Namespace Mythosia.AI.Models
Classes
- AICitation
A provider-supplied source reference. Fields unavailable from the provider remain null.
- AIProcessingInfo
An immutable processing-mode observation for one provider inference attempt.
- AIRequestFeatures
Provider-neutral request options. A service copies these options before using them.
- ChatBlock
Pure conversation container - holds only conversation identity and history. All settings (Temperature, MaxTokens, etc.) are managed by the service.
- ClaudeInputTransformation
A provider-reported change to input thinking blocks on a Claude request.
- FileSearchOptions
Existing provider-hosted document stores searched during a request.
- FileSearchStore
An existing search store owned by one provider/account. Store creation and uploads are managed separately.
- ReasoningOptions
Reasoning settings for the next logical request, including its tool rounds and format repairs.
- RoundResult
라운드 처리 결과
- StructuredOutputPolicy
Structured output 실행 정책. 전체 서비스 기본값(
AIService.StructuredOutputMaxRetries)과 별개로, 특정 호출에만 일회성으로 적용할 수 있습니다.
- SummaryConversationPolicy
Policy that automatically summarizes old conversation messages when the conversation exceeds a configured threshold (token count, message count, or both). The summary is stored as a string and injected into the system message on each request.
- WebSearchOptions
Options for the provider's hosted web search tool.
Enums
- ActorRole
Represents the role of an actor in a conversation
- CachePreservation
Whether a reasoning change must preserve the eligible conversation cache prefix.
- ClaudeReasoningEffort
Adaptive-thinking effort for current Claude models. Auto preserves the legacy
ThinkingBudget-to-effort mapping. With explicit adaptive thinking on Opus 5.5, Auto uses its medium default.
- ClaudeThinkingDisplay
Controls readable output from Claude adaptive thinking. Omitted hides readable reasoning; Summarized requests summaries. Updates requests supported user-facing progress updates without reasoning summaries.
- ClaudeThinkingMode
Selects the thinking phase on Claude Sonnet 5.5 independently of effort.
- ClaudeThinkingPrefixMismatchBehavior
How supported Claude requests handle thinking blocks bound to a changed conversation prefix.
- DeepSeekReasoning
Reasoning effort for DeepSeek thinking mode. Enable thinking separately with ThinkingEnabled.
- Gpt5Reasoning
Reasoning effort level for GPT-5 base models. Auto: Uses model default (Medium).
- Gpt5_1Reasoning
Reasoning effort level for GPT-5.1 models. Auto: Uses model default (None).
- Gpt5_2Reasoning
Reasoning effort level for GPT-5.2 models. Auto: Uses model default (None for GPT-5.2, Medium for GPT-5.2 Pro).
- Gpt5_3Reasoning
Reasoning effort level for GPT-5.3 models. Auto: Uses model default (Medium for Codex). GPT-5.3 Codex supports: low, medium (default), high, xhigh. GPT-5.3 Codex Spark and Instant use simplified configs.
- Gpt5_4Reasoning
Reasoning effort level for GPT-5.4 models. Auto: Uses model default (None for GPT-5.4, Medium for GPT-5.4 Pro). GPT-5.4 supports: none (default), low, medium, high, xhigh. GPT-5.4 Pro supports: medium, high, xhigh.
- Gpt5_5Reasoning
Reasoning effort level for GPT-5.5 models. Auto: Uses model default (None for GPT-5.5, Medium for GPT-5.5 Pro). GPT-5.5 supports: none (default), low, medium, high, xhigh. GPT-5.5 Pro supports: medium, high, xhigh.
- Gpt5_6Reasoning
Reasoning effort level for GPT-5.6 models. Auto uses the model default (Medium). GPT-5.6 supports: none, low, medium (default), high, xhigh, max.
- Gpt5_6ReasoningMode
Reasoning execution mode for GPT-5.6 models. Standard omits the API mode parameter; Pro sends reasoning.mode as pro.
- Gpt6Reasoning
Reasoning effort level for GPT-6 family models, including GPT-6.1 Sol. Auto uses the library default (Medium). GPT-6 supports low, medium, high, xhigh, and max. GPT-6 Sol and Luna also support None; GPT-6 Astra and GPT-6.1 Sol always reason and do not support None.
- Gpt6ReasoningMode
Reasoning execution mode for GPT-6 models. Standard omits the API mode parameter; Pro sends reasoning.mode as pro.
- GrokReasoning
Reasoning effort for xAI Grok models. Auto omits the provider parameter. Grok 4.3 supports None through High; Grok 4.5 supports Low through High; Grok 4.6 and 4.7 also support XHigh. Grok 4.5 and 4.6 cannot disable reasoning.
- InferenceSpeed
Selects a provider processing mode without changing the model or reasoning effort.
- ReasoningLevel
Requested reasoning effort. Providers validate the levels supported by the selected model.
- ReasoningSummary
Reasoning summary mode for GPT-5 family models. Controls how the model summarizes its reasoning process. Set to null (use nullable) to disable reasoning summaries entirely.
- Verbosity
Text verbosity level for supported GPT-5 family models. Controls how verbose the model's text output is.