refactor: rewrite model-context to use js-yaml, add context_length to provider form (#177)

* fix: context-length API returns 200K instead of actual model context Two bugs cause the /api/hermes/sessions/context-length endpoint to always return DEFAULT_CONTEXT_LENGTH (200K): 1. getModelContextLength ignores config.yaml model.context_length The function only checks models_dev_cache.json (which doesn't exist in default installations) and falls back to the hardcoded 200K default, completely ignoring the user's explicit model.context_length setting in config.yaml. 2. getDefaultModel regex fails when api_key/base_url come before default The regex /^model:\s*\n\s+default:\s*(.+)$/m assumes 'default' is the first child key under 'model:', but when api_key or base_url appear first in the YAML, the match fails. This causes getModelContextLength to short-circuit to DEFAULT_CONTEXT_LENGTH before even reaching the cache lookup. Fix: - Add getDefaultModelRobust() that extracts the entire model: block first, then searches for default: within it - Add getConfigContextLength() that reads model.context_length from config.yaml as a fallback (matching hermes-agent priority) - Update getModelContextLength() resolution order: 1. models_dev_cache.json (existing) 2. config.yaml model.context_length (new) 3. DEFAULT_CONTEXT_LENGTH (existing fallback) Closes #169 * refactor: rewrite model-context to use js-yaml, add context_length to provider form - Replace fragile regex-based YAML parsing with js-yaml for reliable config.yaml reads - Fix context_length resolution priority: config.yaml override > custom_providers > models_dev_cache > 200K default - Add context_length input field when adding custom providers in ProviderFormModal - Backend: persist context_length to custom_providers models.<model>.context_length in config.yaml - Add i18n keys (contextLength, contextLengthPlaceholder) to all 8 locales Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: use NInputNumber instead of NInput type=number for context_length NInput does not support type="number" in Naive UI. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: devilardis <53129661@qq.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-04-24 11:18:11 +08:00
parent 30e88797ef
commit 82965ae6e2
12 changed files with 143 additions and 18 deletions
@@ -45,6 +45,7 @@ export interface CustomProvider {
  base_url: string
  api_key: string
  model: string
+  context_length?: number
  providerKey?: string | null
 }

@@ -1,6 +1,6 @@
 <script setup lang="ts">
 import { ref, watch, computed, onMounted } from 'vue'
-import { NModal, NForm, NFormItem, NInput, NButton, NSelect, useMessage } from 'naive-ui'
+import { NModal, NForm, NFormItem, NInput, NInputNumber, NButton, NSelect, useMessage } from 'naive-ui'
 import { useModelsStore } from '@/stores/hermes/models'
 import { useI18n } from 'vue-i18n'
 import CodexLoginModal from './CodexLoginModal.vue'
@@ -29,6 +29,7 @@ const formData = ref({
  base_url: '',
  api_key: '',
  model: '',
+  context_length: null as number | null,
 })

 const modelOptions = ref<Array<{ label: string; value: string }>>([])
@@ -75,7 +76,7 @@ watch(() => formData.value.base_url, (url) => {

 watch(providerType, () => {
  modelOptions.value = []
-  formData.value = { name: '', base_url: '', api_key: '', model: '' }
+  formData.value = { name: '', base_url: '', api_key: '', model: '', context_length: null }
  selectedPreset.value = null
 })

@@ -154,11 +155,13 @@ async function handleSave() {
      ? selectedPreset.value
      : null

+    const contextLength = formData.value.context_length ?? undefined
    await modelsStore.addProvider({
      name: formData.value.name.trim(),
      base_url: formData.value.base_url.trim(),
      api_key: formData.value.api_key.trim(),
      model: formData.value.model,
+      context_length: contextLength,
      providerKey,
    })
    message.success(t('models.providerAdded'))
@@ -270,6 +273,16 @@ function handleClose() {
          </NButton>
        </div>
      </NFormItem>
+
+      <NFormItem v-if="providerType === 'custom'" :label="t('models.contextLength')">
+        <NInputNumber
+          v-model:value="formData.context_length as number | null"
+          :placeholder="t('models.contextLengthPlaceholder')"
+          :min="0"
+          clearable
+          style="width: 100%"
+        />
+      </NFormItem>
    </NForm>

    <template #footer>
@@ -259,6 +259,8 @@ export default {
    builtIn: 'Integriert',
    customType: 'Benutzerdefiniert',
    provider: 'Anbieter',
+    contextLength: 'Kontextlange',
+    contextLengthPlaceholder: 'z.B. 200000 (optional)',
    local: 'Lokal ({host})',
    selectProviderRequired: 'Bitte wahlen Sie einen Anbieter',
    baseUrlRequired: 'Basis-URL ist erforderlich',
@@ -284,6 +284,8 @@ export default {
    builtIn: 'Built-in',
    customType: 'Custom',
    provider: 'Provider',
+    contextLength: 'Context Length',
+    contextLengthPlaceholder: 'e.g. 200000 (optional)',
    local: 'Local ({host})',
    selectProviderRequired: 'Please select a provider',
    baseUrlRequired: 'Base URL is required',
@@ -259,6 +259,8 @@ export default {
    builtIn: 'Integrado',
    customType: 'Personalizado',
    provider: 'Proveedor',
+    contextLength: 'Longitud del contexto',
+    contextLengthPlaceholder: 'ej. 200000 (opcional)',
    local: 'Local ({host})',
    selectProviderRequired: 'Por favor, selecciona un proveedor',
    baseUrlRequired: 'La URL base es obligatoria',
@@ -259,6 +259,8 @@ export default {
    builtIn: 'Integre',
    customType: 'Personnalise',
    provider: 'Fournisseur',
+    contextLength: 'Longueur du contexte',
+    contextLengthPlaceholder: 'ex. 200000 (facultatif)',
    local: 'Local ({host})',
    selectProviderRequired: 'Veuillez selectionner un fournisseur',
    baseUrlRequired: 'L\'URL de base est requise',
@@ -259,6 +259,8 @@ export default {
    builtIn: '組み込み',
    customType: 'カスタム',
    provider: 'プロバイダー',
+    contextLength: 'コンテキスト長',
+    contextLengthPlaceholder: '例: 200000（任意）',
    local: 'ローカル ({host})',
    selectProviderRequired: 'プロバイダーを選択してください',
    baseUrlRequired: 'ベース URL は必須です',
@@ -259,6 +259,8 @@ export default {
    builtIn: '내장',
    customType: '사용자 지정',
    provider: 'Provider',
+    contextLength: '컨텍스트 길이',
+    contextLengthPlaceholder: '예: 200000 (선택사항)',
    local: '로컬 ({host})',
    selectProviderRequired: 'Provider를 선택해 주세요',
    baseUrlRequired: 'Base URL을 입력해 주세요',
@@ -259,6 +259,8 @@ export default {
    builtIn: 'Integrado',
    customType: 'Personalizado',
    provider: 'Provedor',
+    contextLength: 'Tamanho do contexto',
+    contextLengthPlaceholder: 'ex: 200000 (opcional)',
    local: 'Local ({host})',
    selectProviderRequired: 'Por favor, selecione um provedor',
    baseUrlRequired: 'A URL base e obrigatoria',
@@ -284,6 +284,8 @@ export default {
    builtIn: '内置',
    customType: '自定义',
    provider: 'Provider',
+    contextLength: '上下文长度',
+    contextLengthPlaceholder: '例如 200000（可选）',
    local: '本地 ({host})',
    selectProviderRequired: '请选择 Provider',
    baseUrlRequired: 'Base URL 为必填项',
@@ -5,9 +5,17 @@ import * as hermesCli from '../../services/hermes/hermes-cli'
 import { readConfigYaml, writeConfigYaml, saveEnvValue, PROVIDER_ENV_MAP } from '../../services/config-helpers'
 import { logger } from '../../services/logger'

+function buildProviderEntry(name: string, base_url: string, api_key: string, model: string, context_length?: number) {
+  const entry: any = { name, base_url, api_key, model }
+  if (context_length && context_length > 0) {
+    entry.models = { [model]: { context_length } }
+  }
+  return entry
+}
+
 export async function create(ctx: any) {
-  const { name, base_url, api_key, model, providerKey } = ctx.request.body as {
-    name: string; base_url: string; api_key: string; model: string; providerKey?: string | null
+  const { name, base_url, api_key, model, context_length, providerKey } = ctx.request.body as {
+    name: string; base_url: string; api_key: string; model: string; context_length?: number; providerKey?: string | null
  }
  console.log(name, base_url, api_key, model, providerKey)
  if (!name || !base_url || !model) {
@@ -30,8 +38,13 @@ export async function create(ctx: any) {
        existing.base_url = base_url
        existing.api_key = api_key
        existing.model = model
+        if (context_length && context_length > 0) {
+          if (!existing.models) existing.models = {}
+          existing.models[model] = existing.models[model] || {}
+          existing.models[model].context_length = context_length
+        }
      } else {
-        config.custom_providers.push({ name: name.trim().toLowerCase().replace(/ /g, '-'), base_url, api_key, model })
+        config.custom_providers.push(buildProviderEntry(name.trim().toLowerCase().replace(/ /g, '-'), base_url, api_key, model, context_length))
      }
      config.model.default = model
      config.model.provider = poolKey
@@ -51,8 +64,13 @@ export async function create(ctx: any) {
          existing.base_url = base_url
          existing.api_key = api_key
          existing.model = model
+          if (context_length && context_length > 0) {
+            if (!existing.models) existing.models = {}
+            existing.models[model] = existing.models[model] || {}
+            existing.models[model].context_length = context_length
+          }
        } else {
-          config.custom_providers.push({ name: poolKey, base_url, api_key, model })
+          config.custom_providers.push(buildProviderEntry(poolKey, base_url, api_key, model, context_length))
        }
        config.model.default = model
        config.model.provider = `custom:${poolKey}`
@@ -1,6 +1,7 @@
 import { resolve, join } from 'path'
 import { homedir } from 'os'
 import { readFileSync, existsSync, statSync } from 'fs'
+import yaml from 'js-yaml'

 const HERMES_BASE = resolve(homedir(), '.hermes')
 const MODELS_DEV_CACHE = resolve(HERMES_BASE, 'models_dev_cache.json')
@@ -21,6 +22,18 @@ interface ProviderEntry {
  models?: Record<string, ModelEntry>
 }

+// --- Config YAML helpers (js-yaml) ---
+
+function loadConfig(profileDir: string): any | null {
+  const configPath = join(profileDir, 'config.yaml')
+  if (!existsSync(configPath)) return null
+  try {
+    return yaml.load(readFileSync(configPath, 'utf-8')) as any
+  } catch {
+    return null
+  }
+}
+
 // --- In-memory cache: parsed models_dev_cache (1.7MB), invalidated by mtime ---

 let _cache: Record<string, ProviderEntry> | null = null
@@ -55,16 +68,59 @@ function getProfileDir(profile?: string): string {
  return existsSync(dir) ? dir : HERMES_BASE
 }

-function getDefaultModel(profileDir: string): string | null {
-  const configPath = join(profileDir, 'config.yaml')
-  if (!existsSync(configPath)) return null
-  try {
-    const content = readFileSync(configPath, 'utf-8')
-    const match = content.match(/^model:\s*\n\s+default:\s*(.+)$/m)
-    return match ? match[1].trim() : null
-  } catch {
-    return null
+function getDefaultModel(config: any): string | null {
+  const model = config?.model
+  if (!model || typeof model !== 'object') return null
+  return typeof model.default === 'string' ? model.default.trim() || null : null
+}
+
+function getDefaultProvider(config: any): string | null {
+  const model = config?.model
+  if (!model || typeof model !== 'object') return null
+  return typeof model.provider === 'string' ? model.provider.trim() || null : null
+}
+
+/**
+ * Read context_length from config.yaml, only as a sibling of default.
+ * e.g. model:\n  default: gpt-5.4\n  context_length: 200000
+ */
+function getConfigContextLength(config: any): number | null {
+  const model = config?.model
+  if (!model || typeof model !== 'object') return null
+  const val = model.context_length
+  if (typeof val !== 'number' || !Number.isFinite(val) || val <= 0) return null
+  return val
+}
+
+/**
+ * Lookup context_length from custom_providers in config.yaml.
+ * - "custom:xxx" → strip prefix, match by name
+ * - "custom" → match by model name
+ */
+function lookupCustomProviderContextLength(config: any, modelName: string, provider: string | null): number | null {
+  const providers: any[] = Array.isArray(config?.custom_providers) ? config.custom_providers : []
+  if (!provider || !provider.startsWith('custom')) return null
+
+  let matched: any = null
+
+  if (provider === 'custom') {
+    matched = providers.find((cp: any) => cp.model === modelName)
+  } else {
+    const suffix = provider.slice('custom:'.length)
+    matched = providers.find((cp: any) => cp.name === suffix)
  }
+
+  if (!matched) return null
+
+  const models = matched.models
+  if (!models || typeof models !== 'object') return null
+
+  const modelEntry = models[modelName]
+  if (!modelEntry || typeof modelEntry !== 'object') return null
+
+  const val = modelEntry.context_length
+  if (typeof val !== 'number' || !Number.isFinite(val) || val <= 0) return null
+  return val
 }

 // --- Context lookup ---
@@ -95,12 +151,33 @@ function lookupContextFromCache(modelName: string): number | null {

 /**
 * Get the context length for the current profile's default model.
- * Results are cached in memory (5min TTL) and invalidated by file mtime.
+ * Resolution order:
+ *   1. config.yaml model.context_length (highest priority, user override)
+ *   2. custom_providers models.<model>.context_length
+ *   3. models_dev_cache.json (built-in model database)
+ *   4. DEFAULT_CONTEXT_LENGTH (200K hardcoded fallback)
 */
 export function getModelContextLength(profile?: string): number {
  const profileDir = getProfileDir(profile)
-  const model = getDefaultModel(profileDir)
+  const config = loadConfig(profileDir)
+  if (!config) return DEFAULT_CONTEXT_LENGTH
+
+  const model = getDefaultModel(config)
  if (!model) return DEFAULT_CONTEXT_LENGTH

-  return lookupContextFromCache(model) || DEFAULT_CONTEXT_LENGTH
+  // 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
+
+  // 3. models_dev_cache.json
+  const cached = lookupContextFromCache(model)
+  if (cached) return cached
+
+  // 4. Fallback
+  return DEFAULT_CONTEXT_LENGTH
 }