feat: 8 AI providers, rich text editor, agent notifications, UI contrast & font settings

- Add DeepSeek, OpenRouter, Mistral, Z.AI, LM Studio as AI providers with editable model names via Combobox in admin settings - Fix OpenRouter broken by normalizeProvider bug in config.ts - Convert agent-created notes from Markdown to HTML (TipTap rich text) - Add Notification model + in-app notifications for agent results - Agent notification click opens the created note directly - Add note count display on notebook and inbox headers - Fix checklist toggle in card view (persist state via localCheckItems) - Add checklist creation option in tabs/list view (dropdown on + button) - Fix image description ENOENT error with HTTP fallback - Improve UI contrast across all themes (input, border, checkbox visibility) - Add font family setting (Inter vs System Default) in Appearance settings - Fix CSS font-sans variable conflict (removed dead Geist references) - Update README with new features and 8 providers Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
2026-05-01 16:14:07 +02:00
parent 1345403a31
commit dbd49d6fcb
64 changed files with 4124 additions and 1392 deletions
--- a/memento-note/lib/ai/factory.ts
+++ b/memento-note/lib/ai/factory.ts
@@ -3,7 +3,36 @@ import { OllamaProvider } from './providers/ollama';
 import { CustomOpenAIProvider } from './providers/custom-openai';
 import { AIProvider } from './types';

-type ProviderType = 'ollama' | 'openai' | 'custom' | 'deepseek' | 'openrouter';
+type ProviderType = 'ollama' | 'openai' | 'custom' | 'deepseek' | 'openrouter' | 'mistral' | 'zai' | 'lmstudio';
+
+// --- Provider defaults ---
+const PROVIDER_DEFAULTS: Record<string, { baseUrl: string; model: string; embeddingModel: string }> = {
+  deepseek: {
+    baseUrl: 'https://api.deepseek.com/v1',
+    model: 'deepseek-chat',
+    embeddingModel: '',
+  },
+  openrouter: {
+    baseUrl: 'https://openrouter.ai/api/v1',
+    model: 'openai/gpt-4o-mini',
+    embeddingModel: 'openai/text-embedding-3-small',
+  },
+  mistral: {
+    baseUrl: 'https://api.mistral.ai/v1',
+    model: 'mistral-small-latest',
+    embeddingModel: 'mistral-embed',
+  },
+  zai: {
+    baseUrl: 'https://api.zukijourney.com/v1',
+    model: 'gpt-4o-mini',
+    embeddingModel: 'text-embedding-3-small',
+  },
+  lmstudio: {
+    baseUrl: 'http://localhost:1234/v1',
+    model: '',
+    embeddingModel: '',
+  },
+};

 function createOllamaProvider(config: Record<string, string>, modelName: string, embeddingModelName: string, baseUrlOverride?: string): OllamaProvider {
  let baseUrl = baseUrlOverride || config?.OLLAMA_BASE_URL || process.env.OLLAMA_BASE_URL
@@ -19,7 +48,7 @@ function createOllamaProvider(config: Record<string, string>, modelName: string,

  // Ensure baseUrl doesn't end with /api, we'll add it in OllamaProvider
  if (baseUrl.endsWith('/api')) {
-    baseUrl = baseUrl.slice(0, -4); // Remove /api
+    baseUrl = baseUrl.slice(0, -4);
  }

  return new OllamaProvider(baseUrl, modelName, embeddingModelName);
@@ -51,15 +80,39 @@ function createCustomOpenAIProvider(config: Record<string, string>, modelName: s
 }

 function createDeepSeekProvider(config: Record<string, string>, modelName: string, embeddingModelName: string): CustomOpenAIProvider {
-  const apiKey = config?.DEEPSEEK_API_KEY || config?.CUSTOM_OPENAI_API_KEY || process.env.DEEPSEEK_API_KEY || process.env.CUSTOM_OPENAI_API_KEY || '';
+  const apiKey = config?.DEEPSEEK_API_KEY || process.env.DEEPSEEK_API_KEY || '';
  if (!apiKey) throw new Error('DEEPSEEK_API_KEY is required when using DeepSeek provider');
-  return new CustomOpenAIProvider(apiKey, 'https://api.deepseek.com/v1', modelName, embeddingModelName);
+  const defaults = PROVIDER_DEFAULTS.deepseek;
+  return new CustomOpenAIProvider(apiKey, defaults.baseUrl, modelName || defaults.model, embeddingModelName || defaults.embeddingModel);
 }

 function createOpenRouterProvider(config: Record<string, string>, modelName: string, embeddingModelName: string): CustomOpenAIProvider {
-  const apiKey = config?.OPENROUTER_API_KEY || config?.CUSTOM_OPENAI_API_KEY || process.env.OPENROUTER_API_KEY || process.env.CUSTOM_OPENAI_API_KEY || '';
+  const apiKey = config?.OPENROUTER_API_KEY || process.env.OPENROUTER_API_KEY || '';
  if (!apiKey) throw new Error('OPENROUTER_API_KEY is required when using OpenRouter provider');
-  return new CustomOpenAIProvider(apiKey, 'https://openrouter.ai/api/v1', modelName, embeddingModelName);
+  const defaults = PROVIDER_DEFAULTS.openrouter;
+  return new CustomOpenAIProvider(apiKey, defaults.baseUrl, modelName || defaults.model, embeddingModelName || defaults.embeddingModel);
+}
+
+function createMistralProvider(config: Record<string, string>, modelName: string, embeddingModelName: string): CustomOpenAIProvider {
+  const apiKey = config?.MISTRAL_API_KEY || process.env.MISTRAL_API_KEY || '';
+  if (!apiKey) throw new Error('MISTRAL_API_KEY is required when using Mistral provider');
+  const defaults = PROVIDER_DEFAULTS.mistral;
+  return new CustomOpenAIProvider(apiKey, defaults.baseUrl, modelName || defaults.model, embeddingModelName || defaults.embeddingModel);
+}
+
+function createZAIProvider(config: Record<string, string>, modelName: string, embeddingModelName: string): CustomOpenAIProvider {
+  const apiKey = config?.ZAI_API_KEY || process.env.ZAI_API_KEY || '';
+  if (!apiKey) throw new Error('ZAI_API_KEY is required when using Z.AI provider');
+  const defaults = PROVIDER_DEFAULTS.zai;
+  return new CustomOpenAIProvider(apiKey, defaults.baseUrl, modelName || defaults.model, embeddingModelName || defaults.embeddingModel);
+}
+
+function createLMStudioProvider(config: Record<string, string>, modelName: string, embeddingModelName: string): CustomOpenAIProvider {
+  const baseUrl = config?.LMSTUDIO_BASE_URL || process.env.LMSTUDIO_BASE_URL || PROVIDER_DEFAULTS.lmstudio.baseUrl;
+  // LM Studio doesn't require an API key, but the CustomOpenAI provider needs one
+  // Use a dummy key if not provided
+  const apiKey = config?.LMSTUDIO_API_KEY || process.env.LMSTUDIO_API_KEY || 'lm-studio';
+  return new CustomOpenAIProvider(apiKey, baseUrl, modelName, embeddingModelName);
 }

 function getProviderInstance(providerType: ProviderType, config: Record<string, string>, modelName: string, embeddingModelName: string, ollamaBaseUrl?: string): AIProvider {
@@ -74,28 +127,47 @@ function getProviderInstance(providerType: ProviderType, config: Record<string,
      return createDeepSeekProvider(config, modelName, embeddingModelName);
    case 'openrouter':
      return createOpenRouterProvider(config, modelName, embeddingModelName);
+    case 'mistral':
+      return createMistralProvider(config, modelName, embeddingModelName);
+    case 'zai':
+      return createZAIProvider(config, modelName, embeddingModelName);
+    case 'lmstudio':
+      return createLMStudioProvider(config, modelName, embeddingModelName);
    default:
      return createOllamaProvider(config, modelName, embeddingModelName, ollamaBaseUrl);
  }
 }

+// Resolve the effective provider type and config keys for a given provider
+// Returns { providerType, apiKeyConfigKey, baseUrlConfigKey }
+function getProviderConfigKeys(providerType: string): { apiKeyConfigKey: string; baseUrlConfigKey: string } {
+  switch (providerType) {
+    case 'deepseek':   return { apiKeyConfigKey: 'DEEPSEEK_API_KEY', baseUrlConfigKey: '' };
+    case 'openrouter': return { apiKeyConfigKey: 'OPENROUTER_API_KEY', baseUrlConfigKey: '' };
+    case 'mistral':    return { apiKeyConfigKey: 'MISTRAL_API_KEY', baseUrlConfigKey: '' };
+    case 'zai':        return { apiKeyConfigKey: 'ZAI_API_KEY', baseUrlConfigKey: '' };
+    case 'lmstudio':   return { apiKeyConfigKey: 'LMSTUDIO_API_KEY', baseUrlConfigKey: 'LMSTUDIO_BASE_URL' };
+    case 'openai':     return { apiKeyConfigKey: 'OPENAI_API_KEY', baseUrlConfigKey: '' };
+    case 'custom':     return { apiKeyConfigKey: 'CUSTOM_OPENAI_API_KEY', baseUrlConfigKey: 'CUSTOM_OPENAI_BASE_URL' };
+    default:           return { apiKeyConfigKey: '', baseUrlConfigKey: 'OLLAMA_BASE_URL' };
+  }
+}
+
 export function getTagsProvider(config?: Record<string, string>): AIProvider {
-  // Check database config first, then environment variables
  const providerType = (
-    config?.AI_PROVIDER_TAGS || 
-    config?.AI_PROVIDER_EMBEDDING || 
+    config?.AI_PROVIDER_TAGS ||
+    config?.AI_PROVIDER_EMBEDDING ||
    config?.AI_PROVIDER ||
    process.env.AI_PROVIDER_TAGS ||
    process.env.AI_PROVIDER_EMBEDDING ||
    process.env.AI_PROVIDER
  );

-  // If no provider is configured, throw a clear error
  if (!providerType) {
    console.error('[getTagsProvider] FATAL: No provider configured. Config received:', config);
    throw new Error(
      'AI_PROVIDER_TAGS is not configured. Please set it in the admin settings or environment variables. ' +
-      'Options: ollama, openai, custom'
+      'Options: ollama, openai, deepseek, openrouter, mistral, zai, lmstudio, custom'
    );
  }

@@ -108,22 +180,20 @@ export function getTagsProvider(config?: Record<string, string>): AIProvider {
 }

 export function getEmbeddingsProvider(config?: Record<string, string>): AIProvider {
-  // Check database config first, then environment variables
  const providerType = (
-    config?.AI_PROVIDER_EMBEDDING || 
-    config?.AI_PROVIDER_TAGS || 
+    config?.AI_PROVIDER_EMBEDDING ||
+    config?.AI_PROVIDER_TAGS ||
    config?.AI_PROVIDER ||
    process.env.AI_PROVIDER_EMBEDDING ||
    process.env.AI_PROVIDER_TAGS ||
    process.env.AI_PROVIDER
  );

-  // If no provider is configured, throw a clear error
  if (!providerType) {
    console.error('[getEmbeddingsProvider] FATAL: No provider configured. Config received:', config);
    throw new Error(
      'AI_PROVIDER_EMBEDDING is not configured. Please set it in the admin settings or environment variables. ' +
-      'Options: ollama, openai, custom'
+      'Options: ollama, openai, deepseek, openrouter, mistral, zai, lmstudio, custom'
    );
  }

@@ -140,8 +210,6 @@ export function getAIProvider(config?: Record<string, string>): AIProvider {
 }

 export function getChatProvider(config?: Record<string, string>): AIProvider {
-  // Check database config first, then environment variables
-  // Fallback cascade: chat -> tags -> embeddings
  const providerType = (
    config?.AI_PROVIDER_CHAT ||
    config?.AI_PROVIDER_TAGS ||
@@ -153,12 +221,11 @@ export function getChatProvider(config?: Record<string, string>): AIProvider {
    process.env.AI_PROVIDER
  );

-  // If no provider is configured, throw a clear error
  if (!providerType) {
    console.error('[getChatProvider] FATAL: No provider configured. Config received:', config);
    throw new Error(
      'AI_PROVIDER_CHAT is not configured. Please set it in the admin settings or environment variables. ' +
-      'Options: ollama, openai, custom'
+      'Options: ollama, openai, deepseek, openrouter, mistral, zai, lmstudio, custom'
    );
  }

@@ -173,3 +240,6 @@ export function getChatProvider(config?: Record<string, string>): AIProvider {

  return getProviderInstance(provider, config || {}, modelName, embeddingModelName, ollamaBaseUrl);
 }
+
+// Export for use by admin settings form and deploy scripts
+export { PROVIDER_DEFAULTS, getProviderConfigKeys };
--- a/memento-note/lib/ai/services/agent-executor.service.ts
+++ b/memento-note/lib/ai/services/agent-executor.service.ts
@@ -15,6 +15,8 @@ import { sendEmail } from '@/lib/mail'
 import { getAgentEmailTemplate } from '@/lib/agent-email-template'
 import { extractAndDownloadImages, extractImageUrlsFromHtml, downloadImage } from '../tools/extract-images'
 import { calculateNextRun } from '@/lib/agents/schedule'
+import { markdownToHtml } from '@/lib/markdown-to-html'
+import { createNotification } from '@/app/actions/notifications'

 // Import tools for side-effect registration
 import '../tools'
@@ -30,6 +32,32 @@ export interface AgentExecutionResult {
  error?: string
 }

+// --- Note creation helper ---
+
+/** Create an agent note as rich text (TipTap-compatible HTML).
+ *  Converts the markdown content to HTML and sets type='richtext'. */
+async function createAgentNote(data: {
+  title: string
+  content: string
+  userId: string
+  notebookId: string | null
+  autoGenerated?: boolean
+}) {
+  const htmlContent = markdownToHtml(data.content)
+  return prisma.note.create({
+    data: {
+      title: data.title,
+      content: htmlContent,
+      type: 'richtext',
+      isMarkdown: false,
+      autoGenerated: data.autoGenerated ?? true,
+      userId: data.userId,
+      notebookId: data.notebookId,
+    },
+    select: { id: true },
+  })
+}
+
 // --- Language Helper ---

 type Lang = 'fr' | 'en'
@@ -324,15 +352,11 @@ async function executeScraperAgent(

  const title = await generateTitle(fullContent, agent.name, lang)

-  const note = await prisma.note.create({
-    data: {
-      title,
-      content: fullContent,
-      isMarkdown: true,
-      autoGenerated: true,
-      userId: agent.userId,
-      notebookId: agent.targetNotebookId,
-    }
+  const note = await createAgentNote({
+    title,
+    content: fullContent,
+    userId: agent.userId,
+    notebookId: agent.targetNotebookId,
  })

  const logMsg = lang === 'fr'
@@ -424,15 +448,11 @@ async function executeResearcherAgent(

  const title = await generateTitle(fullContent, agent.name, lang)

-  const note = await prisma.note.create({
-    data: {
-      title,
-      content: fullContent,
-      isMarkdown: true,
-      autoGenerated: true,
-      userId: agent.userId,
-      notebookId: agent.targetNotebookId,
-    }
+  const note = await createAgentNote({
+    title,
+    content: fullContent,
+    userId: agent.userId,
+    notebookId: agent.targetNotebookId,
  })

  const logMsg = lang === 'fr'
@@ -526,15 +546,11 @@ async function executeMonitorAgent(

  const title = await generateTitle(fullContent, agent.name, lang)

-  const note = await prisma.note.create({
-    data: {
-      title,
-      content: fullContent,
-      isMarkdown: true,
-      autoGenerated: true,
-      userId: agent.userId,
-      notebookId: agent.targetNotebookId,
-    }
+  const note = await createAgentNote({
+    title,
+    content: fullContent,
+    userId: agent.userId,
+    notebookId: agent.targetNotebookId,
  })

  const logMsg = lang === 'fr' ? `Analyse de ${notes.length} notes. Note créée: ${note.id}` : `Analyzed ${notes.length} notes. Note created: ${note.id}`
@@ -597,15 +613,11 @@ async function executeCustomAgent(

  const title = await generateTitle(fullContent, agent.name, lang)

-  const note = await prisma.note.create({
-    data: {
-      title,
-      content: fullContent,
-      isMarkdown: true,
-      autoGenerated: true,
-      userId: agent.userId,
-      notebookId: agent.targetNotebookId,
-    }
+  const note = await createAgentNote({
+    title,
+    content: fullContent,
+    userId: agent.userId,
+    notebookId: agent.targetNotebookId,
  })

  const toolLogData = JSON.stringify([{
@@ -966,15 +978,11 @@ async function executeToolUseAgent(
    const fullContent = `# ${agent.name}\n\n${text}\n\n---\n\n_Agent execution: ${totalToolCalls} tool calls in ${Math.round(duration / 1000)}s_`
    const title = await generateTitle(fullContent, agent.name, lang)

-    const note = await prisma.note.create({
-      data: {
-        title,
-        content: fullContent,
-        isMarkdown: true,
-        autoGenerated: true,
-        userId: agent.userId,
-        notebookId: agent.targetNotebookId || null,
-      }
+    const note = await createAgentNote({
+      title,
+      content: fullContent,
+      userId: agent.userId,
+      notebookId: agent.targetNotebookId || null,
    })
    noteId = note.id
  }
@@ -1119,6 +1127,20 @@ export async function executeAgent(agentId: string, userId: string, promptOverri
      }
    }

+    // Create in-app notification for agent result
+    if (result.success) {
+      await createNotification({
+        userId,
+        type: 'agent_success',
+        title: agent.name,
+        message: result.noteId
+          ? (lang === 'fr' ? `L'agent a terminé avec succès — note créée.` : `Agent completed successfully — note created.`)
+          : (lang === 'fr' ? `L'agent a terminé avec succès.` : `Agent completed successfully.`),
+        actionUrl: result.noteId ? `/?openNote=${result.noteId}` : '/agents',
+        relatedId: result.noteId || agentId,
+      })
+    }
+
    return result
  } catch (error) {
    const message = error instanceof Error ? error.message : 'Unknown error'
@@ -1129,6 +1151,16 @@ export async function executeAgent(agentId: string, userId: string, promptOverri
      data: { status: 'failure', log: message }
    })

+    // Notify user of agent failure
+    await createNotification({
+      userId,
+      type: 'agent_failure',
+      title: agent?.name || 'Agent',
+      message: message.length > 200 ? message.substring(0, 200) + '...' : message,
+      actionUrl: '/agents',
+      relatedId: agentId,
+    })
+
    return { success: false, actionId: action.id, error: message }
  }
 }
--- a/memento-note/lib/ai/services/image-description.service.ts
+++ b/memento-note/lib/ai/services/image-description.service.ts
@@ -19,23 +19,43 @@ export interface ImageDescriptionResult {

 const UPLOAD_DIR = path.join(process.cwd(), 'data', 'uploads')

-async function resolveImageAsBase64(imageUrl: string): Promise<string> {
+async function resolveImageAsBase64(imageUrl: string): Promise<string | null> {
  const localMatch = imageUrl.match(/\/uploads\/(.+)/)
  if (localMatch) {
-    const filePath = path.join(UPLOAD_DIR, localMatch[1])
-    const buffer = await readFile(filePath)
-    const ext = path.extname(imageUrl).toLowerCase()
-    const mime = ext === '.png' ? 'image/png' : ext === '.gif' ? 'image/gif' : ext === '.webp' ? 'image/webp' : 'image/jpeg'
-    return `data:${mime};base64,${buffer.toString('base64')}`
+    // Try reading from filesystem first
+    try {
+      const filePath = path.join(UPLOAD_DIR, localMatch[1])
+      const buffer = await readFile(filePath)
+      const ext = path.extname(imageUrl).toLowerCase()
+      const mime = ext === '.png' ? 'image/png' : ext === '.gif' ? 'image/gif' : ext === '.webp' ? 'image/webp' : 'image/jpeg'
+      return `data:${mime};base64,${buffer.toString('base64')}`
+    } catch {
+      // File not on disk — fallback to internal HTTP API (same path the browser uses)
+      try {
+        const baseUrl = process.env.NEXTAUTH_URL || process.env.NEXT_PUBLIC_APP_URL || 'http://localhost:3000'
+        const res = await fetch(`${baseUrl}${imageUrl}`)
+        if (!res.ok) return null
+        const contentType = res.headers.get('content-type') || 'image/jpeg'
+        const arrayBuffer = await res.arrayBuffer()
+        const base64 = Buffer.from(arrayBuffer).toString('base64')
+        return `data:${contentType};base64,${base64}`
+      } catch {
+        return null
+      }
+    }
  }

  // Remote URL — fetch and convert
-  const res = await fetch(imageUrl)
-  if (!res.ok) throw new Error(`Failed to fetch image: ${imageUrl}`)
-  const contentType = res.headers.get('content-type') || 'image/jpeg'
-  const arrayBuffer = await res.arrayBuffer()
-  const base64 = Buffer.from(arrayBuffer).toString('base64')
-  return `data:${contentType};base64,${base64}`
+  try {
+    const res = await fetch(imageUrl)
+    if (!res.ok) return null
+    const contentType = res.headers.get('content-type') || 'image/jpeg'
+    const arrayBuffer = await res.arrayBuffer()
+    const base64 = Buffer.from(arrayBuffer).toString('base64')
+    return `data:${contentType};base64,${base64}`
+  } catch {
+    return null
+  }
 }

 export async function describeImages(
@@ -55,8 +75,9 @@ export async function describeImages(
  }
  const langName = langMap[language] || 'English'

-  // Resolve all images as base64 data URLs (same approach as the chat route)
-  const imageDataUrls = await Promise.all(imageUrls.map(url => resolveImageAsBase64(url)))
+  // Resolve all images as base64 data URLs — skip any that can't be found
+  const resolved = await Promise.all(imageUrls.map(url => resolveImageAsBase64(url)))
+  const imageDataUrls = resolved.filter((d): d is string => d !== null)

  if (isTitleMode) {
    const prompt = imageUrls.length === 1
--- a/memento-note/lib/ai/tools/note-crud.tool.ts
+++ b/memento-note/lib/ai/tools/note-crud.tool.ts
@@ -7,6 +7,7 @@ import { tool } from 'ai'
 import { z } from 'zod'
 import { toolRegistry } from './registry'
 import { prisma } from '@/lib/prisma'
+import { markdownToHtml } from '@/lib/markdown-to-html'

 // --- note_read ---
 toolRegistry.register({
@@ -50,11 +51,13 @@ toolRegistry.register({
      }),
      execute: async ({ title, content, notebookId, images }) => {
        try {
+          const htmlContent = markdownToHtml(content)
          const note = await prisma.note.create({
            data: {
              title,
-              content,
-              isMarkdown: true,
+              content: htmlContent,
+              type: 'richtext',
+              isMarkdown: false,
              autoGenerated: true,
              userId: ctx.userId,
              notebookId: notebookId || null,
--- a/memento-note/lib/config.ts
+++ b/memento-note/lib/config.ts
@@ -1,24 +1,32 @@
 import prisma from './prisma'

-// "openrouter" était une ancienne valeur de provider — on la normalise en "custom"
-function normalizeProvider(val: string | undefined): string {
-  if (!val) return ''
-  return val === 'openrouter' ? 'custom' : val
-}
-
 // Environment variable fallbacks for system config keys
 const ENV_FALLBACKS: Record<string, string> = {
-  // AI providers (openrouter → custom)
-  AI_PROVIDER_TAGS: normalizeProvider(process.env.AI_PROVIDER_TAGS),
+  // AI providers
+  AI_PROVIDER_TAGS: process.env.AI_PROVIDER_TAGS || '',
  AI_MODEL_TAGS: process.env.AI_MODEL_TAGS || '',
-  AI_PROVIDER_EMBEDDING: normalizeProvider(process.env.AI_PROVIDER_EMBEDDING),
+  AI_PROVIDER_EMBEDDING: process.env.AI_PROVIDER_EMBEDDING || '',
  AI_MODEL_EMBEDDING: process.env.AI_MODEL_EMBEDDING || '',
-  AI_PROVIDER_CHAT: normalizeProvider(process.env.AI_PROVIDER_CHAT),
+  AI_PROVIDER_CHAT: process.env.AI_PROVIDER_CHAT || '',
  AI_MODEL_CHAT: process.env.AI_MODEL_CHAT || '',
+  // Ollama
  OLLAMA_BASE_URL: process.env.OLLAMA_BASE_URL || '',
+  // OpenAI
  OPENAI_API_KEY: process.env.OPENAI_API_KEY || '',
+  // Custom OpenAI
  CUSTOM_OPENAI_API_KEY: process.env.CUSTOM_OPENAI_API_KEY || '',
  CUSTOM_OPENAI_BASE_URL: process.env.CUSTOM_OPENAI_BASE_URL || '',
+  // DeepSeek
+  DEEPSEEK_API_KEY: process.env.DEEPSEEK_API_KEY || '',
+  // OpenRouter
+  OPENROUTER_API_KEY: process.env.OPENROUTER_API_KEY || '',
+  // Mistral
+  MISTRAL_API_KEY: process.env.MISTRAL_API_KEY || '',
+  // Z.AI
+  ZAI_API_KEY: process.env.ZAI_API_KEY || '',
+  // LM Studio
+  LMSTUDIO_BASE_URL: process.env.LMSTUDIO_BASE_URL || '',
+  LMSTUDIO_API_KEY: process.env.LMSTUDIO_API_KEY || '',
  // Email
  EMAIL_PROVIDER: process.env.EMAIL_PROVIDER || (process.env.RESEND_API_KEY ? 'resend' : 'smtp'),
  RESEND_API_KEY: process.env.RESEND_API_KEY || '',
@@ -51,11 +59,6 @@ export async function getSystemConfig() {
    console.error('Failed to load system config from DB:', e)
  }

-  // Normalise les valeurs openrouter → custom dans la DB aussi
-  for (const key of ['AI_PROVIDER_TAGS', 'AI_PROVIDER_EMBEDDING', 'AI_PROVIDER_CHAT']) {
-    if (dbConfig[key] === 'openrouter') dbConfig[key] = 'custom'
-  }
-
  // Merge: DB values take precedence, env vars as fallback
  const merged = { ...ENV_FALLBACKS, ...dbConfig }
  return merged
--- a/memento-note/lib/i18n/detect-user-language.ts
+++ b/memento-note/lib/i18n/detect-user-language.ts
@@ -1,6 +1,9 @@
 /**
- * Detect user's preferred language from their existing notes
- * Uses a single DB-level GROUP BY query — no note content is loaded
+ * Detect user's preferred language.
+ * Priority:
+ * 1. Most common language among user's notes (DB GROUP BY)
+ * 2. Browser language hint (passed from server component via Accept-Language)
+ * 3. Default: 'en'
 */

 import { auth } from '@/auth'
@@ -10,10 +13,37 @@ import { SupportedLanguage } from './load-translations'

 const SUPPORTED_LANGUAGES = new Set(['en', 'fr', 'es', 'de', 'fa', 'it', 'pt', 'ru', 'zh', 'ja', 'ko', 'ar', 'hi', 'nl', 'pl'])

+/**
+ * Parse an Accept-Language header string and find the best matching supported language
+ */
+export function parseAcceptLanguage(acceptLanguage: string | null): SupportedLanguage | null {
+  if (!acceptLanguage) return null
+
+  // Parse Accept-Language: "fr-FR,fr;q=0.9,en-US;q=0.8,en;q=0.7"
+  const languages = acceptLanguage
+    .split(',')
+    .map(lang => {
+      const [code, q] = lang.trim().split(';q=')
+      return { code: code.trim().toLowerCase(), quality: q ? parseFloat(q) : 1.0 }
+    })
+    .sort((a, b) => b.quality - a.quality)
+
+  for (const { code } of languages) {
+    if (SUPPORTED_LANGUAGES.has(code)) {
+      return code as SupportedLanguage
+    }
+    const base = code.split('-')[0]
+    if (SUPPORTED_LANGUAGES.has(base)) {
+      return base as SupportedLanguage
+    }
+  }
+
+  return null
+}
+
 const getCachedUserLanguage = unstable_cache(
-  async (userId: string): Promise<SupportedLanguage> => {
+  async (userId: string): Promise<SupportedLanguage | null> => {
    try {
-      // Single aggregated query — no notes are fetched, only language counts
      const result = await prisma.note.groupBy({
        by: ['language'],
        where: {
@@ -33,22 +63,39 @@ const getCachedUserLanguage = unstable_cache(
        }
      }

-      return 'en'
+      return null
    } catch (error) {
-      console.error('Error detecting user language:', error)
-      return 'en'
+      console.error('Error detecting user language from notes:', error)
+      return null
    }
  },
  ['user-language'],
  { tags: ['user-language'] }
 )

-export async function detectUserLanguage(): Promise<SupportedLanguage> {
+/**
+ * Detect user language.
+ * @param browserLanguageHint - Optional browser language parsed from Accept-Language header.
+ *                               Should be passed from server components that have access to headers().
+ */
+export async function detectUserLanguage(browserLanguageHint?: SupportedLanguage | null): Promise<SupportedLanguage> {
  const session = await auth()

  if (!session?.user?.id) {
-    return 'en'
+    return browserLanguageHint || 'en'
  }

-  return getCachedUserLanguage(session.user.id)
+  // 1. Try to detect from user's notes
+  const noteLanguage = await getCachedUserLanguage(session.user.id)
+  if (noteLanguage) {
+    return noteLanguage
+  }
+
+  // 2. Fall back to browser language hint
+  if (browserLanguageHint) {
+    return browserLanguageHint
+  }
+
+  // 3. Default
+  return 'en'
 }
--- a/memento-note/lib/markdown-to-html.ts
+++ b/memento-note/lib/markdown-to-html.ts
@@ -0,0 +1,225 @@
+/**
+ * Server-side Markdown → HTML converter.
+ * Converts AI-generated markdown notes into TipTap-compatible rich text HTML.
+ * Uses a lightweight regex-based approach to avoid heavy remark/rehype dependencies.
+ *
+ * Handles: headings, bold, italic, strikethrough, code blocks, inline code,
+ * links, images, lists (ul/ol), blockquotes, horizontal rules, tables, paragraphs.
+ */
+
+export function markdownToHtml(markdown: string): string {
+  if (!markdown || !markdown.trim()) return ''
+
+  let html = markdown
+
+  // Escape HTML entities (but preserve markdown)
+  html = html.replace(/&/g, '&amp;')
+  html = html.replace(/</g, '&lt;')
+  html = html.replace(/>/g, '&gt;')
+
+  // Code blocks (``` ... ```) — protect from further processing
+  const codeBlocks: string[] = []
+  html = html.replace(/```(\w*)\n([\s\S]*?)```/g, (_match, lang, code) => {
+    const idx = codeBlocks.length
+    codeBlocks.push(`<pre><code class="language-${lang || 'plaintext'}">${code.trim()}</code></pre>`)
+    return `%%CODEBLOCK_${idx}%%`
+  })
+
+  // Inline code (`...`)
+  html = html.replace(/`([^`]+)`/g, '<code>$1</code>')
+
+  // Images (![alt](url))
+  html = html.replace(/!\[([^\]]*)\]\(([^)]+)\)/g, '<img src="$2" alt="$1" />')
+
+  // Links ([text](url))
+  html = html.replace(/\[([^\]]+)\]\(([^)]+)\)/g, '<a href="$2" target="_blank" rel="noopener noreferrer">$1</a>')
+
+  // Headings (h1-h6)
+  html = html.replace(/^######\s+(.+)$/gm, '<h6>$1</h6>')
+  html = html.replace(/^#####\s+(.+)$/gm, '<h5>$1</h5>')
+  html = html.replace(/^####\s+(.+)$/gm, '<h4>$1</h4>')
+  html = html.replace(/^###\s+(.+)$/gm, '<h3>$1</h3>')
+  html = html.replace(/^##\s+(.+)$/gm, '<h2>$1</h2>')
+  html = html.replace(/^#\s+(.+)$/gm, '<h1>$1</h1>')
+
+  // Bold (**text** or __text__)
+  html = html.replace(/\*\*([^*]+)\*\*/g, '<strong>$1</strong>')
+  html = html.replace(/__([^_]+)__/g, '<strong>$1</strong>')
+
+  // Italic (*text* or _text_)
+  html = html.replace(/(?<!\*)\*([^*]+)\*(?!\*)/g, '<em>$1</em>')
+  html = html.replace(/(?<!_)_([^_]+)_(?!_)/g, '<em>$1</em>')
+
+  // Strikethrough (~~text~~)
+  html = html.replace(/~~([^~]+)~~/g, '<s>$1</s>')
+
+  // Horizontal rules (---, ***, ___)
+  html = html.replace(/^(-{3,}|\*{3,}|_{3,})$/gm, '<hr />')
+
+  // Tables
+  html = convertTables(html)
+
+  // Blockquotes (> text)
+  html = html.replace(/^&gt;\s+(.+)$/gm, '<blockquote><p>$1</p></blockquote>')
+  // Merge consecutive blockquotes
+  html = html.replace(/<\/blockquote>\n<blockquote>/g, '\n')
+
+  // Unordered lists (- item or * item)
+  html = convertUnorderedLists(html)
+
+  // Ordered lists (1. item)
+  html = convertOrderedLists(html)
+
+  // Restore code blocks
+  codeBlocks.forEach((block, idx) => {
+    html = html.replace(`%%CODEBLOCK_${idx}%%`, block)
+  })
+
+  // Paragraphs — wrap remaining loose text in <p> tags
+  html = wrapParagraphs(html)
+
+  // Clean up empty paragraphs
+  html = html.replace(/<p>\s*<\/p>/g, '')
+
+  return html.trim()
+}
+
+function convertTables(html: string): string {
+  // Simple table conversion: | header | header |\n| --- | --- |\n| cell | cell |
+  const tableRegex = /(?:^|\n)((?:\|[^\n]+\|\n)+)/g
+
+  return html.replace(tableRegex, (match) => {
+    const rows = match.trim().split('\n').filter(r => r.trim())
+    if (rows.length < 2) return match
+
+    // Check if second row is separator
+    const separator = rows[1].trim()
+    if (!/^[\s|:-]+$/.test(separator)) return match
+
+    let table = '<table>'
+
+    // Header row
+    const headers = parseTableRow(rows[0])
+    if (headers.length > 0) {
+      table += '<thead><tr>'
+      headers.forEach(h => { table += `<th>${h}</th>` })
+      table += '</tr></thead>'
+    }
+
+    // Body rows (skip separator)
+    const bodyRows = rows.slice(2)
+    if (bodyRows.length > 0) {
+      table += '<tbody>'
+      bodyRows.forEach(row => {
+        const cells = parseTableRow(row)
+        table += '<tr>'
+        cells.forEach(c => { table += `<td>${c}</td>` })
+        table += '</tr>'
+      })
+      table += '</tbody>'
+    }
+
+    table += '</table>'
+    return '\n' + table + '\n'
+  })
+}
+
+function parseTableRow(row: string): string[] {
+  return row.split('|')
+    .map(cell => cell.trim())
+    .filter((_, i, arr) => i > 0 && i < arr.length) // Skip first and last empty from leading/trailing |
+}
+
+function convertUnorderedLists(html: string): string {
+  const lines = html.split('\n')
+  const result: string[] = []
+  let inList = false
+
+  for (const line of lines) {
+    const listMatch = line.match(/^(\s*)[-*]\s+(.+)$/)
+    if (listMatch) {
+      if (!inList) {
+        result.push('<ul>')
+        inList = true
+      }
+      result.push(`<li>${listMatch[2]}</li>`)
+    } else {
+      if (inList) {
+        result.push('</ul>')
+        inList = false
+      }
+      result.push(line)
+    }
+  }
+  if (inList) result.push('</ul>')
+
+  return result.join('\n')
+}
+
+function convertOrderedLists(html: string): string {
+  const lines = html.split('\n')
+  const result: string[] = []
+  let inList = false
+
+  for (const line of lines) {
+    const listMatch = line.match(/^(\s*)\d+\.\s+(.+)$/)
+    if (listMatch) {
+      if (!inList) {
+        result.push('<ol>')
+        inList = true
+      }
+      result.push(`<li>${listMatch[2]}</li>`)
+    } else {
+      if (inList) {
+        result.push('</ol>')
+        inList = false
+      }
+      result.push(line)
+    }
+  }
+  if (inList) result.push('</ol>')
+
+  return result.join('\n')
+}
+
+function wrapParagraphs(html: string): string {
+  const blockTags = new Set(['h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'ul', 'ol', 'li', 'blockquote', 'pre', 'table', 'thead', 'tbody', 'tr', 'th', 'td', 'hr', 'p', 'div', 'img'])
+
+  const lines = html.split('\n')
+  const result: string[] = []
+  let buffer: string[] = []
+
+  const flushBuffer = () => {
+    const text = buffer.join('\n').trim()
+    if (text) {
+      // Don't double-wrap if already starts with a block tag
+      const firstTag = text.match(/^<(\w+)/)
+      if (firstTag && blockTags.has(firstTag[1].toLowerCase())) {
+        result.push(text)
+      } else {
+        result.push(`<p>${text}</p>`)
+      }
+    }
+    buffer = []
+  }
+
+  for (const line of lines) {
+    const trimmed = line.trim()
+
+    // Check if this line is a block-level element
+    const isBlockLine = trimmed.startsWith('<') && (() => {
+      const tag = trimmed.match(/^<(\w+)/)
+      return tag ? blockTags.has(tag[1].toLowerCase()) : false
+    })()
+
+    if (isBlockLine || trimmed === '') {
+      flushBuffer()
+      if (isBlockLine) result.push(trimmed)
+    } else {
+      buffer.push(trimmed)
+    }
+  }
+  flushBuffer()
+
+  return result.join('\n')
+}