicityResponse
        • ToxicityResponse.score
        • ToxicityResponse.scores
      • PIIEntity
        • PIIEntity.interval
        • PIIEntity.string
        • PIIEntity.category
      • PIIRequest
        • PIIRequest.text
        • PIIRequest.allowlist
        • PIIRequest.blocklist
        • PIIRequest.categories
      • PIIResponse
        • PIIResponse.pii
      • HelpfulnessRequest
        • HelpfulnessRequest.text
      • HelpfulnessResponse
        • HelpfulnessResponse.score
      • SensitivityRequest
        • SensitivityRequest.text
        • SensitivityRequest.topics
      • SensitivityResponse
        • SensitivityResponse.scores
      • RefusalRequest
        • RefusalRequest.query
        • RefusalRequest.response
      • RefusalResponse
        • RefusalResponse.score
      • CompletionRequest
        • CompletionRequest.query
        • CompletionRequest.response
      • CompletionResponse
        • CompletionResponse.score
      • AdherenceRequest
        • AdherenceRequest.policy
        • AdherenceRequest.response
      • AdherenceResponse
        • AdherenceResponse.score
      • StabilityRequest
        • StabilityRequest.responses
      • StabilityResponse
        • StabilityResponse.min
        • StabilityResponse.avg
      • CarbonValue
        • CarbonValue.value
        • CarbonValue.unit
      • CarbonComponent
        • CarbonComponent.component
        • CarbonComponent.carbon
      • CarbonRequest
        • CarbonRequest.provider
        • CarbonRequest.region
        • CarbonRequest.instance_type
        • CarbonRequest.latency
      • CarbonResponse
        • CarbonResponse.carbon
        • CarbonResponse.components
    • V3 Types (Legacy)
      • SDKBaseModel
        • SDKBaseModel.to_json()
        • SDKBaseModel.to_dict()
        • SDKBaseModel.validate_score_range()
        • SDKBaseModel.format_validation_error()
        • SDKBaseModel._get_field_description()
        • SDKBaseModel._format_output()
        • SDKBaseModel._format_nested_dict()
        • SDKBaseModel.__repr__()
        • SDKBaseModel._format_json_value()
        • SDKBaseModel._repr_mimebundle_()
        • SDKBaseModel._format_html()
        • SDKBaseModel._get_value_css_class()
        • SDKBaseModel._format_dict_html()
        • SDKBaseModel.__repr_html__()
      • SDKRequestModel
        • SDKRequestModel.metadata
      • SDKResponseModel
      • CostResponseV3
        • CostResponseV3.cost_estimate_per_run
        • CostResponseV3.total_project_cost_estimate
      • CostRequestV3
        • CostRequestV3.model_name
        • CostRequestV3.model_type
        • CostRequestV3.model_provider
        • CostRequestV3.number_of_queries
        • CostRequestV3.total_prompt_tokens
        • CostRequestV3.total_completion_tokens
        • CostRequestV3.total_tokens
        • CostRequestV3.instance_type
        • CostRequestV3.average_latency
      • GuardrailResponse
        • GuardrailResponse.passed
        • GuardrailResponse.blocked
        • GuardrailResponse.results
        • GuardrailResponse.to_json()
        • GuardrailResponse.to_dict()
    • Guardrails
      • Guardrail
        • Guardrail.__init__()
        • Guardrail._get_metric_kwargs()
        • Guardrail.check_pii()
        • Guardrail.evaluate()
  • API Versioning
    • Version Management
      • Available Versions
      • Current Version
    • Usage Patterns
      • Explicit Version Path
      • Default Version
    • Supported Versions
  • Migration Guide: v3 to v4
    • Prerequisites
    • Migration Steps
    • Step 1: Update Context Format
    • Step 2: Update Metric Calls
    • Step 3: Handle Metric-Specific Changes
      • Prompt Injection → Prompt Manipulation
      • New Metrics in v4
      • Removed Metrics in v4
      • Parameter Name Changes
    • Step 4: Handle Response Schema Changes
      • Response Schema Changes Summary
        • Faithfulness Response
        • Prompt Manipulation Response
        • Toxicity Response
        • PII Response
        • Tone Response
        • Formality Response
        • Context Relevancy Response
    • Complete Migration Example
    • Async Migration
    • Support
  • Changelog
    • v1.0.0b10 (10-07-2025)
      • Deprecation Notices
      • Features
      • Migration
    • v1.0.0 (06-20-2025)
      • Features
      • Metrics
      • Guardrails System (Experimental)
      • Version Management
      • Configuration
Trustwise SDK
  • Search

© Copyright 2025, Trustwise Team.

Built with Sphinx using a theme provided by Read the Docs.