From 8658af1018794cf7d7968c3d852a4362aef1e25b Mon Sep 17 00:00:00 2001
From: Concedo <39025047+LostRuins@users.noreply.github.com>
Date: Thu, 5 Mar 2026 11:11:46 +0800
Subject: [PATCH] qwen3tts default to cpu unless gpu selected

---
 embd_res/klite.embd                           | 142 ++++++++++++++++--
 koboldcpp.py                                  |   2 +-
 .../qwen3tts/audio_tokenizer_decoder.cpp      |   2 +-
 .../qwen3tts/audio_tokenizer_encoder.cpp      |   4 +-
 otherarch/qwen3tts/gguf_loader.h              |   2 +
 otherarch/qwen3tts/tts_transformer.cpp        |   4 +-
 otherarch/tts_adapter.cpp                     |   2 +
 7 files changed, 140 insertions(+), 18 deletions(-)

diff --git a/embd_res/klite.embd b/embd_res/klite.embd
index f4174972d..e60bf6114 100644
--- a/embd_res/klite.embd
+++ b/embd_res/klite.embd
@@ -4355,6 +4355,7 @@ Current version indicated by LITEVER below.
 		saved_oai_role: 3, //0=user,1=assistant,2=system, 3=auto
 		saved_a1111_url: default_a1111_base,
 		saved_comfy_url: default_comfy_base,
+		saved_comfy_bearer_token: "",
 		saved_xtts_url: default_xtts_base,
 		saved_alltalk_url: default_alltalk_base,
 		saved_mcp_urls: "",
@@ -8670,8 +8671,23 @@ Current version indicated by LITEVER below.
 	{
 		console.log("Attempt ComfyUI Connection...");
 		//establish initial connection to a1111 api
-		fetch(localsettings.saved_comfy_url + comfy_models_endpoint)
-		.then(x => x.json())
+		let headers = {};
+		if (localsettings.saved_comfy_bearer_token != "") {
+			headers['Authorization'] = 'Bearer '+localsettings.saved_comfy_bearer_token
+		}
+		fetch(localsettings.saved_comfy_url + comfy_models_endpoint, {
+			headers: headers
+		})
+		.then(x => {
+			if (!silent && !x.ok) {
+				msg = "Connection returned "+x.status
+				if (x.status == 401) {
+					msg +=". Check Bearer Token."
+				}
+				msgbox(msg)
+			}
+			return x.json()
+		})
 		.then(modelsdata => {
 
 			if (modelsdata == null || modelsdata.length == 0) {
@@ -8690,12 +8706,12 @@ Current version indicated by LITEVER below.
 		}).catch((error) => {
 			if(!silent)
 			{
-				msgbox("ComfyUI Connect Error: " + error+"\nPlease make sure ComfyUI is running at "+localsettings.saved_comfy_url+" and properly configured!\n\nIt must be launched with the flags --listen --enable-cors-header '*' to enable API access\n");
+				msgbox("ComfyUI Connect Error: " + error+"\nPlease make sure ComfyUI is running at "+localsettings.saved_comfy_url+" and properly configured!\n\nIt must be launched with the flags --listen --enable-cors-header '*' to enable API access. If using ComfyUI-Login, the Bearer Token must match.\n");
 			}
 			comfyui_is_connected = false;
 		});
 
-		fetch(localsettings.saved_comfy_url + comfy_loras_endpoint)
+		fetch(localsettings.saved_comfy_url + comfy_loras_endpoint, {headers:headers})
         .then(x => x.json())
         .then(lorasdata => {
             //repopulate our lora list
@@ -8735,9 +8751,14 @@ Current version indicated by LITEVER below.
 		const formData = new FormData();
 		formData.append('image', blob, filename);
 		formData.append('overwrite',1);
+		let headers = {};
+		if (localsettings.saved_comfy_bearer_token != "") {
+			headers['Authorization'] = 'Bearer '+localsettings.saved_comfy_bearer_token
+		}
 		return fetch(`${upload_endpoint}`, {
 			method: 'POST',
 			body: formData,
+			headers: headers,
 		}).then(response => {
 			if (!response.ok) {
 				throw new Error(`Upload failed with status ${response.status}`);
@@ -8874,11 +8895,13 @@ Current version indicated by LITEVER below.
 		image_db[imgid].type = 0; //0=image, 1=audio
 
 		uploadBase64ImgToComfy(req_payload["source_image"],comfyimg2imgname).then(() => {
+		headers = {'Content-Type': 'application/json'}
+		if (localsettings.saved_comfy_bearer_token != "") {
+			headers['Authorization'] = 'Bearer '+localsettings.saved_comfy_bearer_token
+		}
 		fetch(gen_endpoint, {
 			method: 'POST',
-			headers: {
-				'Content-Type': 'application/json',
-			},
+			headers: headers,
 			body: JSON.stringify(genimg_payload),
 		})
 		.then(x => x.json())
@@ -9033,6 +9056,15 @@ Current version indicated by LITEVER below.
 		},false);
 	}
 
+	function set_comfy_key()
+	{
+		inputBox("Enter ComfyUI API Auth Bearer Token","ComfyUI API Key",localsettings.saved_comfy_bearer_token,"Input ComfyUI API Key", ()=>{
+			let userinput = getInputBoxValue();
+			userinput = userinput.trim();
+			localsettings.saved_comfy_bearer_token = userinput.trim();
+		},true);
+	}
+
 	function generate_pollinations_image(req_payload, autoappend)
 	{
 		let splits = req_payload.prompt.split("###");
@@ -9835,6 +9867,7 @@ Current version indicated by LITEVER below.
 			new_save_storyobj.savedsettings.saved_kai_key = "";
 			new_save_storyobj.savedsettings.saved_a1111_url = "";
 			new_save_storyobj.savedsettings.saved_comfy_url = "";
+			new_save_storyobj.savedsettings.saved_comfy_bearer_token = "";
 			new_save_storyobj.savedsettings.saved_xtts_url = "";
 			new_save_storyobj.savedsettings.saved_mcp_urls = "";
 
@@ -9976,6 +10009,22 @@ Current version indicated by LITEVER below.
 						console.log("Unzip failed: " + error);
 					}
 
+					//try JSONL
+					try
+					{
+						const lines = text.split('\n');
+						let temparrs = lines.filter(line => line.trim()).map(line => JSON.parse(line));
+						if(temparrs && temparrs.length>1 && temparrs[temparrs.length-1].mes && temparrs[temparrs.length-1].name)
+						{
+							load_tavern_jsonl(temparrs);
+							return;
+						}
+					}
+					catch(error)
+					{
+						console.log("JSONL import failed: " + error);
+					}
+
 					// 5. Fallback to plaintext if .txt
 					if (selectedFilename.endsWith(".txt")) {
 						msgboxYesNo(
@@ -10270,6 +10319,7 @@ Current version indicated by LITEVER below.
 					let tmp_kai2 = localsettings.saved_kai_key;
 					let tmp_a1111 = localsettings.saved_a1111_url;
 					let tmp_comfy = localsettings.saved_comfy_url;
+					let tmp_comfy_bearer_token = localsettings.saved_comfy_bearer_token;
 					let tmp_xtts = localsettings.saved_xtts_url;
 					let tmp_imggen = localsettings.generate_images_mode;
 					let tmp_mcp = localsettings.saved_mcp_urls;
@@ -10327,6 +10377,7 @@ Current version indicated by LITEVER below.
 					localsettings.saved_kai_key = tmp_kai2;
 					localsettings.saved_a1111_url = tmp_a1111;
 					localsettings.saved_comfy_url = tmp_comfy;
+					localsettings.saved_comfy_bearer_token = tmp_comfy_bearer_token;
 					localsettings.saved_xtts_url = tmp_xtts;
 					localsettings.generate_images_mode = tmp_imggen;
 					localsettings.saved_mcp_urls = tmp_mcp;
@@ -10614,6 +10665,68 @@ Current version indicated by LITEVER below.
 		return false;
 	}
 
+	function load_tavern_jsonl(obj)
+	{
+		if(localsettings.opmode!=3 && localsettings.opmode!=4)
+		{
+			//force into instruct
+			localsettings.opmode = 4;
+		}
+		gametext_arr = [];
+		if(localsettings.opmode==3) //import as chat
+		{
+			if(localsettings.gui_type_chat!=3)
+			{
+				localsettings.gui_type_chat = 2;
+			}
+			for(let i=0;i<obj.length;++i)
+			{
+				let curr = obj[i];
+				if(curr.mes)
+				{
+					if(curr.name)
+					{
+						if(curr.is_user)
+						{
+							localsettings.chatname = curr.name;
+						}else{
+							localsettings.chatopponent = curr.name;
+						}
+					}
+					gametext_arr.push("\n"+curr.name+": "+curr.mes);
+				}
+			}
+		}
+		else //import as instruct
+		{
+			localsettings.gui_type_instruct = 2;
+			localsettings.inject_chatnames_instruct = true;
+			for(let i=0;i<obj.length;++i)
+			{
+				let curr = obj[i];
+				if(curr.mes)
+				{
+					if(curr.is_user)
+					{
+						if(curr.name)
+						{
+							localsettings.chatname = curr.name;
+						}
+						gametext_arr.push(get_instructstartplaceholder()+curr.name+": "+curr.mes);
+					}else{
+						if(curr.name)
+						{
+							localsettings.chatopponent = curr.name;
+						}
+						gametext_arr.push(get_instructendplaceholder()+curr.name+": "+curr.mes);
+					}
+				}
+			}
+		}
+		update_for_sidepanel();
+		render_gametext(true);
+		sync_multiplayer(true);
+	}
 	function load_tavern_obj(obj)
 	{
 		let selectedgreeting = "";
@@ -22656,11 +22769,15 @@ Current version indicated by LITEVER below.
 				if(comfyid && comfyid!="")
 				{
 					//comfyui polling
+					let json_headers = {'Content-Type': 'application/json'};
+					let headers={};
+					if (localsettings.saved_comfy_bearer_token != "") {
+						headers['Authorization'] = 'Bearer '+localsettings.saved_comfy_bearer_token;
+						json_headers['Authorization'] = 'Bearer '+localsettings.saved_comfy_bearer_token;
+					}
 					fetch(localsettings.saved_comfy_url + comfy_history_endpoint + "/" + comfyid, {
 						method: 'GET',
-						headers: {
-							'Content-Type': 'application/json',
-						}
+						headers: json_headers,
 					})
 					.then(x => x.json())
 					.then(resp2 => {
@@ -22670,7 +22787,7 @@ Current version indicated by LITEVER below.
 							img.done = true;
 							let finalfilename = resp2[comfyid].outputs["9"].images[0].filename;
 							//fetch final image
-							fetch(localsettings.saved_comfy_url + comfy_results_endpoint + finalfilename)
+							fetch(localsettings.saved_comfy_url + comfy_results_endpoint + finalfilename, {headers: headers})
 							.then((response) => {
 								return response.blob(); // Convert the response into a Blob
 							})
@@ -29616,10 +29733,11 @@ Current version indicated by LITEVER below.
 						<div id="generate_images_comfy_container" class="settinglabel hidden settingsbox">
 							<div class="settinglabel" style="display: flex; width: 100%">
 							<div class="justifyleft">Model</div>
-							<select title="Select Image Model" class="form-control push-right" id="generate_images_comfy_model" style="width:calc(100% - 58px)">
+							<select title="Select Image Model" class="form-control push-right" id="generate_images_comfy_model" style="width:calc(100% - 170px)">
 								<option value="">[None]</option>
 							</select>
 							<button type="button" class="btn btn-primary" onclick="set_comfy_endpoint()" style="width:52px; padding: 2px 2px; margin-left: 3px; font-size:12px;">Set URL</button>
+							<button type="button" class="btn btn-primary" onclick="set_comfy_key()" style="width:52px; padding: 2px 2px; margin-left: 3px; font-size:12px;">Set Key</button>
 							</div>
 							<div class="settinglabel" style="display: flex; width: 100%">
 							<div class="justifyleft">LoRA</div>
diff --git a/koboldcpp.py b/koboldcpp.py
index d2da96236..7df37df3f 100755
--- a/koboldcpp.py
+++ b/koboldcpp.py
@@ -70,7 +70,7 @@ dry_seq_break_max = 128
 extra_images_max = 4 # for kontext/qwen img
 
 # global vars
-KcppVersion = "1.109.1"
+KcppVersion = "1.109.2"
 showdebug = True
 kcpp_instance = None #global running instance
 global_memory = {"tunnel_url": "", "restart_target":"", "input_to_exit":False, "load_complete":False, "restart_override_config_target":""}
diff --git a/otherarch/qwen3tts/audio_tokenizer_decoder.cpp b/otherarch/qwen3tts/audio_tokenizer_decoder.cpp
index fa4bc5097..c58798135 100644
--- a/otherarch/qwen3tts/audio_tokenizer_decoder.cpp
+++ b/otherarch/qwen3tts/audio_tokenizer_decoder.cpp
@@ -339,7 +339,7 @@ bool AudioTokenizerDecoder::load_model(const std::string & model_path) {
         upload_if_present(model_.vq_rest_codebook[i]);
     }
 
-    state_.backend = init_preferred_backend("AudioTokenizerDecoder", &error_msg_, true);
+    state_.backend = init_preferred_backend("AudioTokenizerDecoder", &error_msg_, qwen3tts_allowgpu);
     if (!state_.backend) {
         return false;
     }
diff --git a/otherarch/qwen3tts/audio_tokenizer_encoder.cpp b/otherarch/qwen3tts/audio_tokenizer_encoder.cpp
index 31b4eb31e..47a6d14a2 100644
--- a/otherarch/qwen3tts/audio_tokenizer_encoder.cpp
+++ b/otherarch/qwen3tts/audio_tokenizer_encoder.cpp
@@ -249,11 +249,11 @@ bool AudioTokenizerEncoder::load_model(const std::string & model_path) {
     }
 
     if (!load_tensor_data_from_file(model_path, gguf_ctx, model_.ctx,
-                                     model_.tensors, model_.buffer, error_msg_, GGML_BACKEND_DEVICE_TYPE_GPU)) {
+                                     model_.tensors, model_.buffer, error_msg_, qwen3tts_allowgpu?GGML_BACKEND_DEVICE_TYPE_GPU:GGML_BACKEND_DEVICE_TYPE_CPU)) {
         return false;
     }
 
-    state_.backend = init_preferred_backend("AudioTokenizerEncoder", &error_msg_, true);
+    state_.backend = init_preferred_backend("AudioTokenizerEncoder", &error_msg_, qwen3tts_allowgpu);
     if (!state_.backend) {
         return false;
     }
diff --git a/otherarch/qwen3tts/gguf_loader.h b/otherarch/qwen3tts/gguf_loader.h
index fd949e2a7..0fd45ec38 100644
--- a/otherarch/qwen3tts/gguf_loader.h
+++ b/otherarch/qwen3tts/gguf_loader.h
@@ -9,6 +9,8 @@
 #include <vector>
 #include <memory>
 
+extern bool qwen3tts_allowgpu;
+
 namespace qwen3_tts {
 
 // Generic GGUF model loader class
diff --git a/otherarch/qwen3tts/tts_transformer.cpp b/otherarch/qwen3tts/tts_transformer.cpp
index 2bb0eb98b..9f69b7468 100644
--- a/otherarch/qwen3tts/tts_transformer.cpp
+++ b/otherarch/qwen3tts/tts_transformer.cpp
@@ -97,7 +97,7 @@ bool TTSTransformer::load_model(const std::string & model_path) {
     gguf_free(ctx);
     if (meta_ctx) ggml_free(meta_ctx);
 
-    state_.backend = init_preferred_backend("TTSTransformer", &error_msg_, true);
+    state_.backend = init_preferred_backend("TTSTransformer", &error_msg_, qwen3tts_allowgpu);
     if (!state_.backend) {
         return false;
     }
@@ -553,7 +553,7 @@ bool TTSTransformer::create_tensors(struct gguf_context * ctx) {
  }
 
 bool TTSTransformer::load_tensor_data(const std::string & path, struct gguf_context * ctx) {
-    ggml_backend_t backend = init_preferred_backend("TTSTransformer", &error_msg_, true);
+    ggml_backend_t backend = init_preferred_backend("TTSTransformer", &error_msg_, qwen3tts_allowgpu);
     if (!backend) {
         return false;
     }
diff --git a/otherarch/tts_adapter.cpp b/otherarch/tts_adapter.cpp
index ad7d2849c..feab32a25 100644
--- a/otherarch/tts_adapter.cpp
+++ b/otherarch/tts_adapter.cpp
@@ -501,6 +501,7 @@ static std::string detectedarch = "";
 //qwen3tts specific
 static bool is_qwen3tts_file = false;
 static qwen3_tts::Qwen3TTS qwen3tts_runner;
+bool qwen3tts_allowgpu = false;
 
 int total_tts_gens = 0;
 static std::string tts_executable_path = "";
@@ -576,6 +577,7 @@ bool ttstype_load_model(const tts_load_model_inputs inputs)
     }
     else if(is_qwen3tts_file)
     {
+        qwen3tts_allowgpu = (inputs.gpulayers>0?true:false);
         if (!qwen3tts_runner.load_models(modelfile_ttc,modelfile_cts)) {
             printf("\nQwen3TTS Load Error: %s\n", qwen3tts_runner.get_error().c_str());
             return false;