Merge pull request #129 from LyuLumos/main

feat: add generationConfig of Google Gemini & fix: Gemini abort bug

Author: ddiu8081

src/logics/conversation.ts

@@ -133,8 +133,10 @@ export const callProviderHandler = async(providerId: string, payload: HandlerPay
       baseUrl: payload.globalSettings?.baseUrl,
       model: payload.globalSettings?.model,
       maxTokens: payload.globalSettings?.maxTokens,
+      maxOutputTokens: payload.globalSettings?.maxOutputTokens,
       temperature: payload.globalSettings?.temperature,
-      top_p: payload.globalSettings?.top_p,
+      topP: payload.globalSettings?.topP,
+      topK: payload.globalSettings?.topK,
     },
     botSettings: payload.botSettings,
   })

src/providers/google/api.ts

@@ -3,6 +3,7 @@ export interface GoogleFetchPayload {
   stream: boolean
   body: Record<string, any>
   model?: string
+  signal?: AbortSignal
 }
 
 export const fetchChatCompletion = async(payload: GoogleFetchPayload) => {
@@ -11,6 +12,7 @@ export const fetchChatCompletion = async(payload: GoogleFetchPayload) => {
     headers: { 'Content-Type': 'application/json' },
     method: 'POST',
     body: JSON.stringify({ ...body }),
+    signal: payload.signal,
   }
   return fetch(`https://generativelanguage.googleapis.com/v1beta/models/${model}:streamGenerateContent?${stream ? 'alt=sse&' : ''}key=${apiKey}`, initOptions)
 }

src/providers/google/handler.ts

@@ -18,6 +18,11 @@ export const handleRapidPrompt: Provider['handleRapidPrompt'] = async(prompt, gl
     globalSettings: {
       ...globalSettings,
       model: 'gemini-pro',
+      temperature: 0.4,
+      maxTokens: 10240,
+      maxOutputTokens: 1024,
+      topP: 0.8,
+      topK: 1,
     },
     botSettings: {},
     prompt,
@@ -57,7 +62,14 @@ export const handleChatCompletion = async(payload: HandlerPayload, signal?: Abor
     stream,
     body: {
       contents: parseMessageList(messages),
+      generationConfig: {
+        temperature: payload.globalSettings.temperature as number,
+        maxOutputTokens: payload.globalSettings.maxOutputTokens as number,
+        topP: payload.globalSettings.topP as number,
+        topK: payload.globalSettings.topK as number,
+      }
     },
+    signal,
     model: payload.globalSettings.model as string,
   })
 

src/providers/google/index.ts

@@ -32,7 +32,47 @@ const providerGoogle = () => {
         type: 'slider',
         min: 0,
         max: 32768,
-        default: 2048,
+        default: 10240,
+        step: 1,
+      },
+      {
+        key: 'maxOutputTokens',
+        name: 'Max Output Tokens',
+        description: 'Specifies the maximum number of tokens that can be generated in the response. A token is approximately four characters. 100 tokens correspond to roughly 60-80 words.',
+        type: 'slider',
+        min: 0,
+        max: 4096,
+        default: 1024,
+        step: 1,
+      },
+      {
+        key: 'temperature',
+        name: 'Temperature',
+        description: 'The temperature controls the degree of randomness in token selection. Lower temperatures are good for prompts that require a more deterministic or less open-ended response.',
+        type: 'slider',
+        min: 0,
+        max: 1,
+        default: 0.4,
+        step: 0.01,
+      },
+      {
+        key: 'topP',
+        name: 'Top P',
+        description: 'An alternative to sampling with temperature, called nucleus sampling, where the model considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered.',
+        type: 'slider',
+        min: 0,
+        max: 1,
+        default: 0.95,
+        step: 0.01,
+      },
+      {
+        key: 'topK',
+        name: 'Top K',
+        description: 'Top K sampling chooses from the K most likely tokens.',
+        type: 'slider',
+        min: 0,
+        max: 32768,
+        default: 1,
         step: 1,
       },
       {