feat: Add database table for model context length configuration (#477)

* feat: add database table for model context length configuration - Add model_context table with provider/model/context_limit fields - Implement UPSERT endpoint for model context configuration - Add priority lookup: database > config.yaml > custom_providers > cache - Add frontend click-to-edit UI in ChatInput with tooltip - Add i18n support for context editing dialog (all 8 locales) - Use context_limit field consistently across frontend and backend Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix: use useMessage() composable instead of window.$message in ChatInput - Remove incorrect NMessage import (not a component) - Use useMessage() composable from naive-ui - Replace window.$message?.xxx() with message.xxx() - Fixes TypeScript build errors Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-06 15:05:44 +08:00
parent f338aeea18
commit 479e1feef6
14 changed files with 406 additions and 3 deletions
@@ -0,0 +1,41 @@
+import { request } from '../client'
+
+export interface ModelContext {
+  id: number
+  provider: string
+  model: string
+  context_limit: number
+}
+
+/**
+ * 根据 provider 和 model 查询模型上下文配置
+ */
+export async function getModelContext(provider: string, model: string): Promise<ModelContext | null> {
+  try {
+    const res = await request<{ data: ModelContext }>(
+      `/api/hermes/model-context?provider=${encodeURIComponent(provider)}&model=${encodeURIComponent(model)}`
+    )
+    return res.data
+  } catch (err: any) {
+    if (err.status === 404) return null
+    throw err
+  }
+}
+
+/**
+ * 设置模型上下文配置（UPSERT：存在则更新，不存在则插入）
+ */
+export async function setModelContext(
+  provider: string,
+  model: string,
+  contextLimit: number
+): Promise<ModelContext> {
+  const res = await request<{ success: boolean; data: ModelContext }>(
+    `/api/hermes/model-context/${encodeURIComponent(provider)}/${encodeURIComponent(model)}`,
+    {
+      method: 'PUT',
+      body: JSON.stringify({ provider, model, context_limit: contextLimit }),
+    }
+  )
+  return res.data
+}
@@ -4,12 +4,14 @@ import { useChatStore } from '@/stores/hermes/chat'
 import { useAppStore } from '@/stores/hermes/app'
 import { useProfilesStore } from '@/stores/hermes/profiles'
 import { fetchContextLength } from '@/api/hermes/sessions'
-import { NButton, NTooltip, NSwitch } from 'naive-ui'
+import { setModelContext } from '@/api/hermes/model-context'
+import { NButton, NTooltip, NSwitch, NModal, NInputNumber, useMessage } from 'naive-ui'
 import { computed, ref, onMounted, watch } from 'vue'
 import { useI18n } from 'vue-i18n'

 const chatStore = useChatStore()
 const { t } = useI18n()
+const message = useMessage()
 const inputText = ref('')
 const textareaRef = ref<HTMLTextAreaElement>()
 const fileInputRef = ref<HTMLInputElement>()
@@ -45,6 +47,44 @@ const canSend = computed(() => inputText.value.trim() || attachments.value.lengt
 const contextLength = ref(200000)
 const FALLBACK_CONTEXT = 200000

+// Context length editing
+const showContextEditModal = ref(false)
+const editingContextLimit = ref(200000)
+const isSavingContextLimit = ref(false)
+
+async function handleEditContextLimit() {
+  editingContextLimit.value = contextLength.value
+  showContextEditModal.value = true
+}
+
+async function saveContextLimit() {
+  if (!editingContextLimit.value || editingContextLimit.value <= 0) {
+    message.error(t('chat.contextEditInvalid'))
+    return
+  }
+
+  isSavingContextLimit.value = true
+  try {
+    const appStore = useAppStore()
+    const provider = appStore.selectedProvider || ''
+    const model = appStore.selectedModel || ''
+
+    if (!provider || !model) {
+      message.error(t('chat.contextEditFailed'))
+      return
+    }
+
+    await setModelContext(provider, model, editingContextLimit.value)
+    contextLength.value = editingContextLimit.value
+    showContextEditModal.value = false
+    message.success(t('chat.contextEditSuccess'))
+  } catch (err: any) {
+    message.error(`${t('chat.contextEditFailed')}: ${err.message || ''}`)
+  } finally {
+    isSavingContextLimit.value = false
+  }
+}
+
 async function loadContextLength() {
  try {
    const profile = useProfilesStore().activeProfileName || undefined
@@ -247,7 +287,16 @@ function isImage(type: string): boolean {
      </div>

      <span v-if="totalTokens > 0" class="context-info" :class="{ 'context-warning': usagePercent > 80 }">
-        {{ formatTokens(totalTokens) }} / {{ formatTokens(contextLength) }} · {{ t('chat.contextRemaining') }} {{ formatTokens(remainingTokens) }}
+        {{ formatTokens(totalTokens) }} /
+        <NTooltip trigger="hover">
+          <template #trigger>
+            <span class="context-limit-editable" @click="handleEditContextLimit">
+              {{ formatTokens(contextLength) }}
+            </span>
+          </template>
+          <span>{{ t('chat.contextClickToEdit') }}</span>
+        </NTooltip>
+        · {{ t('chat.contextRemaining') }} {{ formatTokens(remainingTokens) }}
      </span>
      <div v-if="totalTokens > 0" class="context-bar">
        <div
@@ -335,6 +384,47 @@ function isImage(type: string): boolean {
        </NButton>
      </div>
    </div>
+
+    <!-- Context Length Edit Modal -->
+    <NModal
+      v-model:show="showContextEditModal"
+      :title="t('chat.contextEditTitle')"
+      :mask-closable="true"
+      preset="card"
+      style="width: 400px"
+    >
+      <div class="context-edit-content">
+        <p style="margin-bottom: 16px; color: #666;">
+          {{ t('chat.contextEditDesc') }}
+        </p>
+        <NInputNumber
+          v-model:value="editingContextLimit"
+          :min="1000"
+          :max="10000000"
+          :step="1000"
+          :show-button="false"
+          :placeholder="t('chat.contextEditPlaceholder')"
+          style="width: 100%"
+        >
+          <template #suffix>
+            <span style="color: #999;">tokens</span>
+          </template>
+        </NInputNumber>
+        <div style="margin-top: 12px; font-size: 12px; color: #999;">
+          {{ t('chat.contextEditHint') }}
+        </div>
+      </div>
+      <template #footer>
+        <div style="display: flex; justify-content: flex-end; gap: 8px;">
+          <NButton @click="showContextEditModal = false" :disabled="isSavingContextLimit">
+            {{ t('chat.contextEditCancel') }}
+          </NButton>
+          <NButton type="primary" @click="saveContextLimit" :loading="isSavingContextLimit">
+            {{ t('chat.contextEditSave') }}
+          </NButton>
+        </div>
+      </template>
+    </NModal>
  </div>
 </template>

@@ -383,6 +473,19 @@ function isImage(type: string): boolean {
  }
 }

+.context-limit-editable {
+  cursor: pointer;
+  border-bottom: 1px dashed transparent;
+  transition: all 0.2s ease;
+  padding: 0 2px;
+
+  &:hover {
+    border-bottom-color: $text-muted;
+    background: rgba(128, 128, 128, 0.1);
+    border-radius: 2px;
+  }
+}
+
 .context-bar {
  width: 60px;
  height: 4px;
@@ -104,6 +104,16 @@ export default {
  // Chat
  chat: {
    contextRemaining: 'übrig',
+    contextClickToEdit: 'Klicken zum Bearbeiten der Kontextlänge',
+    contextEditTitle: 'Kontextlänge bearbeiten',
+    contextEditDesc: 'Kontextlängenlimit für aktuelles Modell festlegen (in Tokens)',
+    contextEditPlaceholder: 'Kontextlänge eingeben',
+    contextEditHint: 'Häufige Werte: 200k (Claude), 128k (GPT-4), 32k (GPT-3.5)',
+    contextEditSave: 'Speichern',
+    contextEditCancel: 'Abbrechen',
+    contextEditInvalid: 'Bitte geben Sie eine gültige Kontextlänge ein',
+    contextEditSuccess: 'Kontextlänge aktualisiert',
+    contextEditFailed: 'Aktualisierung fehlgeschlagen',
    emptyState: 'Starten Sie eine Konversation mit Hermes Agent',
    inputPlaceholder: 'Nachricht eingeben... (Enter zum Senden, Shift+Enter fur neue Zeile)',
    attachFiles: 'Dateien anhangen',
@@ -114,6 +114,16 @@ export default {
  // Chat
  chat: {
    contextRemaining: 'remaining',
+    contextClickToEdit: 'Click to edit context length',
+    contextEditTitle: 'Edit Context Length',
+    contextEditDesc: 'Set context length limit for current model (in tokens)',
+    contextEditPlaceholder: 'Enter context length',
+    contextEditHint: 'Common values: 200k (Claude), 128k (GPT-4), 32k (GPT-3.5)',
+    contextEditSave: 'Save',
+    contextEditCancel: 'Cancel',
+    contextEditInvalid: 'Please enter a valid context length',
+    contextEditSuccess: 'Context length updated',
+    contextEditFailed: 'Update failed',
    emptyState: 'Start a conversation with Hermes Agent',
    inputPlaceholder: 'Type a message... (Enter to send, Shift+Enter for new line)',
    attachFiles: 'Attach files',
@@ -104,6 +104,16 @@ export default {
  // Chat
  chat: {
    contextRemaining: 'restante',
+    contextClickToEdit: 'Haz clic para editar la longitud del contexto',
+    contextEditTitle: 'Editar longitud del contexto',
+    contextEditDesc: 'Establecer el límite de longitud del contexto para el modelo actual (en tokens)',
+    contextEditPlaceholder: 'Ingresa la longitud del contexto',
+    contextEditHint: 'Valores comunes: 200k (Claude), 128k (GPT-4), 32k (GPT-3.5)',
+    contextEditSave: 'Guardar',
+    contextEditCancel: 'Cancelar',
+    contextEditInvalid: 'Por favor ingresa una longitud de contexto válida',
+    contextEditSuccess: 'Longitud del contexto actualizada',
+    contextEditFailed: 'Error en la actualización',
    emptyState: 'Inicia una conversacion con Hermes Agent',
    inputPlaceholder: 'Escribe un mensaje... (Enter para enviar, Shift+Enter para nueva linea)',
    attachFiles: 'Adjuntar archivos',
@@ -104,6 +104,16 @@ export default {
  // Chat
  chat: {
    contextRemaining: 'restant',
+    contextClickToEdit: 'Cliquez pour modifier la longueur du contexte',
+    contextEditTitle: 'Modifier la longueur du contexte',
+    contextEditDesc: 'Définir la limite de longueur du contexte pour le modèle actuel (en tokens)',
+    contextEditPlaceholder: 'Entrez la longueur du contexte',
+    contextEditHint: 'Valeurs courantes : 200k (Claude), 128k (GPT-4), 32k (GPT-3.5)',
+    contextEditSave: 'Enregistrer',
+    contextEditCancel: 'Annuler',
+    contextEditInvalid: 'Veuillez entrer une longueur de contexte valide',
+    contextEditSuccess: 'Longueur du contexte mise à jour',
+    contextEditFailed: 'Échec de la mise à jour',
    emptyState: 'Demarrer une conversation avec Hermes Agent',
    inputPlaceholder: 'Tapez un message... (Entree pour envoyer, Shift+Entree pour un saut de ligne)',
    attachFiles: 'Joindre des fichiers',
@@ -104,6 +104,16 @@ export default {
  // チャット
  chat: {
    contextRemaining: '残り',
+    contextClickToEdit: 'クリックしてコンテキスト長を編集',
+    contextEditTitle: 'コンテキスト長を編集',
+    contextEditDesc: '現在のモデルのコンテキスト長制限を設定（トークン数）',
+    contextEditPlaceholder: 'コンテキスト長を入力',
+    contextEditHint: '一般的な値：200k (Claude), 128k (GPT-4), 32k (GPT-3.5)',
+    contextEditSave: '保存',
+    contextEditCancel: 'キャンセル',
+    contextEditInvalid: '有効なコンテキスト長を入力してください',
+    contextEditSuccess: 'コンテキスト長を更新しました',
+    contextEditFailed: '更新に失敗しました',
    emptyState: 'Hermes Agent と会話を開始しましょう',
    inputPlaceholder: 'メッセージを入力... (Enter で送信、Shift+Enter で改行)',
    attachFiles: 'ファイルを添付',
@@ -104,6 +104,16 @@ export default {
  // 채팅
  chat: {
    contextRemaining: '남음',
+    contextClickToEdit: '클릭하여 컨텍스트 길이 편집',
+    contextEditTitle: '컨텍스트 길이 편집',
+    contextEditDesc: '현재 모델의 컨텍스트 길이 제한 설정 (토큰 수)',
+    contextEditPlaceholder: '컨텍스트 길이 입력',
+    contextEditHint: '일반적인 값: 200k (Claude), 128k (GPT-4), 32k (GPT-3.5)',
+    contextEditSave: '저장',
+    contextEditCancel: '취소',
+    contextEditInvalid: '유효한 컨텍스트 길이를 입력하세요',
+    contextEditSuccess: '컨텍스트 길이가 업데이트되었습니다',
+    contextEditFailed: '업데이트 실패',
    emptyState: 'Hermes Agent와 대화를 시작하세요',
    inputPlaceholder: '메시지를 입력하세요... (Enter로 전송, Shift+Enter로 줄바꿈)',
    attachFiles: '파일 첨부',
@@ -104,6 +104,16 @@ export default {
  // Chat
  chat: {
    contextRemaining: 'restante',
+    contextClickToEdit: 'Clique para editar o tamanho do contexto',
+    contextEditTitle: 'Editar tamanho do contexto',
+    contextEditDesc: 'Definir o limite de tamanho do contexto para o modelo atual (em tokens)',
+    contextEditPlaceholder: 'Digite o tamanho do contexto',
+    contextEditHint: 'Valores comuns: 200k (Claude), 128k (GPT-4), 32k (GPT-3.5)',
+    contextEditSave: 'Salvar',
+    contextEditCancel: 'Cancelar',
+    contextEditInvalid: 'Por favor, insira um tamanho de contexto válido',
+    contextEditSuccess: 'Tamanho do contexto atualizado',
+    contextEditFailed: 'Falha na atualização',
    emptyState: 'Inicie uma conversa com o Hermes Agent',
    inputPlaceholder: 'Digite uma mensagem... (Enter para enviar, Shift+Enter para nova linha)',
    attachFiles: 'Anexar arquivos',
@@ -114,6 +114,16 @@ export default {
  // 对话
  chat: {
    contextRemaining: '剩余',
+    contextClickToEdit: '点击编辑上下文长度',
+    contextEditTitle: '编辑上下文长度',
+    contextEditDesc: '设置当前模型的上下文长度限制（token 数量）',
+    contextEditPlaceholder: '请输入上下文长度',
+    contextEditHint: '常见值：200k (Claude), 128k (GPT-4), 32k (GPT-3.5)',
+    contextEditSave: '保存',
+    contextEditCancel: '取消',
+    contextEditInvalid: '请输入有效的上下文长度',
+    contextEditSuccess: '上下文长度已更新',
+    contextEditFailed: '更新失败',
    emptyState: '开始与 Hermes Agent 对话',
    inputPlaceholder: '输入消息... (Enter 发送，Shift+Enter 换行)',
    attachFiles: '添加附件',
@@ -5,6 +5,8 @@ import { readConfigYaml, writeConfigYaml, fetchProviderModels, buildModelGroups,
 import { buildProviderModelMap, PROVIDER_PRESETS } from '../../shared/providers'
 import { getCopilotModelsDetailed, resolveCopilotOAuthToken, type CopilotModelMeta } from '../../services/hermes/copilot-models'
 import { readAppConfig } from '../../services/app-config'
+import { getDb } from '../../db'
+import { MODEL_CONTEXT_TABLE } from '../../db/hermes/schemas'

 const PROVIDER_MODEL_CATALOG = buildProviderModelMap()

@@ -236,3 +238,126 @@ export async function setConfigModel(ctx: any) {
    ctx.body = { error: err.message }
  }
 }
+
+/**
+ * 设置模型上下文配置（UPSERT：存在则更新，不存在则插入）
+ * 支持路径参数和查询参数两种方式
+ */
+export async function updateModelContext(ctx: any) {
+  // 支持两种方式：
+  // 1. 路径参数: /api/hermes/model-context/:provider/:model
+  // 2. 查询参数: /api/hermes/model-context?provider=xxx&model=xxx
+  let provider: string | undefined
+  let model: string | undefined
+
+  // 优先从路径参数获取
+  if (ctx.params.provider && ctx.params.model) {
+    provider = ctx.params.provider
+    model = ctx.params.model
+  } else {
+    // 从查询参数获取
+    const query = ctx.query as { provider?: string; model?: string }
+    provider = query.provider
+    model = query.model
+  }
+
+  // 如果没有参数，从请求体获取
+  if (!provider || !model) {
+    const body = ctx.request.body as { provider?: string; model?: string; context_limit?: number }
+    provider = body.provider
+    model = body.model
+  }
+
+  const { context_limit } = ctx.request.body as { context_limit: number }
+
+  if (!provider || !model || !context_limit) {
+    ctx.status = 400
+    ctx.body = { error: 'Missing required fields: provider, model, context_limit' }
+    return
+  }
+
+  if (typeof context_limit !== 'number' || context_limit <= 0) {
+    ctx.status = 400
+    ctx.body = { error: 'Context limit must be a positive number' }
+    return
+  }
+
+  try {
+    const db = getDb()
+    if (!db) {
+      ctx.status = 500
+      ctx.body = { error: 'Database not available' }
+      return
+    }
+
+    // 使用 REPLACE 实现 UPSERT：存在则替换，不存在则插入
+    db.prepare(
+      `REPLACE INTO ${MODEL_CONTEXT_TABLE} (provider, model, context_limit) VALUES (?, ?, ?)`
+    ).run(provider, model, context_limit)
+
+    // 查询并返回更新后的数据
+    const row = db.prepare(
+      `SELECT id, provider, model, context_limit FROM ${MODEL_CONTEXT_TABLE} WHERE provider = ? AND model = ?`
+    ).get(provider, model) as { id: number; provider: string; model: string; context_limit: number }
+
+    ctx.body = {
+      success: true,
+      data: row
+    }
+  } catch (err: any) {
+    ctx.status = 500
+    ctx.body = { error: err.message }
+  }
+}
+
+/**
+ * 查询模型上下文配置
+ */
+export async function getModelContext(ctx: any) {
+  // 支持两种方式：
+  // 1. 路径参数: /api/hermes/model-context/:provider/:model
+  // 2. 查询参数: /api/hermes/model-context?provider=xxx&model=xxx
+  let provider: string | undefined
+  let model: string | undefined
+
+  // 优先从路径参数获取
+  if (ctx.params.provider && ctx.params.model) {
+    provider = ctx.params.provider
+    model = ctx.params.model
+  } else {
+    // 从查询参数获取
+    const query = ctx.query as { provider?: string; model?: string }
+    provider = query.provider
+    model = query.model
+  }
+
+  if (!provider || !model) {
+    ctx.status = 400
+    ctx.body = { error: 'Missing provider or model parameter' }
+    return
+  }
+
+  try {
+    const db = getDb()
+    if (!db) {
+      ctx.status = 500
+      ctx.body = { error: 'Database not available' }
+      return
+    }
+
+    const row = db.prepare(
+      `SELECT id, provider, model, context_limit FROM ${MODEL_CONTEXT_TABLE} WHERE provider = ? AND model = ?`
+    ).get(provider, model) as { id: number; provider: string; model: string; context_limit: number } | undefined
+
+    if (!row) {
+      ctx.status = 404
+      ctx.body = { error: 'Model context not found' }
+      return
+    }
+
+    ctx.body = { data: { ...row, limit: row.context_limit } }
+  } catch (err: any) {
+    ctx.status = 500
+    ctx.body = { error: err.message }
+  }
+}
@@ -89,6 +89,21 @@ export const COMPRESSION_SNAPSHOT_SCHEMA: Record<string, string> = {
  updated_at: 'INTEGER NOT NULL',
 }

+// ============================================================================
+// Model Context (model-context.ts)
+// ============================================================================
+
+export const MODEL_CONTEXT_TABLE = 'model_context'
+
+export const MODEL_CONTEXT_SCHEMA: Record<string, string> = {
+  id: 'INTEGER PRIMARY KEY AUTOINCREMENT',
+  provider: 'TEXT NOT NULL',
+  model: 'TEXT NOT NULL',
+  context_limit: 'INTEGER NOT NULL',
+}
+
+export const MODEL_CONTEXT_INDEX = 'CREATE UNIQUE INDEX IF NOT EXISTS idx_model_context_provider_model ON model_context(provider, model)'
+
 // ============================================================================
 // Group Chat (services/hermes/group-chat/index.ts)
 // ============================================================================
@@ -469,6 +484,13 @@ export function initAllHermesTables(retryCount = 0): void {
    // Compression snapshot
    syncTable(COMPRESSION_SNAPSHOT_TABLE, COMPRESSION_SNAPSHOT_SCHEMA)

+    // Model context
+    syncTable(MODEL_CONTEXT_TABLE, MODEL_CONTEXT_SCHEMA, {
+      indexes: {
+        idx_model_context_provider_model: MODEL_CONTEXT_INDEX,
+      }
+    })
+
    // Group chat - basic tables
    syncTable(GC_ROOMS_TABLE, GC_ROOMS_SCHEMA)
    syncTable(GC_MESSAGES_TABLE, GC_MESSAGES_SCHEMA)
@@ -6,3 +6,9 @@ export const modelRoutes = new Router()
 modelRoutes.get('/api/hermes/available-models', ctrl.getAvailable)
 modelRoutes.get('/api/hermes/config/models', ctrl.getConfigModels)
 modelRoutes.put('/api/hermes/config/model', ctrl.setConfigModel)
+
+// Model context routes
+modelRoutes.get('/api/hermes/model-context', ctrl.getModelContext)
+modelRoutes.get('/api/hermes/model-context/:provider/:model', ctrl.getModelContext)
+modelRoutes.put('/api/hermes/model-context/:provider/:model', ctrl.updateModelContext)
+modelRoutes.put('/api/hermes/model-context', ctrl.updateModelContext)
@@ -2,6 +2,8 @@ import { resolve, join } from 'path'
 import { homedir } from 'os'
 import { readFileSync, existsSync, statSync } from 'fs'
 import yaml from 'js-yaml'
+import { getDb } from '../../db'
+import { MODEL_CONTEXT_TABLE } from '../../db/hermes/schemas'

 const HERMES_BASE = resolve(homedir(), '.hermes')
 const MODELS_DEV_CACHE = resolve(HERMES_BASE, 'models_dev_cache.json')
@@ -234,6 +236,25 @@ function lookupContextFromCache(modelName: string, provider: string | null): num
 *   3. models_dev_cache.json, scoped to model.provider when configured
 *   4. DEFAULT_CONTEXT_LENGTH (200K hardcoded fallback)
 */
+/**
+ * 从数据库 model_context 表查找上下文长度（最高优先级）
+ */
+function lookupContextFromDatabase(modelName: string, provider: string | null): number | null {
+  const db = getDb()
+  if (!db) return null
+
+  try {
+    // 尝试精确匹配 provider 和 model
+    const row = db
+      .prepare(`SELECT context_limit FROM ${MODEL_CONTEXT_TABLE} WHERE provider = ? AND model = ?`)
+      .get(provider || 'default', modelName) as { context_limit: number } | undefined
+
+    return row?.context_limit || null
+  } catch {
+    return null
+  }
+}
+
 export function getModelContextLength(profile?: string): number {
  const profileDir = getProfileDir(profile)
  const config = loadConfig(profileDir)
@@ -242,12 +263,17 @@ export function getModelContextLength(profile?: string): number {
  const model = getDefaultModel(config)
  if (!model) return DEFAULT_CONTEXT_LENGTH

+  const provider = getDefaultProvider(config)
+
+  // 0. Database model_context table (highest priority)
+  const dbCtx = lookupContextFromDatabase(model, provider)
+  if (dbCtx && dbCtx > 0) return dbCtx
+
  // 1. Global context_length override in config.yaml
  const configCtx = getConfigContextLength(config)
  if (configCtx && configCtx > 0) return configCtx

  // 2. Custom provider context_length
-  const provider = getDefaultProvider(config)
  const customCtx = lookupCustomProviderContextLength(config, model, provider)
  if (customCtx && customCtx > 0) return customCtx