feat(agentic): render integration

Author: deepinsect

docs/specs/agentic-renderer-integration/chatConfig-removal-spec.md

@@ -9,6 +9,7 @@ import { modelCapabilities } from '../../configPresenter/modelCapabilities'
 import { enhanceSystemPromptWithDateTime } from '../utility/promptEnhancer'
 import { ToolCallCenter } from '../tool/toolCallCenter'
 import { nanoid } from 'nanoid'
+import { getRuntimeConfig } from '../runtimeConfig'
 
 import {
   addContextMessages,
@@ -83,29 +84,9 @@ function appendPromptSection(base: string, section: string): string {
 }
 
 async function resolveSystemPrompt(conversation: CONVERSATION): Promise<string> {
-  const systemPromptId = conversation.settings.systemPromptId?.trim()
-  if (!systemPromptId) {
-    return conversation.settings.systemPrompt || ''
-  }
-
-  if (systemPromptId === 'empty') {
-    return ''
-  }
-
-  try {
-    if (systemPromptId === 'default') {
-      return await presenter.configPresenter.getDefaultSystemPrompt()
-    }
-    const prompts = await presenter.configPresenter.getSystemPrompts()
-    const matchedPrompt = prompts.find((prompt) => prompt.id === systemPromptId)
-    if (matchedPrompt?.content) {
-      return matchedPrompt.content
-    }
-  } catch (error) {
-    console.warn('AgentPresenter: Failed to resolve system prompt id', error)
-  }
-
-  return conversation.settings.systemPrompt || ''
+  // Phase 6: System prompts removed - use agent default
+  const runtimeConfig = getRuntimeConfig(conversation)
+  return runtimeConfig.systemPrompt
 }
 
 export async function preparePromptContent({
@@ -121,7 +102,9 @@ export async function preparePromptContent({
   finalContent: ChatMessage[]
   promptTokens: number
 }> {
-  const { contextLength, artifacts, enabledMcpTools } = conversation.settings
+  // Phase 6: Get runtime config from agent defaults
+  const runtimeConfig = getRuntimeConfig(conversation)
+  const { contextLength, artifacts, enabledMcpTools } = runtimeConfig
   const systemPrompt = await resolveSystemPrompt(conversation)
   const isAgentMode = true
   const isToolPromptMode = true

docs/specs/agentic-renderer-integration/tasks.md

@@ -18,6 +18,7 @@ import type { StreamGenerationHandler } from '../streaming/streamGenerationHandl
 import type { LLMEventHandler } from '../streaming/llmEventHandler'
 import { BaseHandler, type ThreadHandlerContext } from '../baseHandler'
 import { CommandPermissionService } from '../../permission/commandPermissionService'
+import { getRuntimeConfig } from '../runtimeConfig'
 
 export class PermissionHandler extends BaseHandler {
   private readonly generatingMessages: Map<string, GeneratingMessageState>
@@ -379,6 +380,8 @@ export class PermissionHandler extends BaseHandler {
       const { conversation, contextMessages, userMessage } =
         await this.streamGenerationHandler.prepareConversationContext(conversationId, messageId)
 
+      // Phase 6: Get runtime config instead of reading from settings
+      const runtimeConfig = getRuntimeConfig(conversation)
       const {
         providerId,
         modelId,
@@ -391,7 +394,7 @@ export class PermissionHandler extends BaseHandler {
         enableSearch,
         forcedSearch,
         searchStrategy
-      } = conversation.settings
+      } = runtimeConfig
 
       const modelConfig = this.ctx.configPresenter.getModelConfig(modelId, providerId)
       const completedToolCall = {
@@ -477,6 +480,8 @@ export class PermissionHandler extends BaseHandler {
       const { contextMessages, userMessage } =
         await this.streamGenerationHandler.prepareConversationContext(conversationId, messageId)
 
+      // Phase 6: Get runtime config instead of reading from settings
+      const runtimeConfig = getRuntimeConfig(conversation)
       const modelConfig = this.ctx.configPresenter.getModelConfig(
         conversation.settings.modelId,
         conversation.settings.providerId
@@ -495,15 +500,15 @@ export class PermissionHandler extends BaseHandler {
         finalContent,
         conversation.settings.modelId,
         messageId,
-        conversation.settings.temperature,
-        conversation.settings.maxTokens,
-        conversation.settings.enabledMcpTools,
-        conversation.settings.thinkingBudget,
-        conversation.settings.reasoningEffort,
-        conversation.settings.verbosity,
-        conversation.settings.enableSearch,
-        conversation.settings.forcedSearch,
-        conversation.settings.searchStrategy,
+        runtimeConfig.temperature,
+        runtimeConfig.maxTokens,
+        runtimeConfig.enabledMcpTools,
+        runtimeConfig.thinkingBudget,
+        runtimeConfig.reasoningEffort,
+        runtimeConfig.verbosity,
+        runtimeConfig.enableSearch,
+        runtimeConfig.forcedSearch,
+        runtimeConfig.searchStrategy,
         conversationId
       )
 
@@ -546,6 +551,8 @@ export class PermissionHandler extends BaseHandler {
       const { conversation, contextMessages, userMessage } =
         await this.streamGenerationHandler.prepareConversationContext(conversationId, message.id)
 
+      // Phase 6: Get runtime config instead of reading from settings
+      const runtimeConfig = getRuntimeConfig(conversation)
       const {
         providerId,
         modelId,
@@ -558,7 +565,7 @@ export class PermissionHandler extends BaseHandler {
         enableSearch,
         forcedSearch,
         searchStrategy
-      } = conversation.settings
+      } = runtimeConfig
 
       const modelConfig = this.ctx.configPresenter.getModelConfig(modelId, providerId)
       if (!modelConfig) {

src/main/presenter/agentPresenter/message/messageBuilder.ts

@@ -0,0 +1,64 @@
+/**
+ * Phase 6: Runtime Configuration Helper
+ *
+ * This module provides helper functions to get runtime configuration values
+ * that were previously stored in CONVERSATION_SETTINGS but now come from:
+ * 1. Agent's SessionInfo (runtime state from the agent)
+ * 2. Model configuration defaults
+ * 3. Hardcoded defaults for removed features
+ */
+
+import { CONVERSATION } from '@shared/presenter'
+import { presenter } from '@/presenter'
+
+export interface RuntimeConfig {
+  providerId: string
+  modelId: string
+  systemPrompt: string
+  systemPromptId: string | undefined
+  temperature: number
+  contextLength: number
+  maxTokens: number
+  artifacts: number
+  enabledMcpTools: string[] | undefined
+  thinkingBudget: number | undefined
+  reasoningEffort: 'minimal' | 'low' | 'medium' | 'high' | undefined
+  verbosity: 'low' | 'medium' | 'high' | undefined
+  enableSearch: boolean | undefined
+  forcedSearch: boolean | undefined
+  searchStrategy: 'turbo' | 'max' | undefined
+  selectedVariantsMap: Record<string, string>
+  activeSkills: string[]
+}
+
+/**
+ * Get runtime configuration for a conversation
+ * These values now come from agent's SessionInfo, not from stored settings
+ */
+export function getRuntimeConfig(conversation: CONVERSATION): RuntimeConfig {
+  // Get model config for defaults
+  const modelConfig = presenter.configPresenter.getModelConfig(
+    conversation.settings.modelId,
+    conversation.settings.providerId
+  )
+
+  return {
+    providerId: conversation.settings.providerId,
+    modelId: conversation.settings.modelId,
+    systemPrompt: '', // Phase 6: Use agent default
+    systemPromptId: undefined, // Phase 6: System prompts removed
+    contextLength: modelConfig?.contextLength || 4000,
+    maxTokens: modelConfig?.maxTokens || 2000,
+    temperature: modelConfig?.temperature ?? 0.7,
+    artifacts: 0, // Phase 6: Artifacts feature removed
+    enabledMcpTools: undefined, // Phase 6: All tools enabled by default
+    thinkingBudget: modelConfig?.thinkingBudget,
+    reasoningEffort: undefined, // Phase 6: Use agent default
+    verbosity: undefined, // Phase 6: Use agent default
+    enableSearch: undefined, // Phase 6: Use agent default
+    forcedSearch: undefined, // Phase 6: Use agent default
+    searchStrategy: undefined, // Phase 6: Use agent default
+    selectedVariantsMap: {}, // Phase 6: Variant management removed
+    activeSkills: [] // Phase 6: Active skills feature removed
+  }
+}

src/main/presenter/agentPresenter/permission/permissionHandler.ts

@@ -15,6 +15,6 @@ export function resolveSessionContext(input: SessionResolveInput): SessionContex
     supportsVision: modelConfig?.vision ?? false,
     supportsFunctionCall: modelConfig?.functionCall ?? false,
     agentWorkspacePath: settings.agentWorkspacePath ?? null,
-    enabledMcpTools: settings.enabledMcpTools
+    enabledMcpTools: undefined // Phase 6: All MCP tools enabled by default
   }
 }

src/main/presenter/agentPresenter/runtimeConfig.ts

@@ -18,6 +18,7 @@ import type { LLMEventHandler } from './llmEventHandler'
 import { LoopOrchestrator } from '../loop/loopOrchestrator'
 import type { AgenticEventEmitter } from '@shared/types/presenters/agentic.presenter.d'
 import { normalizeAndEmit } from '../normalizer'
+import { getRuntimeConfig } from '../runtimeConfig'
 
 interface StreamGenerationHandlerDeps {
   generatingMessages: Map<string, GeneratingMessageState>
@@ -105,6 +106,8 @@ export class StreamGenerationHandler extends BaseHandler {
       this.throwIfCancelled(state.message.id)
 
       const currentConversation = await this.getConversation(conversationId)
+      // Phase 6: Get runtime config instead of reading from settings
+      const runtimeConfig = getRuntimeConfig(currentConversation)
       const {
         providerId: currentProviderId,
         modelId: currentModelId,
@@ -117,7 +120,7 @@ export class StreamGenerationHandler extends BaseHandler {
         enableSearch: currentEnableSearch,
         forcedSearch: currentForcedSearch,
         searchStrategy: currentSearchStrategy
-      } = currentConversation.settings
+      } = runtimeConfig
 
       const stream = this.ctx.llmProviderPresenter.startStreamCompletion(
         currentProviderId,
@@ -217,6 +220,8 @@ export class StreamGenerationHandler extends BaseHandler {
 
       this.throwIfCancelled(state.message.id)
 
+      // Phase 6: Get runtime config instead of reading from settings
+      const runtimeConfig = getRuntimeConfig(conversation)
       const {
         providerId,
         modelId,
@@ -229,7 +234,7 @@ export class StreamGenerationHandler extends BaseHandler {
         enableSearch,
         forcedSearch,
         searchStrategy
-      } = conversation.settings
+      } = runtimeConfig
       const modelConfig = this.ctx.configPresenter.getModelConfig(modelId, providerId)
       if (!modelConfig) {
         throw new Error(`Model config not found for ${providerId}/${modelId}`)
@@ -442,9 +447,11 @@ export class StreamGenerationHandler extends BaseHandler {
         throw new Error('Unsupported message type')
       }
 
+      // Phase 6: Get context length from runtime config
+      const runtimeConfig = getRuntimeConfig(conversation)
       contextMessages = await this.ctx.messageManager.getMessageHistory(
         userMessage.id,
-        conversation.settings.contextLength
+        runtimeConfig.contextLength
       )
     } else {
       userMessage = await this.ctx.messageManager.getLastUserMessage(conversationId)
@@ -635,7 +642,9 @@ export class StreamGenerationHandler extends BaseHandler {
   }
 
   private async getContextMessages(conversation: CONVERSATION): Promise<Message[]> {
-    let messageCount = Math.ceil(conversation.settings.contextLength / 300)
+    // Phase 6: Get context length from runtime config
+    const runtimeConfig = getRuntimeConfig(conversation)
+    let messageCount = Math.ceil(runtimeConfig.contextLength / 300)
     if (messageCount < 2) {
       messageCount = 2
     }

src/main/presenter/agentPresenter/session/sessionResolver.ts

@@ -6,7 +6,7 @@ import type {
   MCPToolDefinition,
   MESSAGE_METADATA
 } from '@shared/presenter'
-import type { AssistantMessageBlock, Message, UserMessageContent } from '@shared/chat'
+import type { AssistantMessageBlock, UserMessageContent } from '@shared/chat'
 import { ModelType } from '@shared/model'
 import { presenter } from '@/presenter'
 import { BaseHandler, type ThreadHandlerContext } from '../baseHandler'
@@ -18,6 +18,7 @@ import {
 } from '../../exporter/formats/conversationExporter'
 import { preparePromptContent } from '../message/messageBuilder'
 import type { StreamGenerationHandler } from '../streaming/streamGenerationHandler'
+import { getRuntimeConfig } from '../runtimeConfig'
 
 // Translation constants
 const TRANSLATION_TEMPERATURE = 0.3
@@ -204,13 +205,10 @@ export class UtilityHandler extends BaseHandler {
       // Filter out unsent messages
       const validMessages = messages.filter((msg) => msg.status === 'sent')
 
-      // Apply variant selection
-      const selectedVariantsMap = conversation.settings.selectedVariantsMap || {}
-      const variantAwareMessages = this.applyVariantSelection(validMessages, selectedVariantsMap)
-
+      // Phase 6: Variant management removed - use messages directly
       // Generate filename
       const filename = generateExportFilename(format)
-      const content = buildConversationExportContent(conversation, variantAwareMessages, format)
+      const content = buildConversationExportContent(conversation, validMessages, format)
 
       return { filename, content }
     } catch (error) {
@@ -231,16 +229,16 @@ export class UtilityHandler extends BaseHandler {
     }
     const { providerId, modelId } = conversation.settings
 
-    // Get context messages
-    let messageCount = Math.ceil(conversation.settings.contextLength / DEFAULT_MESSAGE_LENGTH)
+    // Phase 6: Get context length from runtime config
+    const runtimeConfig = getRuntimeConfig(conversation)
+    let messageCount = Math.ceil(runtimeConfig.contextLength / DEFAULT_MESSAGE_LENGTH)
     if (messageCount < 2) {
       messageCount = 2
     }
     const messages = await this.ctx.messageManager.getContextMessages(conversation.id, messageCount)
 
-    const selectedVariantsMap = conversation.settings.selectedVariantsMap || {}
-    const variantAwareMessages = this.applyVariantSelection(messages, selectedVariantsMap)
-    const messagesWithLength = variantAwareMessages
+    // Phase 6: Variant management removed - use messages directly
+    const messagesWithLength = messages
       .map((msg) => {
         if (msg.role === 'user') {
           const userContent = msg.content as UserMessageContent
@@ -288,13 +286,15 @@ export class UtilityHandler extends BaseHandler {
       }
 
       const conversation = await this.ctx.sqlitePresenter.getConversation(message.conversation_id)
+      // Phase 6: Get runtime config instead of reading from settings
+      const runtimeConfig = getRuntimeConfig(conversation)
       const {
         providerId: defaultProviderId,
         modelId: defaultModelId,
         temperature,
         maxTokens,
         enabledMcpTools
-      } = conversation.settings
+      } = runtimeConfig
 
       // Parse metadata to get model_provider and model_id
       let messageMetadata: MESSAGE_METADATA | null = null
@@ -317,9 +317,10 @@ export class UtilityHandler extends BaseHandler {
       const userMessage = this.ctx.messageManager['convertToMessage'](userMessageSqlite)
 
       // Get context messages using getMessageHistory
+      // Phase 6: Use context length from runtime config
       const contextMessages = await this.streamGenerationHandler.getMessageHistory(
         userMessage.id,
-        conversation.settings.contextLength
+        runtimeConfig.contextLength
       )
 
       // Prepare prompt content (reconstruct what was sent)
@@ -437,31 +438,5 @@ export class UtilityHandler extends BaseHandler {
       throw error
     }
   }
-
-  /**
-   * Applies variant selection to messages based on selectedVariantsMap.
-   * Returns messages with selected variant fields applied when a variant is selected.
-   */
-  private applyVariantSelection(
-    messages: Message[],
-    selectedVariantsMap: Record<string, string>
-  ): Message[] {
-    return messages.map((msg) => {
-      if (msg.role === 'assistant' && selectedVariantsMap[msg.id] && msg.variants) {
-        const selectedVariantId = selectedVariantsMap[msg.id]
-        const selectedVariant = msg.variants.find((v) => v.id === selectedVariantId)
-
-        if (selectedVariant) {
-          const newMsg = JSON.parse(JSON.stringify(msg))
-          newMsg.content = selectedVariant.content
-          newMsg.usage = selectedVariant.usage
-          newMsg.model_id = selectedVariant.model_id
-          newMsg.model_provider = selectedVariant.model_provider
-          newMsg.model_name = selectedVariant.model_name
-          return newMsg
-        }
-      }
-      return msg
-    })
-  }
 }
+// Phase 6: applyVariantSelection method removed - variant management feature removed