Merge 937e2b5a54 into fb3af2a08f

Removed deprecated Gemini models

.env.template

@@ -64,6 +64,11 @@ CUSTOM_MODELS=
 # Change default model
 DEFAULT_MODEL=
 
+# (optional)
+# Default: Empty
+# Change default compress model
+DEFAULT_COMPRESS_MODEL=
+
 # anthropic claude Api Key.(optional)
 ANTHROPIC_API_KEY=
 

README.md

@@ -322,6 +322,10 @@ For ByteDance: use `modelName@bytedance=deploymentName` to customize model name
 
 Change default model
 
+### `DEFAULT_COMPRESS_MODEL` （optional）
+
+Change default summary model
+
 ### `VISION_MODELS` (optional)
 
 > Default: Empty
@@ -348,7 +352,6 @@ Stability API key.
 
 Customize Stability API url.
 
-
 ### `ENABLE_MCP` (optional)
 
 Enable MCP（Model Context Protocol）Feature

README_CN.md

@@ -244,6 +244,10 @@ DeepSeek Api Url.
 
 更改默认模型
 
+### `DEFAULT_COMPRESS_MODEL` （可选）
+
+更改默认对话摘要模型
+
 ### `VISION_MODELS` (可选)
 
 > 默认值：空

README_JA.md

@@ -217,6 +217,10 @@ ByteDance モードでは、`modelName@bytedance=deploymentName` 形式でモデ
 
 デフォルトのモデルを変更します。
 
+### `DEFAULT_COMPRESS_MODEL` （オプション）
+
+デフォルトの圧縮モデルを変更します。
+
 ### `VISION_MODELS` (オプション)
 
 > デフォルト：空

app/api/config/route.ts

@@ -14,6 +14,7 @@ const DANGER_CONFIG = {
   disableFastLink: serverConfig.disableFastLink,
   customModels: serverConfig.customModels,
   defaultModel: serverConfig.defaultModel,
+  defaultCompressModel: serverConfig.defaultCompressModel,
   visionModels: serverConfig.visionModels,
 };
 

app/client/platforms/openai.ts

@@ -56,7 +56,7 @@ export interface OpenAIListModelResponse {
 
 export interface RequestPayload {
   messages: {
-    role: "system" | "user" | "assistant";
+    role: "developer" | "system" | "user" | "assistant";
     content: string | MultimodalContent[];
   }[];
   stream?: boolean;
@@ -238,8 +238,16 @@ export class ChatGPTApi implements LLMApi {
         // Please do not ask me why not send max_tokens, no reason, this param is just shit, I dont want to explain anymore.
       };
 
-      // O1 使用 max_completion_tokens 控制token数 (https://platform.openai.com/docs/guides/reasoning#controlling-costs)
       if (isO1OrO3) {
+        // by default the o1/o3 models will not attempt to produce output that includes markdown formatting
+        // manually add "Formatting re-enabled" developer message to encourage markdown inclusion in model responses
+        // (https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/reasoning?tabs=python-secure#markdown-output)
+        requestPayload["messages"].unshift({
+          role: "developer",
+          content: "Formatting re-enabled",
+        });
+
+        // o1/o3 uses max_completion_tokens to control the number of tokens (https://platform.openai.com/docs/guides/reasoning#controlling-costs)
         requestPayload["max_completion_tokens"] = modelConfig.max_tokens;
       }
 

app/components/model-config.tsx

@@ -259,13 +259,15 @@ export function ModelConfigList(props: {
             });
           }}
         >
-          {allModels
-            .filter((v) => v.available)
-            .map((v, i) => (
+          {Object.keys(groupModels).map((providerName, index) => (
+            <optgroup label={providerName} key={index}>
+              {groupModels[providerName].map((v, i) => (
                 <option value={`${v.name}@${v.provider?.providerName}`} key={i}>
-                {v.displayName}({v.provider?.providerName})
+                  {v.displayName}
                 </option>
               ))}
+            </optgroup>
+          ))}
         </Select>
       </ListItem>
     </>

app/config/server.ts

@@ -23,6 +23,7 @@ declare global {
       DISABLE_FAST_LINK?: string; // disallow parse settings from url or not
       CUSTOM_MODELS?: string; // to control custom models
       DEFAULT_MODEL?: string; // to control default model in every new chat window
+      DEFAULT_COMPRESS_MODEL?: string; // to control default compress model
       VISION_MODELS?: string; // to control vision models
 
       // stability only
@@ -135,6 +136,7 @@ export const getServerSideConfig = () => {
   const disableGPT4 = !!process.env.DISABLE_GPT4;
   let customModels = process.env.CUSTOM_MODELS ?? "";
   let defaultModel = process.env.DEFAULT_MODEL ?? "";
+  let defaultCompressModel = process.env.DEFAULT_COMPRESS_MODEL ?? "";
   let visionModels = process.env.VISION_MODELS ?? "";
 
   if (disableGPT4) {
@@ -145,6 +147,9 @@ export const getServerSideConfig = () => {
     if (defaultModel && isGPT4Model(defaultModel)) {
       defaultModel = "";
     }
+    if (defaultCompressModel && isGPT4Model(defaultCompressModel)) {
+      defaultCompressModel = "";
+    }
   }
 
   const isStability = !!process.env.STABILITY_API_KEY;
@@ -262,6 +267,7 @@ export const getServerSideConfig = () => {
     disableFastLink: !!process.env.DISABLE_FAST_LINK,
     customModels,
     defaultModel,
+    defaultCompressModel,
     visionModels,
     allowedWebDavEndpoints,
     enableMcp: process.env.ENABLE_MCP === "true",

app/constant.ts

@@ -523,20 +523,15 @@ const openaiModels = [
 ];
 
 const googleModels = [
-  "gemini-1.0-pro", // Deprecated on 2/15/2025
   "gemini-1.5-pro-latest",
   "gemini-1.5-pro",
   "gemini-1.5-pro-002",
-  "gemini-1.5-pro-exp-0827",
   "gemini-1.5-flash-latest",
   "gemini-1.5-flash-8b-latest",
   "gemini-1.5-flash",
   "gemini-1.5-flash-8b",
   "gemini-1.5-flash-002",
-  "gemini-1.5-flash-exp-0827",
   "learnlm-1.5-pro-experimental",
-  "gemini-exp-1114",
-  "gemini-exp-1121",
   "gemini-exp-1206",
   "gemini-2.0-flash",
   "gemini-2.0-flash-exp",
@@ -546,6 +541,7 @@ const googleModels = [
   "gemini-2.0-flash-thinking-exp-01-21",
   "gemini-2.0-pro-exp",
   "gemini-2.0-pro-exp-02-05",
+  "gemini-2.5-pro-preview-06-05",
 ];
 
 const anthropicModels = [
@@ -632,6 +628,18 @@ const xAIModes = [
   "grok-2-vision-1212",
   "grok-2-vision",
   "grok-2-vision-latest",
+  "grok-3-mini-fast-beta",
+  "grok-3-mini-fast",
+  "grok-3-mini-fast-latest",
+  "grok-3-mini-beta",
+  "grok-3-mini",
+  "grok-3-mini-latest",
+  "grok-3-fast-beta",
+  "grok-3-fast",
+  "grok-3-fast-latest",
+  "grok-3-beta",
+  "grok-3",
+  "grok-3-latest",
 ];
 
 const chatglmModels = [

app/store/access.ts

@@ -140,6 +140,7 @@ const DEFAULT_ACCESS_STATE = {
   disableFastLink: false,
   customModels: "",
   defaultModel: "",
+  defaultCompressModel: "",
   visionModels: "",
 
   // tts config
@@ -255,12 +256,21 @@ export const useAccessStore = createPersistStore(
         .then((res) => res.json())
         .then((res) => {
           const defaultModel = res.defaultModel ?? "";
-          if (defaultModel !== "") {
+          if (defaultModel) {
             const [model, providerName] = getModelProvider(defaultModel);
             DEFAULT_CONFIG.modelConfig.model = model;
             DEFAULT_CONFIG.modelConfig.providerName = providerName as any;
           }
 
+          const defaultCompressModel = res.defaultCompressModel ?? "";
+          if (defaultCompressModel) {
+            const [model, providerName] =
+              getModelProvider(defaultCompressModel);
+            DEFAULT_CONFIG.modelConfig.compressModel = model;
+            DEFAULT_CONFIG.modelConfig.compressProviderName =
+              providerName as any;
+          }
+
           return res;
         })
         .then((res: DangerConfig) => {

app/store/chat.ts

@@ -123,9 +123,16 @@ function getSummarizeModel(
   currentModel: string,
   providerName: string,
 ): string[] {
+  const configStore = useAppConfig.getState();
+  if (configStore.modelConfig.compressModel) {
+    return [
+      configStore.modelConfig.compressModel,
+      configStore.modelConfig.compressProviderName,
+    ];
+  }
+
   // if it is using gpt-* models, force to use 4o-mini to summarize
   if (currentModel.startsWith("gpt") || currentModel.startsWith("chatgpt")) {
-    const configStore = useAppConfig.getState();
     const accessStore = useAccessStore.getState();
     const allModel = collectModelsWithDefaultModel(
       configStore.models,