Reasoning Level in Chat Input Settings

Added a "Reasoning Level" selector to the AI settings popover in the chat input, defaulting to Medium and only visible for models that support it.

Changes

Frontend — chat-ui package

• packages/chat-ui/src/StreamingChatPanel.tsx:
  • Added ReasoningLevel type ('off' | 'low' | 'medium' | 'high' | 'xhigh')
  • Added supportsReasoning?: boolean to ChatModel interface
  • Added reasoningLevel?: ReasoningLevel to AICategoryConfig interface
  • Added reasoning level <select> in the AI settings popover, between the "AI Models" header and the model pickers
  • Only renders when the current conversation model's supportsReasoning is true

Frontend — BotDetailPage

• services/frontend/src/routes/bots/BotDetailPage.tsx:
  • Added supportsReasoning: true/false to each entry in QUANTAGO_MODELS
  • Added reasoningLevel: 'medium' to QUANTAGO_AI_CATEGORY_MODELS.defaults

Frontend — API client

• services/frontend/src/lib/api-client/index.ts:
  • Added reasoningLevel? to streamBotChatMessage opts aiConfig type

Backend — AI types

• services/backend/src/lib/ai/types.ts:
  • Added ReasoningLevel type
  • Added reasoningLevel?: ReasoningLevel to AICategoryConfig
  • Default reasoningLevel: 'medium' in DEFAULT_AI_CATEGORY_CONFIG and resolveAICategoryConfig

Backend — Chat handler

• services/backend/src/routes/bots/bots-chat-handlers.ts:
  • Added reasoningLevel? to DurableBotChatRequest['aiConfig'] and HTTP endpoint body type
  • Updated all createAIAdapter call sites to pass ctx.aiConfig?.reasoningLevel

Backend — Adapter factory

• services/backend/src/lib/ai/providers.ts:
  • Added reasoningLevel?: ReasoningLevel parameter to createAIAdapter
  • Passes reasoning effort to CodexAdapter (mapped from ReasoningLevel to ReasoningEffort, skipping 'off')

Models with reasoning support

Model	Supports Reasoning
DeepSeek V4 Flash	Yes
DeepSeek V4 Pro	Yes
GPT Codex 5.3	Yes
Claude Sonnet 4.6	Yes
Claude Haiku 4.5	No
GPT-5.4 mini	No
Claude Opus 4.7	Yes
GPT-5.5	Yes