@automagik/omni 2.260916.12 → 2.260921.2

package/dist/index.js

@@ -137424,7 +137424,7 @@ import { fileURLToPath } from "url";
 // package.json
 var package_default = {
   name: "@automagik/omni",
-  version: "2.260916.12",
+  version: "2.260921.2",
   repository: {
     type: "git",
     url: "git+https://github.com/automagik-dev/omni.git",

package/dist/server/index.js

@@ -251115,7 +251115,7 @@ var init_sentry_scrub = __esm(() => {
 var require_package7 = __commonJS((exports, module) => {
   module.exports = {
     name: "@omni/api",
-    version: "2.260916.12",
+    version: "2.260921.2",
     private: true,
     type: "module",
     exports: {
@@ -339767,7 +339767,7 @@ var init_client7 = __esm(() => {
     IMAGE_GEN: "gemini-3.1-flash-image",
     VIDEO_GEN: "veo-3.1-generate-preview",
     VISION: "gemini-3.1-flash-lite",
-    STT: "gemini-3.1-flash-lite"
+    STT: "gemini-3.5-flash-lite"
   };
 });
 
@@ -339987,7 +339987,7 @@ function extractAudio(response) {
   }
   return null;
 }
-var log68, DEFAULT_PRO_MODEL = "lyria-3-pro-preview", DEFAULT_CLIP_MODEL = "lyria-3-clip-preview";
+var log68, DEFAULT_PRO_MODEL = "lyria-3.5", DEFAULT_CLIP_MODEL = "lyria-3-clip-preview";
 var init_musicgen = __esm(() => {
   init_src();
   init_client7();
@@ -340604,7 +340604,7 @@ ${base2}`;
     return base2;
   }
 }
-var log72, DEFAULT_VISION_MODEL = "gemini-3.1-flash-lite-preview", DEFAULT_PROMPT2 = "Describe this image in clear, concise detail. Mention key subjects, actions, colors, setting, and any legible text.";
+var log72, DEFAULT_VISION_MODEL = "gemini-3.5-flash-lite", DEFAULT_PROMPT2 = "Describe this image in clear, concise detail. Mention key subjects, actions, colors, setting, and any legible text.";
 var init_vision2 = __esm(() => {
   init_src();
   init_client7();
@@ -342731,7 +342731,7 @@ new Groq({ apiKey, dangerouslyAllowBrowser: true })`);
 });
 
 // ../media-processing/src/models.ts
-var GEMINI_MODEL = "gemini-3-flash-preview", OPENAI_VISION_MODEL = "gpt-4o-mini", OPENAI_AUDIO_CHAT_MODEL = "gpt-audio-mini", OPENAI_TRANSCRIBE_MODEL = "gpt-4o-transcribe", GEMINI_AUDIO_MODEL = "gemini-3.1-flash-lite", GROQ_WHISPER_MODEL = "whisper-large-v3-turbo";
+var GEMINI_MODEL = "gemini-3.5-flash-lite", OPENAI_VISION_MODEL = "gpt-5.6-luna", OPENAI_AUDIO_CHAT_MODEL = "gpt-audio-mini", OPENAI_TRANSCRIBE_MODEL = "gpt-transcribe", GEMINI_AUDIO_MODEL = "gemini-3.5-flash-lite", GROQ_WHISPER_MODEL = "whisper-large-v3-turbo";
 var init_models2 = () => {};
 
 // ../media-processing/src/pricing.ts
@@ -342813,6 +342813,12 @@ var init_pricing = __esm(() => {
       unit: "per_minute",
       inputRate: 0.006
     },
+    "openai_whisper:gpt-transcribe": {
+      model: "gpt-transcribe",
+      provider: "openai",
+      unit: "per_minute",
+      inputRate: 0.006
+    },
     "openai_vision:gpt-4o-mini": {
       model: "gpt-4o-mini",
       provider: "openai",
@@ -342827,6 +342833,20 @@ var init_pricing = __esm(() => {
       inputRate: 2.5,
       outputRate: 10
     },
+    "openai_vision:gpt-5.6-luna": {
+      model: "gpt-5.6-luna",
+      provider: "openai",
+      unit: "per_million_tokens",
+      inputRate: 0.2,
+      outputRate: 1.2
+    },
+    "gemini_vision:gemini-3.5-flash-lite": {
+      model: "gemini-3.5-flash-lite",
+      provider: "google",
+      unit: "per_million_tokens",
+      inputRate: 0.15,
+      outputRate: 0.6
+    },
     "gemini_vision:gemini-3-flash-preview": {
       model: "gemini-3-flash-preview",
       provider: "google",
@@ -342834,6 +342854,13 @@ var init_pricing = __esm(() => {
       inputRate: 0.15,
       outputRate: 0.6
     },
+    "gemini_video:gemini-3.5-flash-lite": {
+      model: "gemini-3.5-flash-lite",
+      provider: "google",
+      unit: "per_million_tokens",
+      inputRate: 0.15,
+      outputRate: 0.6
+    },
     "gemini_video:gemini-3-flash-preview": {
       model: "gemini-3-flash-preview",
       provider: "google",
@@ -350024,10 +350051,16 @@ var init_image = __esm(() => {
     geminiClient = null;
     geminiModel = null;
     openaiClient = null;
+    get geminiModelId() {
+      return this.config.geminiVisionModel ?? GEMINI_MODEL;
+    }
+    get openaiVisionModelId() {
+      return this.config.openaiVisionModel ?? OPENAI_VISION_MODEL;
+    }
     getGeminiModel() {
       if (!this.geminiModel && this.config.geminiApiKey) {
         this.geminiClient = new GoogleGenerativeAI(this.config.geminiApiKey);
-        this.geminiModel = this.geminiClient.getGenerativeModel({ model: GEMINI_MODEL });
+        this.geminiModel = this.geminiClient.getGenerativeModel({ model: this.geminiModelId });
         this.log.info("Gemini model initialized");
       }
       return this.geminiModel;
@@ -350070,7 +350103,7 @@ Additional context (caption): ${options.caption}` : basePrompt;
     async describeWithGemini(imageData, mimeType, prompt) {
       const model = this.getGeminiModel();
       if (!model) {
-        return this.createFailedResult("Gemini not configured (missing API key)", "google", GEMINI_MODEL);
+        return this.createFailedResult("Gemini not configured (missing API key)", "google", this.geminiModelId);
       }
       const timeouts = getMediaTimeouts();
       const runGemini = async (timeoutMs) => this.executeWithResilience("gemini", async () => {
@@ -350108,7 +350141,7 @@ Additional context (caption): ${options.caption}` : basePrompt;
           }
         }
         const { text: text3, inputTokens, outputTokens } = callResult;
-        const costCents = calculateCost("gemini_vision", GEMINI_MODEL, {
+        const costCents = calculateCost("gemini_vision", this.geminiModelId, {
           inputTokens,
           outputTokens
         });
@@ -350118,7 +350151,7 @@ Additional context (caption): ${options.caption}` : basePrompt;
           contentFormat: "text",
           processingType: "description",
           provider: "google",
-          model: GEMINI_MODEL,
+          model: this.geminiModelId,
           processingTimeMs: 0,
           inputTokens,
           outputTokens,
@@ -350128,13 +350161,13 @@ Additional context (caption): ${options.caption}` : basePrompt;
         const errorMsg = error3 instanceof Error ? error3.message : String(error3);
         const isCircuit = this.isCircuitOpen(error3);
         this.log.error("Gemini description error", { error: errorMsg, circuitOpen: isCircuit });
-        return this.createFailedResult(errorMsg, "google", GEMINI_MODEL);
+        return this.createFailedResult(errorMsg, "google", this.geminiModelId);
       }
     }
     async describeWithOpenAI(imageData, mimeType, prompt) {
       const client = this.getOpenAIClient();
       if (!client) {
-        return this.createFailedResult("OpenAI client not configured (missing API key)", "openai", OPENAI_VISION_MODEL);
+        return this.createFailedResult("OpenAI client not configured (missing API key)", "openai", this.openaiVisionModelId);
       }
       const timeouts = getMediaTimeouts();
       try {
@@ -350142,7 +350175,7 @@ Additional context (caption): ${options.caption}` : basePrompt;
           const base64Data = imageData.toString("base64");
           const dataUrl = `data:${mimeType};base64,${base64Data}`;
           const response = await client.chat.completions.create({
-            model: OPENAI_VISION_MODEL,
+            model: this.openaiVisionModelId,
             messages: [
               {
                 role: "user",
@@ -350166,7 +350199,7 @@ Additional context (caption): ${options.caption}` : basePrompt;
             outputTokens: response.usage?.completion_tokens ?? 0
           };
         }, { timeoutMs: timeouts.imageTimeoutMs });
-        const costCents = calculateCost("openai_vision", OPENAI_VISION_MODEL, {
+        const costCents = calculateCost("openai_vision", this.openaiVisionModelId, {
           inputTokens,
           outputTokens
         });
@@ -350176,7 +350209,7 @@ Additional context (caption): ${options.caption}` : basePrompt;
           contentFormat: "text",
           processingType: "description",
           provider: "openai",
-          model: OPENAI_VISION_MODEL,
+          model: this.openaiVisionModelId,
           processingTimeMs: 0,
           inputTokens,
           outputTokens,
@@ -350185,7 +350218,7 @@ Additional context (caption): ${options.caption}` : basePrompt;
       } catch (error3) {
         const errorMsg = error3 instanceof Error ? error3.message : String(error3);
         this.log.error("OpenAI description error", { error: errorMsg });
-        return this.createFailedResult(errorMsg, "openai", OPENAI_VISION_MODEL);
+        return this.createFailedResult(errorMsg, "openai", this.openaiVisionModelId);
       }
     }
     normalizeGeminiMimeType(mimeType) {
@@ -350298,10 +350331,13 @@ var init_video = __esm(() => {
     ];
     geminiClient = null;
     geminiModel = null;
+    get geminiModelId() {
+      return this.config.geminiVisionModel ?? GEMINI_MODEL;
+    }
     getGeminiModel() {
       if (!this.geminiModel && this.config.geminiApiKey) {
         this.geminiClient = new GoogleGenerativeAI(this.config.geminiApiKey);
-        this.geminiModel = this.geminiClient.getGenerativeModel({ model: GEMINI_MODEL });
+        this.geminiModel = this.geminiClient.getGenerativeModel({ model: this.geminiModelId });
         this.log.info("Gemini model initialized for video");
       }
       return this.geminiModel;
@@ -350344,7 +350380,7 @@ Additional context: ${options.caption}` : basePrompt;
     async describeWithGemini(videoData, mimeType, prompt) {
       const model = this.getGeminiModel();
       if (!model) {
-        return this.createFailedResult("Gemini not configured (missing API key)", "google", GEMINI_MODEL);
+        return this.createFailedResult("Gemini not configured (missing API key)", "google", this.geminiModelId);
       }
       const timeouts = getMediaTimeouts();
       try {
@@ -350366,7 +350402,7 @@ Additional context: ${options.caption}` : basePrompt;
             outputTokens: usageMetadata?.candidatesTokenCount ?? 0
           };
         }, { timeoutMs: timeouts.videoTimeoutMs });
-        const costCents = calculateCost("gemini_video", GEMINI_MODEL, {
+        const costCents = calculateCost("gemini_video", this.geminiModelId, {
           inputTokens,
           outputTokens
         });
@@ -350376,7 +350412,7 @@ Additional context: ${options.caption}` : basePrompt;
           contentFormat: "text",
           processingType: "description",
           provider: "google",
-          model: GEMINI_MODEL,
+          model: this.geminiModelId,
           processingTimeMs: 0,
           inputTokens,
           outputTokens,
@@ -350386,7 +350422,7 @@ Additional context: ${options.caption}` : basePrompt;
         const errorMsg = error3 instanceof Error ? error3.message : String(error3);
         const isCircuit = this.isCircuitOpen(error3);
         this.log.error("Gemini video description error", { error: errorMsg, circuitOpen: isCircuit });
-        return this.createFailedResult(errorMsg, "google", GEMINI_MODEL);
+        return this.createFailedResult(errorMsg, "google", this.geminiModelId);
       }
     }
     normalizeVideoMimeType(mimeType) {
@@ -359419,10 +359455,13 @@ var init_document = __esm(() => {
     ];
     geminiClient = null;
     geminiModel = null;
+    get geminiModelId() {
+      return this.config.geminiVisionModel ?? GEMINI_MODEL;
+    }
     getGeminiModel() {
       if (!this.geminiModel && this.config.geminiApiKey) {
         this.geminiClient = new GoogleGenerativeAI(this.config.geminiApiKey);
-        this.geminiModel = this.geminiClient.getGenerativeModel({ model: GEMINI_MODEL });
+        this.geminiModel = this.geminiClient.getGenerativeModel({ model: this.geminiModelId });
         this.log.info("Gemini model initialized for document OCR");
       }
       return this.geminiModel;
@@ -359845,7 +359884,7 @@ ${rows.join(`
             outputTokens: usageMetadata?.candidatesTokenCount ?? 0
           };
         }, { timeoutMs: timeouts.documentTimeoutMs });
-        const costCents = calculateCost("gemini_vision", GEMINI_MODEL, {
+        const costCents = calculateCost("gemini_vision", this.geminiModelId, {
           inputTokens,
           outputTokens
         });
@@ -359855,7 +359894,7 @@ ${rows.join(`
           contentFormat: "markdown",
           processingType: "extraction",
           provider: "google",
-          model: GEMINI_MODEL,
+          model: this.geminiModelId,
           processingTimeMs: 0,
           inputTokens,
           outputTokens,
@@ -359864,7 +359903,7 @@ ${rows.join(`
       } catch (error3) {
         const errorMsg = error3 instanceof Error ? error3.message : String(error3);
         this.log.error("Gemini OCR failed", { error: errorMsg });
-        return this.createFailedResult(errorMsg, "google", GEMINI_MODEL);
+        return this.createFailedResult(errorMsg, "google", this.geminiModelId);
       }
     }
     createFailedResult(errorMessage, provider, model) {
@@ -362225,6 +362264,73 @@ function buildWhatsAppMessageContext(rawPayload, chatId, instance4, text3) {
   return { isDirectMessage, mentionsBot, isReplyToBot, text: text3 };
 }
 
+// ../api/src/services/response-gate.ts
+async function resolveGateProvider(settings) {
+  const configured = await settings.getString("gate.provider", "GATE_PROVIDER");
+  return configured?.trim().toLowerCase() === "openai" ? "openai" : DEFAULT_GATE_PROVIDER;
+}
+async function resolveGateModel(provider, instanceModel, settings) {
+  if (instanceModel)
+    return instanceModel;
+  if (provider === "openai") {
+    return await settings.getString("gate.openai.model", "GATE_OPENAI_MODEL") ?? DEFAULT_GATE_MODEL_OPENAI;
+  }
+  return await settings.getString("gate.gemini.model", "GATE_GEMINI_MODEL") ?? DEFAULT_GATE_MODEL_GEMINI;
+}
+async function callGeminiGate(prompt, model, settings, signal2) {
+  const apiKey = await settings.getSecret("gemini.api_key", "GEMINI_API_KEY");
+  if (!apiKey)
+    return { ok: false, reason: "missing_key" };
+  const url = `${GEMINI_GENERATE_URL2}/${model}:generateContent?key=${apiKey}`;
+  const res = await fetch(url, {
+    method: "POST",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify({
+      contents: [{ parts: [{ text: prompt }] }],
+      generationConfig: { maxOutputTokens: GATE_MAX_TOKENS, temperature: 0 }
+    }),
+    signal: signal2
+  });
+  if (!res.ok)
+    return { ok: false, reason: "http_error", status: res.status };
+  const data = await res.json();
+  return { ok: true, answer: data.candidates?.[0]?.content?.parts?.[0]?.text?.trim().toLowerCase() ?? "" };
+}
+async function callOpenAiGate(prompt, model, settings, signal2) {
+  const apiKey = await settings.getSecret("openai.api_key", "OPENAI_API_KEY");
+  if (!apiKey)
+    return { ok: false, reason: "missing_key" };
+  const baseUrl = (await settings.getString("gate.openai.base_url", "GATE_OPENAI_BASE_URL") ?? DEFAULT_OPENAI_BASE_URL).trim().replace(/\/+$/, "");
+  const res = await fetch(`${baseUrl}/chat/completions`, {
+    method: "POST",
+    headers: { "Content-Type": "application/json", Authorization: `Bearer ${apiKey}` },
+    body: JSON.stringify({
+      model,
+      messages: [{ role: "user", content: prompt }],
+      max_completion_tokens: GATE_MAX_TOKENS,
+      temperature: 0
+    }),
+    signal: signal2
+  });
+  if (!res.ok)
+    return { ok: false, reason: "http_error", status: res.status };
+  const data = await res.json();
+  return { ok: true, answer: data.choices?.[0]?.message?.content?.trim().toLowerCase() ?? "" };
+}
+async function runGateCall(provider, prompt, model, settings, timeoutMs = GATE_TIMEOUT_MS) {
+  const controller = new AbortController;
+  const timeout = setTimeout(() => controller.abort(), timeoutMs);
+  try {
+    return provider === "openai" ? await callOpenAiGate(prompt, model, settings, controller.signal) : await callGeminiGate(prompt, model, settings, controller.signal);
+  } catch (callError) {
+    const errName = callError.name;
+    return errName === "AbortError" ? { ok: false, reason: "timeout" } : { ok: false, reason: "error", error: String(callError) };
+  } finally {
+    clearTimeout(timeout);
+  }
+}
+var DEFAULT_GATE_PROVIDER = "gemini", DEFAULT_GATE_MODEL_GEMINI = "gemini-3.5-flash-lite", DEFAULT_GATE_MODEL_OPENAI = "gpt-5.6-luna", GATE_TIMEOUT_MS = 3000, GATE_MAX_TOKENS = 256, GEMINI_GENERATE_URL2 = "https://generativelanguage.googleapis.com/v1beta/models", DEFAULT_OPENAI_BASE_URL = "https://api.openai.com/v1";
+
 // ../api/src/services/turn-events.ts
 function getTurnEventsConnection() {
   return nc && !nc.isClosed() ? nc : null;
@@ -367174,7 +367280,8 @@ async function shouldRespondViaGate(instance4, messages4, chatType, settings) {
   if (!inst.agentGateEnabled)
     return true;
   const agentName = instance4.name ?? "assistant";
-  const model = inst.agentGateModel ?? DEFAULT_GATE_MODEL;
+  const provider = await resolveGateProvider(settings);
+  const model = await resolveGateModel(provider, inst.agentGateModel, settings);
   const basePrompt = await resolveGatePrompt(inst.agentGatePrompt, settings);
   const messagesText = messages4.map((m2) => {
     const name = m2.payload.rawPayload?.pushName ?? m2.payload.from ?? "Unknown";
@@ -367185,52 +367292,31 @@ async function shouldRespondViaGate(instance4, messages4, chatType, settings) {
   const traceId = messages4[0]?.metadata.traceId ?? "unknown";
   const startMs = Date.now();
   try {
-    const apiKey = await settings.getSecret("gemini.api_key", "GEMINI_API_KEY");
-    if (!apiKey) {
-      log94.warn("Gate: no Gemini API key, fail-open", { traceId });
-      return true;
-    }
-    const controller = new AbortController;
-    const timeout = setTimeout(() => controller.abort(), GATE_TIMEOUT_MS);
-    try {
-      const url = `https://generativelanguage.googleapis.com/v1beta/models/${model}:generateContent?key=${apiKey}`;
-      const res = await fetch(url, {
-        method: "POST",
-        headers: { "Content-Type": "application/json" },
-        body: JSON.stringify({
-          contents: [{ parts: [{ text: prompt }] }],
-          generationConfig: { maxOutputTokens: 10, temperature: 0 }
-        }),
-        signal: controller.signal
-      });
-      clearTimeout(timeout);
-      if (!res.ok) {
-        log94.warn("Gate: API error, fail-open", { traceId, status: res.status, durationMs: Date.now() - startMs });
-        return true;
-      }
-      const data = await res.json();
-      const answer = data.candidates?.[0]?.content?.parts?.[0]?.text?.trim().toLowerCase() ?? "";
-      const shouldRespond = !answer.startsWith("skip");
-      log94.info("Gate decision", {
+    const result = await runGateCall(provider, prompt, model, settings);
+    if (!result.ok) {
+      log94.warn("Gate: call failed, fail-open", {
         traceId,
-        decision: shouldRespond ? "respond" : "skip",
-        rawAnswer: answer,
+        provider,
         model,
-        chatType,
-        messageCount: messages4.length,
+        reason: result.reason,
+        status: result.status,
+        error: result.error,
         durationMs: Date.now() - startMs
       });
-      return shouldRespond;
-    } catch (fetchError) {
-      clearTimeout(timeout);
-      const errName = fetchError.name;
-      if (errName === "AbortError") {
-        log94.warn("Gate: timeout, fail-open", { traceId, durationMs: Date.now() - startMs });
-      } else {
-        log94.warn("Gate: fetch error, fail-open", { traceId, error: String(fetchError) });
-      }
       return true;
     }
+    const shouldRespond = !result.answer.startsWith("skip");
+    log94.info("Gate decision", {
+      traceId,
+      decision: shouldRespond ? "respond" : "skip",
+      rawAnswer: result.answer,
+      provider,
+      model,
+      chatType,
+      messageCount: messages4.length,
+      durationMs: Date.now() - startMs
+    });
+    return shouldRespond;
   } catch (error3) {
     log94.warn("Gate: unexpected error, fail-open", { traceId, error: String(error3) });
     return true;
@@ -367543,7 +367629,7 @@ async function setupAgentDispatcher(eventBus, services, db2) {
     log94.info("Agent dispatcher shutdown complete");
   };
 }
-var import_api31, log94, agentDispatchLimiter, _natsGenieProviderCtor, QUOTED_MESSAGE_MAX_CHARS = 4000, DM_HISTORY_LIMIT = 20, LIFECYCLE_PREVIEW_MAX_CHARS = 160, LIFECYCLE_SENSITIVE_KEY_PARTS, PRESENCE_DEFAULT_MIN_MS = 5000, PRESENCE_DEFAULT_MAX_WAIT_MS = 30000, activeMessageDebouncer = null, DEFAULT_DISPATCH_ERROR_MESSAGE = "Opa, tive um probleminha aqui \uD83D\uDE05 Pode mandar de novo?", DEFAULT_ERROR_HANDOFF_MESSAGE = "T\xF4 com um probleminha t\xE9cnico aqui agora. Logo algu\xE9m do time vai entrar em contato com voc\xEA.", TRANSIENT_DISPATCH_ERROR_PATTERNS, TRANSIENT_DISPATCH_RETRY_DELAYS_MS, CHANNEL_MESSAGE_LIMITS, DEFAULT_MESSAGE_LIMIT = 4000, MEDIA_BASE_PATH3, NON_DISPATCHABLE_CONTENT_TYPES, MEDIA_ICONS, MEDIA_WAIT_NULL, mediaCompletions, mediaResultCache, MEDIA_WAIT_TIMEOUT_MS = 30000, DEFAULT_SEND_MEDIA_PATH_TYPES, BOT_PREFIX = "\uD83E\uDD16 ", activeStreams, activeRunAborts, CancelThreadPayloadSchema, sessionActivityStore, PROC_REACT_START, PROC_REACT_DONE = "\u2705", queuedMessageAcks, QUEUED_ACK_TIMEOUT_MS = 120000, providerCache, openclawClientPool, _openClawClientCtor, nullFilterWarnedInstances, ACTIVE_OWNER_IDENTIFIER_CACHE_TTL_MS = 1e4, cachedActiveOwnerIdentifiers = null, cachedActiveOwnerIdentifiersAt = 0, tenantCachedActiveOwnerIdentifiers, DEFAULT_GATE_MODEL = "gemini-3-flash-preview", GATE_TIMEOUT_MS = 3000, setupAgentResponder;
+var import_api31, log94, agentDispatchLimiter, _natsGenieProviderCtor, QUOTED_MESSAGE_MAX_CHARS = 4000, DM_HISTORY_LIMIT = 20, LIFECYCLE_PREVIEW_MAX_CHARS = 160, LIFECYCLE_SENSITIVE_KEY_PARTS, PRESENCE_DEFAULT_MIN_MS = 5000, PRESENCE_DEFAULT_MAX_WAIT_MS = 30000, activeMessageDebouncer = null, DEFAULT_DISPATCH_ERROR_MESSAGE = "Opa, tive um probleminha aqui \uD83D\uDE05 Pode mandar de novo?", DEFAULT_ERROR_HANDOFF_MESSAGE = "T\xF4 com um probleminha t\xE9cnico aqui agora. Logo algu\xE9m do time vai entrar em contato com voc\xEA.", TRANSIENT_DISPATCH_ERROR_PATTERNS, TRANSIENT_DISPATCH_RETRY_DELAYS_MS, CHANNEL_MESSAGE_LIMITS, DEFAULT_MESSAGE_LIMIT = 4000, MEDIA_BASE_PATH3, NON_DISPATCHABLE_CONTENT_TYPES, MEDIA_ICONS, MEDIA_WAIT_NULL, mediaCompletions, mediaResultCache, MEDIA_WAIT_TIMEOUT_MS = 30000, DEFAULT_SEND_MEDIA_PATH_TYPES, BOT_PREFIX = "\uD83E\uDD16 ", activeStreams, activeRunAborts, CancelThreadPayloadSchema, sessionActivityStore, PROC_REACT_START, PROC_REACT_DONE = "\u2705", queuedMessageAcks, QUEUED_ACK_TIMEOUT_MS = 120000, providerCache, openclawClientPool, _openClawClientCtor, nullFilterWarnedInstances, ACTIVE_OWNER_IDENTIFIER_CACHE_TTL_MS = 1e4, cachedActiveOwnerIdentifiers = null, cachedActiveOwnerIdentifiersAt = 0, tenantCachedActiveOwnerIdentifiers, setupAgentResponder;
 var init_agent_dispatcher = __esm(() => {
   init_src2();
   init_src();
@@ -373904,6 +373990,14 @@ var init_settings = __esm(() => {
       description: "Default DeepSeek vision/text model for vision provider dogfood",
       defaultValue: "deepseek-v4-flash"
     },
+    {
+      key: "vision.model",
+      category: "vision",
+      valueType: "string",
+      isSecret: false,
+      description: "Default Gemini vision model for `omni see` (default: gemini-3.5-flash-lite)",
+      defaultValue: "gemini-3.5-flash-lite"
+    },
     {
       key: "deepseek.anthropic_url",
       category: "media",
@@ -373920,6 +374014,22 @@ var init_settings = __esm(() => {
       description: "Default language for media processing",
       defaultValue: "pt"
     },
+    {
+      key: "media.gemini.model",
+      category: "media",
+      valueType: "string",
+      isSecret: false,
+      description: "Gemini model for inbound image description, video analysis and document OCR (default: gemini-3.5-flash-lite)",
+      defaultValue: "gemini-3.5-flash-lite"
+    },
+    {
+      key: "media.openai.vision_model",
+      category: "media",
+      valueType: "string",
+      isSecret: false,
+      description: "OpenAI vision model used when the Gemini vision lane is unavailable (default: gpt-5.6-luna)",
+      defaultValue: "gpt-5.6-luna"
+    },
     {
       key: "prompt.image_description",
       category: "prompts",
@@ -373955,6 +374065,37 @@ var init_settings = __esm(() => {
       isSecret: false,
       description: "Override prompt for response gate (null = code default)"
     },
+    {
+      key: "gate.provider",
+      category: "providers",
+      valueType: "string",
+      isSecret: false,
+      description: 'Response gate provider: "gemini" (default) or "openai".',
+      defaultValue: "gemini"
+    },
+    {
+      key: "gate.gemini.model",
+      category: "providers",
+      valueType: "string",
+      isSecret: false,
+      description: "Default Gemini model for the response gate.",
+      defaultValue: "gemini-3.5-flash-lite"
+    },
+    {
+      key: "gate.openai.model",
+      category: "providers",
+      valueType: "string",
+      isSecret: false,
+      description: "Default OpenAI model for the response gate.",
+      defaultValue: "gpt-5.6-luna"
+    },
+    {
+      key: "gate.openai.base_url",
+      category: "providers",
+      valueType: "string",
+      isSecret: false,
+      description: "OpenAI-compatible base URL for the response gate (default: https://api.openai.com/v1). Set it to route the gate through a gateway."
+    },
     {
       key: "tts.provider",
       category: "providers",
@@ -374031,8 +374172,8 @@ var init_settings = __esm(() => {
       category: "providers",
       valueType: "string",
       isSecret: false,
-      description: "Default OpenAI STT model. Primary: gpt-4o-transcribe (dedicated STT); gpt-audio-* models use the chat lane, whose output is validated with a transcriptions-endpoint fallback.",
-      defaultValue: "gpt-4o-transcribe"
+      description: "Default OpenAI STT model. Primary: gpt-transcribe (dedicated STT); gpt-audio-* models use the chat lane, whose output is validated with a transcriptions-endpoint fallback.",
+      defaultValue: "gpt-transcribe"
     },
     {
       key: "stt.gemini.model",
@@ -374040,7 +374181,7 @@ var init_settings = __esm(() => {
       valueType: "string",
       isSecret: false,
       description: "Default Gemini direct-audio STT model.",
-      defaultValue: "gemini-3.1-flash-lite"
+      defaultValue: "gemini-3.5-flash-lite"
     },
     {
       key: "videogen.gemini.model",
@@ -375304,8 +375445,10 @@ var init_tts3 = __esm(() => {
     }
     async synthesize(text3, options) {
       const apiKey = await this.getApiKey();
-      const defaultVoice = await this.settings.getString("elevenlabs.default_voice", "ELEVENLABS_DEFAULT_VOICE", "JBFqnCBsd6RMkjVDRZzb");
-      const defaultModel = await this.settings.getString("elevenlabs.default_model", undefined, "eleven_v3");
+      const configuredVoice = await this.settings.getString("elevenlabs.default_voice", "ELEVENLABS_DEFAULT_VOICE", "JBFqnCBsd6RMkjVDRZzb");
+      const defaultVoice = !configuredVoice || configuredVoice === "undefined" ? undefined : configuredVoice;
+      const configuredModel = await this.settings.getString("elevenlabs.default_model", undefined, "eleven_v3");
+      const defaultModel = !configuredModel || configuredModel === "undefined" ? undefined : configuredModel;
       const voiceId = options?.voiceId || defaultVoice || "JBFqnCBsd6RMkjVDRZzb";
       const modelId = options?.modelId || defaultModel || "eleven_v3";
       const stability = options?.stability ?? 0.5;
@@ -389665,7 +389808,7 @@ var init_instances3 = __esm(() => {
     messageSplitDelayMinMs: exports_external.number().int().min(0).default(300).describe('Minimum delay between split chunks in milliseconds (used when messageSplitDelayMode="randomized")'),
     messageSplitDelayMaxMs: exports_external.number().int().min(0).default(1000).describe('Maximum delay between split chunks in milliseconds (used when messageSplitDelayMode="randomized")'),
     agentGateEnabled: exports_external.boolean().default(false).describe("Enable LLM response gate (pre-filter before agent dispatch)"),
-    agentGateModel: exports_external.string().nullable().default(null).describe("Model for response gate (default: gemini-3-flash-preview)"),
+    agentGateModel: exports_external.string().nullable().default(null).describe("Model for response gate (default: gate.gemini.model / gemini-3.5-flash-lite). Must match the provider selected by the gate.provider setting."),
     agentGatePrompt: exports_external.string().nullable().default(null).describe("Custom prompt for response gate (null = use default)"),
     telegramBotToken: exports_external.string().optional().nullable().describe("Telegram bot token (persisted for reconnection)"),
     discordBotToken: exports_external.string().optional().nullable().describe("Discord bot token (persisted for reconnection)"),
@@ -399414,19 +399557,23 @@ async function setupMediaProcessor(eventBus, db2, services) {
     audioPrompt,
     imagePrompt,
     videoPrompt,
-    documentPrompt
+    documentPrompt,
+    geminiVisionModel,
+    openaiVisionModel
   ] = await Promise.all([
     services.settings.getSecret("groq.api_key", "GROQ_API_KEY"),
     services.settings.getSecret("openai.api_key", "OPENAI_API_KEY"),
     services.settings.getSecret("gemini.api_key", "GEMINI_API_KEY"),
     services.settings.getString("media.default_language", "DEFAULT_LANGUAGE", "pt"),
     services.settings.getString("stt.provider", "STT_PROVIDER", "openai"),
-    services.settings.getString("stt.openai.model", "OPENAI_STT_MODEL", "gpt-4o-transcribe"),
+    services.settings.getString("stt.openai.model", "OPENAI_STT_MODEL", OPENAI_TRANSCRIBE_MODEL),
     services.settings.getString("stt.gemini.model", "GEMINI_STT_MODEL", GEMINI_AUDIO_MODEL),
     services.settings.getString("prompt.audio_transcription"),
     services.settings.getString("prompt.image_description"),
     services.settings.getString("prompt.video_description"),
-    services.settings.getString("prompt.document_ocr")
+    services.settings.getString("prompt.document_ocr"),
+    services.settings.getString("media.gemini.model", "GEMINI_MEDIA_MODEL", GEMINI_MODEL),
+    services.settings.getString("media.openai.vision_model", "OPENAI_VISION_MODEL", OPENAI_VISION_MODEL)
   ]);
   const mediaService = createMediaProcessingService({
     groqApiKey,
@@ -399434,8 +399581,10 @@ async function setupMediaProcessor(eventBus, db2, services) {
     geminiApiKey,
     defaultLanguage,
     audioProvider: audioProvider ?? "openai",
-    audioModel: audioModel ?? "gpt-4o-transcribe",
+    audioModel: audioModel ?? OPENAI_TRANSCRIBE_MODEL,
     geminiAudioModel: geminiAudioModel ?? GEMINI_AUDIO_MODEL,
+    geminiVisionModel: geminiVisionModel ?? GEMINI_MODEL,
+    openaiVisionModel: openaiVisionModel ?? OPENAI_VISION_MODEL,
     audioPrompt: audioPrompt ?? undefined
   });
   const mediaStorage2 = new MediaStorageService(db2);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@automagik/omni",
-  "version": "2.260916.12",
+  "version": "2.260921.2",
   "repository": {
     "type": "git",
     "url": "git+https://github.com/automagik-dev/omni.git",