levix-bot 3.3.0-rc.2 → 4.0.0-beta

package/src/services/aiAgent.cjs

@@ -21,11 +21,18 @@ const logger = require("../utils/logger.cjs");
 const aiIdentity = require("../config/ai-identity.cjs");
 const settings = require("../config/settings.cjs");
 const memory = require("../utils/memory.cjs");
+const {
+  assertProviderBaseUrl,
+  assertProviderRequestUrl,
+} = require("../utils/providerUrl.cjs");
 const { toolDeclarations, describeCall, runTool } = require("./aiTools.cjs");
 const {
-  runProviderAgent,
-  activeProviderKeySetting,
-} = require("./aiProviders.cjs");
+  BaseAIProvider,
+  registerProvider,
+  getProvider,
+  detectModelCapabilities,
+} = require("./aiRouter.cjs");
+const { runProviderAgent, activeProviderKeySetting } = require("./aiProviders.cjs");
 
 // Model, key, tool budget and history length all come from config/settings.cjs
 // (what the dashboard saved, else the default) and are read per call, so changing
@@ -45,7 +52,9 @@ let clientCache = { key: null, baseUrl: null, client: null };
 function geminiClient() {
   const key = settings.get("gemini_api_key");
   if (!key) return null;
-  const baseUrl = String(settings.get("gemini_base_url")).replace(/\/+$/, "");
+  const baseUrl = assertProviderBaseUrl(settings.get("gemini_base_url"), {
+    allowLoopback: true,
+  });
   if (clientCache.key !== key || clientCache.baseUrl !== baseUrl) {
     clientCache = {
       key,
@@ -100,7 +109,12 @@ function loadPersona() {
     const raw = fs.readFileSync(PERSONA_FILE, "utf8").trim();
     // Everything above the `---` separator is a note to the human editing the
     // file, not part of the prompt.
-    const body = raw.includes("\n---") ? raw.split(/\n---\s*\n/).slice(1).join("\n---\n") : raw;
+    const body = raw.includes("\n---")
+      ? raw
+          .split(/\n---\s*\n/)
+          .slice(1)
+          .join("\n---\n")
+      : raw;
     personaCache = { text: body.trim() || DEFAULT_PERSONA, stamp };
     return personaCache.text;
   } catch {
@@ -120,6 +134,17 @@ function nowLabel() {
   }
 }
 
+function botLangDirective() {
+  const botLang = settings.get("bot_language") || "auto";
+  if (botLang === "ar") {
+    return "- Language instruction: Always reply in Modern Standard Arabic (العربية الفصحى), clear, professional, and well-phrased.";
+  }
+  if (botLang === "en") {
+    return "- Language instruction: Always reply in English, clear, professional, natural, and direct.";
+  }
+  return "- Language instruction: Automatically reply in the language the user writes in (Modern Standard Arabic if Arabic, English if English).";
+}
+
 /**
  * product identity + persona + who/where/when + long-term memory.
  *
@@ -129,15 +154,7 @@ function nowLabel() {
  * src/config/ai-identity.cjs) and is never rendered by the panel.
  */
 function buildSystemInstruction(context = {}) {
-  const {
-    chatId,
-    senderName,
-    senderId,
-    isOwner,
-    isAdmin,
-    isGroup,
-    chatName,
-  } = context;
+  const { chatId, senderName, senderId, isOwner, isAdmin, isGroup, chatName } = context;
 
   const runtime = [
     "# Current context",
@@ -148,13 +165,12 @@ function buildSystemInstruction(context = {}) {
     isGroup
       ? "- This is a group: every participant sees your reply. Say nothing here that belongs to one person."
       : "- This is a private chat: only this person sees your reply.",
+    botLangDirective(),
   ].join("\n");
 
   const memoryBlock = memory.buildMemoryContext(chatId);
 
-  return [aiIdentity.systemBlock, loadPersona(), runtime, memoryBlock]
-    .filter(Boolean)
-    .join("\n\n");
+  return [aiIdentity.systemBlock, loadPersona(), runtime, memoryBlock].filter(Boolean).join("\n\n");
 }
 
 /** A turn the conversation can legally start from. */
@@ -188,11 +204,7 @@ function trimHistory(history) {
 function isFileReferenceError(error) {
   if (!error) return false;
   const status =
-    error.status ||
-    error.statusCode ||
-    error.code ||
-    error.response?.status ||
-    error.error?.code;
+    error.status || error.statusCode || error.code || error.response?.status || error.error?.code;
   const message = `${error.message || ""} ${error.details || ""}`.toLowerCase();
 
   if ([403, 404, 410].includes(Number(status))) return true;
@@ -202,7 +214,7 @@ function isFileReferenceError(error) {
   return (
     message.includes("file") &&
     /not found|not exist|expired|not in an active|not active|permission to access|forbidden|file_?uri/.test(
-      message
+      message,
     )
   );
 }
@@ -227,7 +239,7 @@ function withTimeout(promise, ms, label) {
     new Promise((_, reject) => {
       timer = setTimeout(
         () => reject(new Error(`${label} تأخرت أكتر من ${Math.round(ms / 1000)} ثانية`)),
-        ms
+        ms,
       );
     }),
   ]).finally(() => clearTimeout(timer));
@@ -391,33 +403,109 @@ function formatSources(sources) {
  * @param {boolean}  [options.useTools=true]
  * @returns {Promise<{text: string, history: Array, toolCalls: Array, steps: number}>}
  */
-async function runAgent({
+class GeminiProvider extends BaseAIProvider {
+  constructor() {
+    super({
+      id: "gemini",
+      label: "Google Gemini",
+      keySetting: "gemini_api_key",
+      modelSetting: "gemini_model",
+      baseUrlSetting: "gemini_base_url",
+      defaultBaseUrl: "",
+      supportsFiles: true,
+    });
+  }
+
+  async prepareMedia(
+    parts,
+    mediaMessage,
+    mimeOverride,
+    { downloadContentFromMessage, genAI, tempDir } = {},
+  ) {
+    if (!genAI) throw new Error("Gemini client unavailable");
+    const mime = mimeOverride || mediaMessage?.mimetype || "";
+    const caps = detectModelCapabilities("gemini");
+    if (mime.startsWith("image/") && (!settings.get("ai_vision_enabled") || !caps.supportsVision)) {
+      parts.push({
+        text: "[تم إرفاق صورة — الرؤية غير مفعّلة للنموذج المحدد / Vision is disabled for the selected model]",
+      });
+      return null;
+    }
+    if (mime.startsWith("audio/") && (!settings.get("ai_stt_enabled") || !caps.supportsAudioStt)) {
+      parts.push({
+        text: "[تم إرفاق صوت — تحويل الكلام غير مفعّل للنموذج المحدد / Speech-to-text is disabled for the selected model]",
+      });
+      return null;
+    }
+    const tempFilePath = path.join(tempDir || __dirname, `temp_media_${Date.now()}`);
+    const stream = await downloadContentFromMessage(
+      mediaMessage,
+      mediaMessage.mimetype?.startsWith("image/")
+        ? "image"
+        : mediaMessage.mimetype?.startsWith("video/")
+          ? "video"
+          : mediaMessage.mimetype?.startsWith("audio/")
+            ? "audio"
+            : "document",
+    );
+    let buffer = Buffer.from([]);
+    for await (const chunk of stream) buffer = Buffer.concat([buffer, chunk]);
+    if (!buffer.length) throw new Error("Empty media buffer");
+    await fs.promises.writeFile(tempFilePath, buffer);
+
+    try {
+      const uploadResult = await genAI.files.upload({
+        file: tempFilePath,
+        mimeType: mediaMessage.mimetype,
+      });
+      parts.push({
+        fileData: {
+          fileUri: uploadResult.uri,
+          mimeType: uploadResult.mimeType,
+        },
+      });
+      return uploadResult.uri;
+    } finally {
+      await fs.promises.unlink(tempFilePath).catch(() => {});
+    }
+  }
+
+  async runTurn(options) {
+    return runGeminiTurn(options);
+  }
+}
+
+registerProvider("gemini", new GeminiProvider());
+
+/**
+ * Run one agent turn.
+ * Modular dispatch through getProvider().
+ */
+async function runAgent(options = {}) {
+  const provider = getProvider();
+  const result = await provider.runTurn({
+    ...options,
+    history: trimHistory(options.history),
+    systemInstruction: buildSystemInstruction(options.context),
+  });
+  return { ...result, history: trimHistory(result.history) };
+}
+
+async function runGeminiTurn({
   parts,
   history = [],
   status = null,
   context = {},
   useTools = true,
   maxSteps = null,
+  systemInstruction = null,
 } = {}) {
-  const provider = settings.get("ai_provider");
-  if (provider !== "gemini") {
-    const result = await runProviderAgent(provider, {
-      parts,
-      history: trimHistory(history),
-      systemInstruction: buildSystemInstruction(context),
-      status,
-      context,
-      useTools,
-      maxSteps,
-    });
-    return { ...result, history: trimHistory(result.history) };
-  }
-
   const genAI = geminiClient();
   if (!genAI) throw new Error("GEMINI_API_KEY غير معرف");
+  await assertProviderRequestUrl(settings.get("gemini_base_url"), { allowLoopback: true });
 
   const stepBudget = maxSteps ?? settings.get("ai_max_tool_steps");
-  const systemInstruction = buildSystemInstruction(context);
+  const instruction = systemInstruction || buildSystemInstruction(context);
   const trimmed = trimHistory(history);
   const model = settings.get("gemini_model");
 
@@ -435,7 +523,7 @@ async function runAgent({
       model,
       history: trimmed,
       config: {
-        systemInstruction,
+        systemInstruction: instruction,
         ...(tools.length ? { tools } : {}),
         ...(toolConfig ? { toolConfig } : {}),
       },
@@ -459,7 +547,7 @@ async function runAgent({
     if (!searchOffered || !isToolCombinationError(error)) throw error;
     logger.warn(
       { err: error?.message, model },
-      "[aiAgent] this model refused Google Search together with function tools — retrying with functions only"
+      "[aiAgent] this model refused Google Search together with function tools — retrying with functions only",
     );
     searchOffered = false;
     chat = openChat(false);
@@ -486,10 +574,7 @@ async function runAgent({
     const responses = [];
     for (const call of calls) {
       toolCalls.push({ name: call.name, args: call.args });
-      logger.info(
-        { tool: call.name, chatId: context.chatId },
-        "[aiAgent] tool call"
-      );
+      logger.info({ tool: call.name, chatId: context.chatId }, "[aiAgent] tool call");
 
       // Named apart from the outer `response` on purpose: that one is the
       // model's reply and is reassigned at the bottom of this loop.
@@ -498,7 +583,7 @@ async function runAgent({
         toolResult = await withTimeout(
           runTool(call.name, call.args, context),
           settings.get("ai_tool_timeout_ms"),
-          call.name
+          call.name,
         );
       } catch (err) {
         toolResult = { error: err?.message || String(err) };
@@ -532,8 +617,7 @@ async function runAgent({
   }
 
   if (!text && steps >= stepBudget) {
-    text =
-      "شغّلت الأدوات المتاحة بس مقدرتش أوصل لإجابة نهائية. جرّب تسأل بصيغة أوضح.";
+    text = "شغّلت الأدوات المتاحة بس مقدرتش أوصل لإجابة نهائية. جرّب تسأل بصيغة أوضح.";
   }
 
   let newHistory = [];
@@ -574,4 +658,7 @@ module.exports = {
   sanitizeHistoryForFiles,
   trimHistory,
   PERSONA_FILE,
+  detectModelCapabilities,
+  getProvider,
+  GeminiProvider,
 };

package/src/services/aiProviders.cjs

@@ -31,6 +31,7 @@
 const logger = require("../utils/logger.cjs");
 const settings = require("../config/settings.cjs");
 const { toolDeclarations, describeCall, runTool } = require("./aiTools.cjs");
+const { assertProviderBaseUrl, safeProviderFetch } = require("../utils/providerUrl.cjs");
 
 // Shown in place of a media part, on the wire and in the canonical history.
 const MEDIA_NOTE = "[تم إرفاق ملف/وسائط في الرسالة]";
@@ -49,9 +50,7 @@ const ADAPTERS = {
     baseUrlSetting: "openai_base_url",
     defaultBaseUrl: "https://api.openai.com/v1",
     tools: openaiTools,
-    systemMessage: (systemInstruction) => [
-      { role: "system", content: systemInstruction },
-    ],
+    systemMessage: (systemInstruction) => [{ role: "system", content: systemInstruction }],
     userMessage: (text) => ({ role: "user", content: text }),
     /** Canonical history -> chat-completions messages. */
     historyToMessages: historyToOpenAI,
@@ -97,6 +96,110 @@ const ADAPTERS = {
   },
 };
 
+const {
+  BaseAIProvider,
+  registerProvider,
+  getProvider,
+  detectModelCapabilities,
+} = require("./aiRouter.cjs");
+
+function visionAllowed(providerId) {
+  const id = providerId || settings.get("ai_provider");
+  return Boolean(settings.get("ai_vision_enabled") && detectModelCapabilities(id).supportsVision);
+}
+
+async function prepareInlineImageMedia(
+  providerId,
+  parts,
+  mediaMessage,
+  mimeOverride,
+  downloadContentFromMessage,
+) {
+  const isVisionOn = visionAllowed(providerId);
+  const mime = mimeOverride || mediaMessage?.mimetype || "ملف";
+  if (isVisionOn && mime.startsWith("image/") && typeof downloadContentFromMessage === "function") {
+    try {
+      const stream = await downloadContentFromMessage(mediaMessage, "image");
+      let buffer = Buffer.from([]);
+      for await (const chunk of stream) buffer = Buffer.concat([buffer, chunk]);
+      if (buffer.length) {
+        parts.push({
+          inlineData: {
+            mimeType: mime,
+            data: buffer.toString("base64"),
+          },
+        });
+        return "inline:image";
+      }
+    } catch (err) {
+      logger.warn({ err }, `[${providerId}] failed to extract inline image for vision`);
+    }
+  }
+  parts.push({
+    text: `[تم إرفاق وسائط (${mime}) — المزود الحالي لا يستطيع قراءة هذا النوع من الوسائط أو تم تعطيل ميزة تحليل الصور]`,
+  });
+  return null;
+}
+
+class OpenAIProvider extends BaseAIProvider {
+  constructor() {
+    super({
+      id: "openai",
+      label: "OpenAI-compatible",
+      keySetting: "openai_api_key",
+      modelSetting: "openai_model",
+      baseUrlSetting: "openai_base_url",
+      defaultBaseUrl: "https://api.openai.com/v1",
+      supportsFiles: false,
+    });
+  }
+
+  async prepareMedia(parts, mediaMessage, mimeOverride, { downloadContentFromMessage } = {}) {
+    return prepareInlineImageMedia(
+      "openai",
+      parts,
+      mediaMessage,
+      mimeOverride,
+      downloadContentFromMessage,
+    );
+  }
+
+  async runTurn(options) {
+    return runProviderAgent("openai", options);
+  }
+}
+
+class AnthropicProvider extends BaseAIProvider {
+  constructor() {
+    super({
+      id: "anthropic",
+      label: "Anthropic",
+      keySetting: "anthropic_api_key",
+      modelSetting: "anthropic_model",
+      baseUrlSetting: "anthropic_base_url",
+      defaultBaseUrl: "https://api.anthropic.com",
+      supportsFiles: false,
+    });
+  }
+
+  async prepareMedia(parts, mediaMessage, mimeOverride, { downloadContentFromMessage } = {}) {
+    return prepareInlineImageMedia(
+      "anthropic",
+      parts,
+      mediaMessage,
+      mimeOverride,
+      downloadContentFromMessage,
+    );
+  }
+
+  async runTurn(options) {
+    return runProviderAgent("anthropic", options);
+  }
+}
+
+registerProvider("openai", new OpenAIProvider());
+registerProvider("anthropic", new AnthropicProvider());
+
 function adapterFor(provider) {
   const adapter = ADAPTERS[provider];
   if (!adapter) throw new Error(`مزود غير معروف: ${provider}`);
@@ -160,15 +263,7 @@ function toJsonSchema(schema) {
     } else if (key === "anyOf") {
       out.anyOf = toJsonSchema(value);
     } else if (
-      [
-        "description",
-        "enum",
-        "required",
-        "format",
-        "minimum",
-        "maximum",
-        "default",
-      ].includes(key)
+      ["description", "enum", "required", "format", "minimum", "maximum", "default"].includes(key)
     ) {
       out[key] = value;
     }
@@ -204,11 +299,17 @@ function anthropicTools() {
 // ===========================================================================
 
 function partsText(parts) {
+  const isVisionOn = visionAllowed();
   return (parts || [])
     .map((part) => {
       if (part?.text) return part.text;
-      // Media on this path is a note, never bytes: there is no upload API to
-      // send it through (see the header comment).
+      if (
+        isVisionOn &&
+        part?.inlineData?.mimeType?.startsWith("image/") &&
+        part?.inlineData?.data
+      ) {
+        return "";
+      }
       if (part?.fileData || part?.inlineData) return MEDIA_NOTE;
       return "";
     })
@@ -242,9 +343,7 @@ function historyToOpenAI(history) {
   (history || []).forEach((turn, turnIndex) => {
     if (!turn || !Array.isArray(turn.parts)) return;
     const role = turn.role === "model" ? "assistant" : "user";
-    const calls = turn.parts
-      .filter((part) => part?.functionCall)
-      .map((part) => part.functionCall);
+    const calls = turn.parts.filter((part) => part?.functionCall).map((part) => part.functionCall);
     const responses = turn.parts
       .filter((part) => part?.functionResponse)
       .map((part) => part.functionResponse);
@@ -268,6 +367,29 @@ function historyToOpenAI(history) {
       return;
     }
 
+    const isVisionOn = visionAllowed("openai");
+    const imageParts =
+      isVisionOn && role === "user"
+        ? turn.parts.filter(
+            (p) => p?.inlineData?.mimeType?.startsWith("image/") && p?.inlineData?.data,
+          )
+        : [];
+
+    let userContent = text;
+    if (imageParts.length) {
+      const items = [];
+      if (text) items.push({ type: "text", text });
+      for (const img of imageParts) {
+        items.push({
+          type: "image_url",
+          image_url: {
+            url: `data:${img.inlineData.mimeType};base64,${img.inlineData.data}`,
+          },
+        });
+      }
+      userContent = items;
+    }
+
     if (responses.length) {
       for (const response of responses) {
         messages.push({
@@ -276,11 +398,11 @@ function historyToOpenAI(history) {
           content: JSON.stringify(response.response ?? {}),
         });
       }
-      if (text) messages.push({ role: "user", content: text });
+      if (userContent) messages.push({ role: "user", content: userContent });
       return;
     }
 
-    if (text) messages.push({ role, content: text });
+    if (userContent) messages.push({ role, content: userContent });
   });
 
   return messages;
@@ -330,7 +452,23 @@ function historyToAnthropic(history) {
           content: JSON.stringify(part.functionResponse.response ?? {}),
         });
       } else if (part?.fileData || part?.inlineData) {
-        blocks.push({ type: "text", text: MEDIA_NOTE });
+        const isVisionOn = visionAllowed("anthropic");
+        if (
+          isVisionOn &&
+          part?.inlineData?.mimeType?.startsWith("image/") &&
+          part?.inlineData?.data
+        ) {
+          blocks.push({
+            type: "image",
+            source: {
+              type: "base64",
+              media_type: part.inlineData.mimeType,
+              data: part.inlineData.data,
+            },
+          });
+        } else {
+          blocks.push({ type: "text", text: MEDIA_NOTE });
+        }
       }
     }
 
@@ -352,8 +490,8 @@ function parseOpenAIResponse(payload) {
   const content = Array.isArray(message.content)
     ? message.content.map((block) => block?.text || "").join("\n")
     : typeof message.content === "string"
-    ? message.content
-    : "";
+      ? message.content
+      : "";
 
   const calls = (message.tool_calls || [])
     .map((toolCall, i) => ({
@@ -377,7 +515,7 @@ function parseOpenAIResponse(payload) {
 
 function parseAnthropicResponse(payload) {
   const blocks = (payload?.content || []).filter(
-    (block) => block?.type === "text" || block?.type === "tool_use"
+    (block) => block?.type === "text" || block?.type === "tool_use",
   );
 
   const calls = blocks
@@ -411,17 +549,17 @@ async function callApi({ url, headers, body, label }) {
 
   let response;
   try {
-    response = await fetch(url, {
+    response = await safeProviderFetch(url, {
       method: "POST",
       headers: { "Content-Type": "application/json", ...headers },
       body: JSON.stringify(body),
       signal: controller.signal,
-    });
+    }, { allowLoopback: true });
   } catch (err) {
     throw new Error(
       err?.name === "AbortError"
         ? `${label} API request timed out after ${API_TIMEOUT_MS / 1000}s`
-        : `${label} API request failed: ${err?.message || err}`
+        : `${label} API request failed: ${err?.message || err}`,
     );
   } finally {
     clearTimeout(timer);
@@ -445,7 +583,11 @@ async function callApi({ url, headers, body, label }) {
   return payload;
 }
 
-function requestFor(adapter, provider, { baseUrl, apiKey, model, systemInstruction, messages, tools }) {
+function requestFor(
+  adapter,
+  provider,
+  { baseUrl, apiKey, model, systemInstruction, messages, tools },
+) {
   if (provider === "anthropic") {
     return {
       url: `${baseUrl}/v1/messages`,
@@ -481,8 +623,8 @@ function withTimeout(promise, ms, label) {
     promise,
     new Promise((_, reject) => {
       timer = setTimeout(
-        () => reject(new Error(`${label} تأخرت أكتر من ${Math.round(ms / 1000)} ثانية`)),
-        ms
+        () => reject(new Error(`${label} استغرقت العملية أكثر من ${Math.round(ms / 1000)} ثانية`)),
+        ms,
       );
     }),
   ]).finally(() => clearTimeout(timer));
@@ -505,15 +647,18 @@ function withTimeout(promise, ms, label) {
  * @param {boolean} [options.useTools=true]
  * @returns {Promise<{text: string, history: Array, toolCalls: Array, steps: number, sources: Array, searchOffered: boolean}>}
  */
-async function runProviderAgent(provider, {
-  parts,
-  systemInstruction,
-  history = [],
-  status = null,
-  context = {},
-  useTools = true,
-  maxSteps = null,
-} = {}) {
+async function runProviderAgent(
+  provider,
+  {
+    parts,
+    systemInstruction,
+    history = [],
+    status = null,
+    context = {},
+    useTools = true,
+    maxSteps = null,
+  } = {},
+) {
   const adapter = adapterFor(provider);
 
   const apiKey = settings.get(adapter.keySetting);
@@ -521,9 +666,10 @@ async function runProviderAgent(provider, {
 
   const stepBudget = maxSteps ?? settings.get("ai_max_tool_steps");
   const model = settings.get(adapter.modelSetting);
-  const baseUrl = String(
-    settings.get(adapter.baseUrlSetting) || adapter.defaultBaseUrl
-  ).replace(/\/+$/, "");
+  const baseUrl = assertProviderBaseUrl(
+    settings.get(adapter.baseUrlSetting) || adapter.defaultBaseUrl,
+    { allowLoopback: true },
+  );
 
   const turnText = partsText(parts);
 
@@ -585,8 +731,7 @@ async function runProviderAgent(provider, {
       // Same contract as the Gemini loop: the budget answer, not a lie.
       return finish({
         text:
-          parsed.text ||
-          "شغّلت الأدوات المتاحة بس مقدرتش أوصل لإجابة نهائية. جرّب تسأل بصيغة أوضح.",
+          parsed.text || "شغّلت الأدوات المتاحة بس مقدرتش أوصل لإجابة نهائية. جرّب تسأل بصيغة أوضح.",
         canonical,
         toolCalls,
         steps,
@@ -596,9 +741,7 @@ async function runProviderAgent(provider, {
     steps += 1;
 
     if (status) {
-      const line = parsed.calls
-        .map((call) => describeCall(call.name, call.args))
-        .join("\n");
+      const line = parsed.calls.map((call) => describeCall(call.name, call.args)).join("\n");
       await status.update(line);
     }
 
@@ -612,7 +755,7 @@ async function runProviderAgent(provider, {
         toolResult = await withTimeout(
           runTool(call.name, call.args, context),
           settings.get("ai_tool_timeout_ms"),
-          call.name
+          call.name,
         );
       } catch (err) {
         toolResult = { error: err?.message || String(err) };
@@ -663,4 +806,7 @@ module.exports = {
   parseOpenAIResponse,
   parseAnthropicResponse,
   ADAPTERS,
+  getProvider,
+  OpenAIProvider,
+  AnthropicProvider,
 };