| Constant |
Value |
Description |
| Ports |
| LLM_PORT_DEFAULT | 0 | Use the provider-specific default port. |
| LLM_PORT_SECURE | 443 | Default port for secure HTTPS connections. |
| Protocols |
| LLM_PROTOCOL_DEFAULT | 0 | Use the provider-specific default protocol. |
| LLM_PROTOCOL_HTTP | 1 | Use the HTTP protocol. |
| LLM_PROTOCOL_HTTPS | 2 | Use the HTTPS protocol. |
| LLM_PROTOCOL_UNKNOWN | -1 | Unknown or invalid protocol. |
| Options |
| LLM_OPTION_NONE | 0 | No options specified. |
| LLM_OPTION_NOCACHE | 0x000001 | Disable provider and model caching for the session. |
| LLM_OPTION_KEEPALIVE | 0x000002 | Enable persistent (keep-alive) HTTP connections. |
| LLM_OPTION_PROXY | 0x000008 | Use a configured proxy server for the connection. |
| LLM_OPTION_NOUSERAGENT | 0x000020 | Do not send a User-Agent header with the request. |
| LLM_OPTION_SECURE | 0x001000 | Force use of a secure (HTTPS) connection. |
| LLM_OPTION_DEFAULT | LLM_OPTION_NONE | Default option set (no flags enabled). |
| Providers |
| LLM_PROVIDER_NONE | 0 | No provider specified. |
| LLM_PROVIDER_LOCAL | 1 | Local inference provider (e.g., LM Studio or Ollama). |
| LLM_PROVIDER_OPENAI | 2 | OpenAI provider using the ChatGPT models. |
| LLM_PROVIDER_ANTHROPIC | 3 | Anthropic provider using the Claude models. |
| LLM_PROVIDER_GOOGLE | 4 | Google AI provider using the Gemini models. |
| LLM_PROVIDER_MICROSOFT | 5 | Microsoft Foundry provider using OpenAI-compatible models. |
| LLM_PROVIDER_MISTRAL | 6 | Mistral AI provider using the Mistral models. |
| LLM_PROVIDER_UNKNOWN | ((DWORD)-1) | Unknown or unrecognized service provider. |
| LLM_PROVIDER_DEFAULT | LLM_PROVIDER_LOCAL | Default service provider. |
| Reasoning |
| LLM_REASONING_NONE | 0 | No reasoning, even if it is available. |
| LLM_REASONING_DEFAULT | 1 | Provider or model default. |
| LLM_REASONING_LOW | 2 | Prefer faster responses; reduces usage costs. |
| LLM_REASONING_MEDIUM | 3 | Balanced reasoning. |
| LLM_REASONING_HIGH | 4 | Prefer quality responses; increases usage costs. |
| LLM_REASONING_MAXIMUM | 5 | Maximum reasoning quality. |
| Temperature |
| LLM_TEMPERATURE_MINIMUM | 0.0 | Deterministic output
with minimal randomness. |
| LLM_TEMPERATURE_CONSERVATIVE | 0.3 | Low variability, more predictable responses. |
| LLM_TEMPERATURE_BALANCED | 0.7 | Balanced creativity and coherence. |
| LLM_TEMPERATURE_DEFAULT | 1.0 | Standard default sampling behavior. |
| LLM_TEMPERATURE_CREATIVE | 1.2 | Increased creativity and variation. |
| LLM_TEMPERATURE_MAXIMUM | 2.0 | Maximum randomness
with highly varied output. |
| Top-P Sampling |
| LLM_TOP_P_DISABLED | 0.0 | Nucleus sampling disabled. |
| LLM_TOP_P_MINIMUM | 0.1 | Very restrictive token selection. |
| LLM_TOP_P_CONSERVATIVE | 0.8 | Conservative token selection. |
| LLM_TOP_P_BALANCED | 0.9 | Balanced nucleus sampling. |
| LLM_TOP_P_CREATIVE | 0.95 | More diverse token selection. |
| LLM_TOP_P_MAXIMUM | 1.0 | Full probability distribution
with no filtering. |
| LLM_TOP_P_DEFAULT | LLM_TOP_P_DISABLED | Default disables nucleus sampling. |
| Top-K Sampling |
| LLM_TOP_K_DISABLED | 0 | Top-K sampling disabled. |
| LLM_TOP_K_MINIMUM | 0 | Equivalent to disabled sampling. |
| LLM_TOP_K_CONSERVATIVE | 10 | Restrict to top 10 tokens. |
| LLM_TOP_K_BALANCED | 40 | Balanced token selection range. |
| LLM_TOP_K_CREATIVE | 60 | More diverse token selection. |
| LLM_TOP_K_MAXIMUM | 100 | Wide token selection range. |
| LLM_TOP_K_DEFAULT | LLM_TOP_K_DISABLED | Default disables Top-K sampling. |
| Frequency |
| LLM_FREQUENCY_DISABLED | 0.0 | No frequency penalty applied. |
| LLM_FREQUENCY_MINIMUM | 0.0 | Minimum frequency penalty. |
| LLM_FREQUENCY_MAXIMUM | 2.0 | Maximum frequency penalty. |
| LLM_FREQUENCY_DEFAULT | LLM_FREQUENCY_DISABLED | Default disables frequency penalty. |
| Presence |
| LLM_PRESENCE_DISABLED | 0.0 | No presence penalty applied. |
| LLM_PRESENCE_MINIMUM | 0.0 | Minimum presence penalty. |
| LLM_PRESENCE_MAXIMUM | 2.0 | Maximum presence penalty. |
| LLM_PRESENCE_DEFAULT | LLM_PRESENCE_DISABLED | Default disables presence penalty. |
| Provider Flags |
| LLM_PROVIDER_FLAG_NONE | 0 | No provider requirements specified. |
| LLM_PROVIDER_FLAG_REQUIRE_TLS | 0x0001 | Requires a secure TLS (HTTPS) connection. |
| LLM_PROVIDER_FLAG_REQUIRE_AUTH | 0x0002 | Requires authentication (e.g., API key or token). |
| LLM_PROVIDER_FLAG_API_VERSION | 0x0004 | Requires an explicit API version parameter. |
| LLM_PROVIDER_FLAG_LOCAL | 0x0008 | Locally hosted provider (LM Studio, Ollama, etc.) |
| LLM_PROVIDER_FLAG_MODEL_LIST | 0x0010 | Supports enumerating available models. |
| LLM_PROVIDER_FLAG_MODEL_INFO | 0x0020 | Supports retrieving detailed model information. |
| Model Options |
| LLM_MODEL_OPTION_DEFAULT | 0x0000 | default list
of chat models for a provider. |
| LLM_MODEL_OPTION_PREVIEW | 0x0001 | Include preview or experimental models. |
| LLM_MODEL_OPTION_STREAMING | 0x0002 | Include models that support streaming responses. |
| LLM_MODEL_OPTION_STRUCTURED | 0x0004 | Include models that support structured outputs. |
| LLM_MODEL_OPTION_REASONING | 0x0008 | Include reasoning-oriented models. |
| LLM_MODEL_OPTION_VISION | 0x0010 | Include models that support image input. |
| LLM_MODEL_OPTION_AUDIO | 0x0200 | Include models that support audio output (text-to-speech). |
| LLM_MODEL_OPTION_IMAGE | 0x0400 | Include models that support image generation or input. |
| LLM_MODEL_OPTION_VIDEO | 0x0800 | Include models that support video generation. |
| LLM_MODEL_OPTION_TRANSCRIPTION | 0x1000 | Include models that support speech-to-text transcription. |
| LLM_MODEL_OPTION_DEPRECATED | 0x8000 | Include deprecated or retired models. |
| LLM_MODEL_OPTION_ALL_MODELS | 0xFFFF | Include all available model types. |
| Model Flags |
| LLM_MODEL_FLAG_NONE | 0x0000 | No model capabilities defined. |
| LLM_MODEL_FLAG_CHAT | 0x0001 | Supports conversational text generation. |
| LLM_MODEL_FLAG_SYSTEM_ROLE | 0x0002 | Accepts a system role message in the input. |
| LLM_MODEL_FLAG_STREAMING | 0x0004 | Supports streaming responses. |
| LLM_MODEL_FLAG_STRUCTURED | 0x0008 | Supports structured output (e.g., JSON or schema). |
| LLM_MODEL_FLAG_VISION | 0x0010 | Supports image inputs (multimodal). |
| LLM_MODEL_FLAG_TOOLS | 0x0020 | Supports tool or function calling. |
| LLM_MODEL_FLAG_REASONING | 0x0040 | Supports reasoning-oriented generation. |
| LLM_MODEL_FLAG_RESPONSES | 0x0080 | Supports the Responses API (unified endpoint). |
| LLM_MODEL_FLAG_FUNCTIONS | 0x0100 | Supports OpenAI-style function calling with schemas. |
| LLM_MODEL_FLAG_AUDIO | 0x0200 | Supports audio output (text-to-speech) |
| LLM_MODEL_FLAG_IMAGE | 0x0400 | Supports image generation. |
| LLM_MODEL_FLAG_VIDEO | 0x0800 | Supports video generation. |
| LLM_MODEL_FLAG_TRANSCRIPTION | 0x1000 | Supports speech-to-text transcription. |
| LLM_MODEL_FLAG_INSTRUCTIONS | 0x2000 | Supports a dedicated instruction field outside message lists. |
| LLM_MODEL_FLAG_MAX_COMPLETION_TOKENS | 0x4000 | Uses max_completion_tokens instead of max_tokens in requests. |
| LLM_MODEL_FLAG_INFERRED | 0x80000 | Capabilities are inferred rather than explicitly defined. |
| Model Status |
| LLM_MODEL_STATUS_NONE | 0x0000 | No status assigned. |
| LLM_MODEL_STATUS_AVAILABLE | 0x0001 | Model is actively available. |
| LLM_MODEL_STATUS_PREVIEW | 0x0002 | Model is in preview or experimental stage. |
| LLM_MODEL_STATUS_DEPRECATED | 0x0004 | Model is deprecated or legacy. |
| Limits |
| LLM_MAX_MESSAGE_LENGTH | 65536 | Maximum length of a single input message (64 KB) |
| LLM_MAX_RESPONSE_LENGTH | 262144 | Maximum expected response length (256 KB) |
| LLM_MAX_SYSTEM_PROMPT_LENGTH | 32768 | Maximum length of the system prompt (32 KB) |
| LLM_MAX_CHAT_HISTORY | 1000 | Maximum number of messages stored per session. |
| LLM_MAX_TOKENS | 200000 | Maximum token limit per request. |
| Roles |
| LLM_ROLE_UNKNOWN | 0 | Unknown or invalid role. |
| LLM_ROLE_USER | 1 | User input message. |
| LLM_ROLE_ASSISTANT | 2 | Assistant response message. |
| LLM_ROLE_SYSTEM | 3 | System-level instruction message. |
| LLM_ROLE_FUNCTION | 4 | Function call message (legacy) |
| LLM_ROLE_TOOL | 5 | Tool-generated message. |
| LLM_ROLE_TOOLUSER | 6 | User input directed to a tool. |
| LLM_ROLE_DEFAULT | LLM_ROLE_USER | Default role is user input. |
| Responses |
| LLM_RESPONSE_UNKNOWN | 0 | No stop reason specified or unrecognized response state. |
| LLM_RESPONSE_COMPLETE | 1 | Response completed normally. |
| LLM_RESPONSE_LIMITED | 2 | Stopped due to token or context limits. |
| LLM_RESPONSE_ABORTED | 3 | Request was cancelled or interrupted. |
| LLM_RESPONSE_ERROR | 4 | An error occurred during processing. |
| LLM_RESPONSE_TIMEOUT | 5 | Request timed out. |
| Session |
| LLM_SESSION_FLAG_NONE | 0 | No session flags specified. |
| LLM_SESSION_FLAG_DEFAULT | LLM_SESSION_FLAG_NONE | Default session behavior. |
| Cache |
| LLM_INVALID_CACHE_TIME | -1 | Indicates an invalid or unspecified cache duration. |