LlmClient Control Constants  
 
Constant Value Description
Providers
llmProviderNone0No provider specified
llmProviderLocal1Local inference provider (e.g., LM Studio or Ollama)
llmProviderOpenAI2OpenAI provider using the ChatGPT models
llmProviderAnthropic3Anthropic provider using the Claude models
llmProviderGoogle4Google AI provider using the Gemini models
llmProviderMicrosoft5Microsoft Foundry provider using OpenAI-compatible models
llmProviderMistral6Mistral AI provider using the Mistral models
llmProviderUnknown-1Unknown or unrecognized service provider
llmProviderDefaultllmProviderLocalDefault service provider
Options
llmOptionNone0No options specified
llmOptionNoCache&H0001Disable provider and model caching for the session
llmOptionKeepAlive&H0002Enable persistent (keep-alive) HTTP connections
llmOptionProxy&H0008Use a configured proxy server for the connection
llmOptionNoUserAgent&H0020Do not send a User-Agent header with the request
llmOptionSecure&H1000Force use of a secure (HTTPS) connection
llmOptionDefaultllmOptionNoneDefault option set (no flags enabled)
Provider Flags
llmProviderFlagNone0No provider requirements specified
llmProviderFlagRequireTls&H0001Requires a secure TLS (HTTPS) connection
llmProviderFlagRequireAuth&H0002Requires authentication (e.g., API key or token)
llmProviderFlagApiVersion&H0004Requires an explicit API version parameter
llmProviderFlagLocal&H0008Locally hosted provider (LM Studio, Ollama, etc.)
llmProviderFlagModelList&H0010Supports enumerating available models
llmProviderFlagModelInfo&H0020Supports retrieving detailed model information
Model Options
llmModelOptionDefault&H0000Default model filtering (no additional criteria)
llmModelOptionPreview&H0001Include preview or experimental models
llmModelOptionStreaming&H0002Include models that support streaming responses
llmModelOptionStructured&H0004Include models that support structured outputs
llmModelOptionReasoning&H0008Include reasoning-oriented models
llmModelOptionVision&H0010Include models that support image input
llmModelOptionAudio&H0200Include models that support audio output (text-to-speech)
llmModelOptionImage&H0400Include models that support image generation or input
llmModelOptionVideo&H0800Include models that support video generation
llmModelOptionTranscription&H1000Include models that support speech-to-text transcription
llmModelOptionDeprecated&H8000Include deprecated or retired models
llmModelOptionAllModels&HFFFFInclude all available model types
Model Flags
llmModelFlagNone&H0000No model capabilities defined
llmModelFlagChat&H0001Supports conversational text generation
llmModelFlagSystemRole&H0002Accepts a system role message in the input
llmModelFlagStreaming&H0004Supports streaming responses
llmModelFlagStructured&H0008Supports structured output (e.g., JSON or schema)
llmModelFlagVision&H0010Supports image inputs (multimodal)
llmModelFlagTools&H0020Supports tool or function calling
llmModelFlagReasoning&H0040Supports reasoning-oriented generation
llmModelFlagResponses&H0080Supports the Responses API (unified endpoint)
llmModelFlagFunctions&H0100Supports OpenAI-style function calling with schemas
llmModelFlagAudio&H0200Supports audio output (text-to-speech)
llmModelFlagImage&H0400Supports image generation
llmModelFlagVideo&H0800Supports video generation
llmModelFlagTranscription&H1000Supports speech-to-text transcription
llmModelFlagInstructions&H2000Supports a dedicated instruction field outside message lists
llmModelFlagMaxCompletionTokens&H4000Uses max_completion_tokens instead of max_tokens in requests
llmModelFlagInferred&H80000Capabilities are inferred rather than explicitly defined
Responses
llmResponseUnknown0No stop reason specified or unrecognized response state
llmResponseComplete1Response completed normally
llmResponseLimited2Stopped due to token or context limits
llmResponseAborted3Request was cancelled or interrupted
llmResponseError4An error occurred during processing
llmResponseTimeout5Request timed out
Reasoning
llmReasoningDefault0Provider or model default
llmReasoningNone1Disable reasoning if available
llmReasoningLow2Prefer faster responses; reduces usage costs
llmReasoningMedium3Balanced reasoning
llmReasoningHigh4Prefer quality responses; increases usage costs
llmReasoningMaximum5Maximum reasoning quality
Temperature
llmTemperatureMinimum0.0Deterministic output (minimal randomness)
llmTemperatureConservative0.3Low variability, more predictable responses
llmTemperatureBalanced0.7Balanced creativity and coherence
llmTemperatureDefault1.0Standard default sampling behavior
llmTemperatureCreative1.2Increased creativity and variation
llmTemperatureMaximum2.0Maximum randomness (highly varied output)
Top-P Sampling
llmTopPDisabled0.0Nucleus sampling disabled
llmTopPMinimum0.1Very restrictive token selection
llmTopPConservative0.8Conservative token selection
llmTopPBalanced0.9Balanced nucleus sampling
llmTopPCreative0.95More diverse token selection
llmTopPMaximum1.0Full probability distribution (no filtering)
llmTopPDefaultllmTopPDisabledDefault disables nucleus sampling
Top-K Sampling
llmTopKDisabled0Top-K sampling disabled
llmTopKMinimum0Equivalent to disabled sampling
llmTopKConservative10Restrict to top 10 tokens
llmTopKBalanced40Balanced token selection range
llmTopKCreative60More diverse token selection
llmTopKMaximum100Wide token selection range
llmTopKDefaultllmTopKDisabledDefault disables Top-K sampling
Frequency
llmFrequencyDisabled0.0No frequency penalty applied
llmFrequencyMinimum0.0Minimum frequency penalty
llmFrequencyMaximum2.0Maximum frequency penalty
llmFrequencyDefaultllmFrequencyDisabledDefault disables frequency penalty
Presence
llmPresenceDisabled0.0No presence penalty applied
llmPresenceMinimum0.0Minimum presence penalty
llmPresenceMaximum2.0Maximum presence penalty
llmPresenceDefaultllmPresenceDisabledDefault disables presence penalty
Limits
LlmMaxMessageLength65536Maximum length of a single input message (64 KB)
LlmMaxResponseLength262144Maximum expected response length (256 KB)
LlmMaxSystemPromptLength32768Maximum length of the system prompt (32 KB)
LlmMaxChatHistory1000Maximum number of messages stored per session
LlmMaxTokens200000Maximum token limit per request