add prompt command and refactor context logic

src/bot.ts

@@ -516,12 +516,24 @@ const logErrorHandler = (ex: any): void => {
 bot.command('new', async (ctx) => {
   writeCommandLog(ctx as OnMessageContext).catch(logErrorHandler)
   await openAiBot.onStop(ctx as OnMessageContext)
+  await claudeBot.onStop(ctx as OnMessageContext) // any Bot with 'llms' as sessionKey works.
   return await ctx.reply('Chat history reseted', {
     parse_mode: 'Markdown',
     message_thread_id: ctx.message?.message_thread_id
   })
 })
 
+bot.command('prompt', async (ctx) => {
+  const context = ctx.match
+  if (context) {
+    ctx.session.currentPrompt = context
+  }
+  await ctx.reply(`Prompt set to: _${ctx.session.currentPrompt}_`, {
+    parse_mode: 'Markdown',
+    message_thread_id: ctx.message?.message_thread_id
+  })
+})
+
 bot.command('more', async (ctx) => {
   writeCommandLog(ctx as OnMessageContext).catch(logErrorHandler)
   return await ctx.reply(commandsHelpText.more, {

src/helpers.ts

@@ -56,6 +56,7 @@ export function createInitialSessionData (): BotSessionData {
       isVoiceForwardingEnabled: config.voiceMemo.isVoiceForwardingEnabled
     },
     currentModel: LlmModelsEnum.GPT_4O,
+    currentPrompt: config.openAi.chatGpt.chatCompletionContext,
     lastBroadcast: ''
   }
 }

src/modules/llms/api/athropic.ts

@@ -24,11 +24,12 @@ const API_ENDPOINT = config.llms.apiEndpoint // 'http://127.0.0.1:5000' // confi
 export const anthropicCompletion = async (
   conversation: ChatConversation[],
   model = LlmModelsEnum.CLAUDE_3_OPUS,
+  ctx: OnMessageContext | OnCallBackQueryData,
   parameters?: ModelParameters
 ): Promise<LlmCompletion> => {
   logger.info(`Handling ${model} completion`)
   parameters = parameters ?? {
-    system: config.openAi.chatGpt.chatCompletionContext,
+    system: ctx.session.currentPrompt,
     max_tokens: +config.openAi.chatGpt.maxTokens
   }
   const data = {
@@ -68,11 +69,12 @@ export const anthropicCompletion = async (
 export const xaiCompletion = async (
   conversation: ChatConversation[],
   model = LlmModelsEnum.GROK,
+  ctx: OnMessageContext | OnCallBackQueryData,
   parameters?: ModelParameters
 ): Promise<LlmCompletion> => {
   logger.info(`Handling ${model} completion`)
   parameters = parameters ?? {
-    system: config.openAi.chatGpt.chatCompletionContext,
+    system: ctx.session.currentPrompt,
     max_tokens: +config.openAi.chatGpt.maxTokens
   }
   const data = {
@@ -119,7 +121,7 @@ export const anthropicStreamCompletion = async (
 ): Promise<LlmCompletion> => {
   logger.info(`Handling ${model} stream completion`)
   parameters = parameters ?? {
-    system: config.openAi.chatGpt.chatCompletionContext,
+    system: ctx.session.currentPrompt,
     max_tokens: +config.openAi.chatGpt.maxTokens
   }
   const data = {
@@ -217,11 +219,12 @@ export const anthropicStreamCompletion = async (
 export const toolsChatCompletion = async (
   conversation: ChatConversation[],
   model = LlmModelsEnum.CLAUDE_3_OPUS,
+  ctx: OnMessageContext | OnCallBackQueryData,
   parameters?: ModelParameters
 ): Promise<LlmCompletion> => {
   logger.info(`Handling ${model} completion`)
   parameters = parameters ?? {
-    system: config.openAi.chatGpt.chatCompletionContext,
+    system: ctx.session.currentPrompt,
     max_tokens: +config.openAi.chatGpt.maxTokens
   }
   const input = {

src/modules/llms/api/openai.ts

@@ -81,25 +81,26 @@ export async function alterGeneratedImg (
 
 type ConversationOutput = Omit<ChatConversation, 'timestamp' | 'model' | 'id' | 'author' | 'numSubAgents'>
 
-const prepareConversation = (conversation: ChatConversation[], model: string): ConversationOutput[] => {
+const prepareConversation = (conversation: ChatConversation[], model: string, ctx: OnMessageContext | OnCallBackQueryData): ConversationOutput[] => {
   const messages = conversation.filter(c => c.model === model).map(m => { return { content: m.content, role: m.role } })
   if (messages.length !== 1 || model === LlmModelsEnum.O1) {
     return messages
   }
   const systemMessage = {
     role: 'system',
-    content: config.openAi.chatGpt.chatCompletionContext
+    content: ctx.session.currentPrompt
   }
   return [systemMessage, ...messages]
 }
 
 export async function chatCompletion (
   conversation: ChatConversation[],
   model = config.openAi.chatGpt.model,
+  ctx: OnMessageContext | OnCallBackQueryData,
   limitTokens = true,
   parameters?: ModelParameters
 ): Promise<LlmCompletion> {
-  const messages = prepareConversation(conversation, model)
+  const messages = prepareConversation(conversation, model, ctx)
   parameters = parameters ?? {
     max_completion_tokens: config.openAi.chatGpt.maxTokens,
     temperature: config.openAi.dalle.completions.temperature
@@ -139,15 +140,15 @@ export async function chatCompletion (
 
 export const streamChatCompletion = async (
   conversation: ChatConversation[],
-  ctx: OnMessageContext | OnCallBackQueryData,
   model = LlmModelsEnum.GPT_4,
+  ctx: OnMessageContext | OnCallBackQueryData,
   msgId: number,
   limitTokens = true,
   parameters?: ModelParameters
 ): Promise<LlmCompletion> => {
   let completion = ''
   let wordCountMinimum = 2
-  const messages = prepareConversation(conversation, model)
+  const messages = prepareConversation(conversation, model, ctx)
   parameters = parameters ?? {
     max_completion_tokens: config.openAi.chatGpt.maxTokens,
     temperature: config.openAi.dalle.completions.temperature || 0.8
@@ -322,10 +323,10 @@ export const streamChatVisionCompletion = async (
   }
 }
 
-export async function improvePrompt (promptText: string, model: string): Promise<string> {
+export async function improvePrompt (promptText: string, model: string, ctx: OnMessageContext | OnCallBackQueryData): Promise<string> {
   const prompt = `Improve this picture description using max 100 words and don't add additional text to the image: ${promptText} `
   const conversation = [{ role: 'user', content: prompt, timestamp: Date.now() }]
-  const response = await chatCompletion(conversation, model)
+  const response = await chatCompletion(conversation, model, ctx)
   return response.completion?.content as string ?? ''
 }
 

src/modules/llms/api/vertex.ts

@@ -22,10 +22,12 @@ const logger = pino({
 export const vertexCompletion = async (
   conversation: ChatConversation[],
   model = config.llms.model,
+  ctx: OnMessageContext | OnCallBackQueryData,
   parameters?: ModelParameters
 ): Promise<LlmCompletion> => {
   const data = {
     model,
+    system: ctx.session.currentPrompt,
     stream: false,
     messages: conversation.filter(c => c.model === model)
       .map((msg) => {
@@ -71,10 +73,9 @@ export const vertexStreamCompletion = async (
   parameters?: ModelParameters
 ): Promise<LlmCompletion> => {
   parameters = parameters ?? {
-    system: config.openAi.chatGpt.chatCompletionContext,
+    system: ctx.session.currentPrompt,
     max_tokens: +config.openAi.chatGpt.maxTokens
   }
-
   const data = {
     model,
     stream: true, // Set stream to true to receive the completion as a stream

src/modules/llms/claudeBot.ts

@@ -42,6 +42,9 @@ export class ClaudeBot extends LlmsBase {
     msgId: number,
     limitTokens: boolean,
     parameters?: ModelParameters): Promise<LlmCompletion> {
+    if (parameters) {
+      parameters.system = ctx.session.currentPrompt
+    }
     return await anthropicStreamCompletion(
       conversation,
       model,
@@ -55,13 +58,17 @@ export class ClaudeBot extends LlmsBase {
   async chatCompletion (
     conversation: ChatConversation[],
     model: ModelVersion,
+    ctx: OnMessageContext | OnCallBackQueryData,
     hasTools: boolean,
     parameters?: ModelParameters
   ): Promise<LlmCompletion> {
+    if (parameters) {
+      parameters.system = ctx.session.currentPrompt
+    }
     if (hasTools) {
-      return await toolsChatCompletion(conversation, model, parameters)
+      return await toolsChatCompletion(conversation, model, ctx, parameters)
     }
-    return await anthropicCompletion(conversation, model, parameters)
+    return await anthropicCompletion(conversation, model, ctx, parameters)
   }
 
   public async onEvent (ctx: OnMessageContext | OnCallBackQueryData): Promise<void> {

src/modules/llms/dalleBot.ts

@@ -133,18 +133,19 @@ export class DalleBot extends LlmsBase {
   ): Promise<LlmCompletion> {
     return await streamChatCompletion(
       conversation,
-      ctx,
       model,
+      ctx,
       msgId,
       true // telegram messages has a character limit
     )
   }
 
   async chatCompletion (
     conversation: ChatConversation[],
-    model: ModelVersion
+    model: ModelVersion,
+    ctx: OnMessageContext | OnCallBackQueryData
   ): Promise<LlmCompletion> {
-    return await chatCompletion(conversation, model)
+    return await chatCompletion(conversation, model, ctx)
   }
 
   hasPrefix (prompt: string): string {

src/modules/llms/llmsBase.ts

@@ -120,6 +120,7 @@ export abstract class LlmsBase implements PayableBot {
   protected abstract chatCompletion (
     conversation: ChatConversation[],
     model: ModelVersion,
+    ctx: OnMessageContext | OnCallBackQueryData,
     usesTools: boolean,
     parameters?: ModelParameters
   ): Promise<LlmCompletion>
@@ -373,7 +374,7 @@ export abstract class LlmsBase implements PayableBot {
         }
       } else {
         const parameters = this.modelManager.getModelParameters(model)
-        const response = await this.chatCompletion(conversation, model, usesTools, parameters)
+        const response = await this.chatCompletion(conversation, model, ctx, usesTools, parameters)
         conversation.push({
           role: 'assistant',
           content: response.completion?.content ?? '',
@@ -406,7 +407,7 @@ export abstract class LlmsBase implements PayableBot {
     ).message_id
     ctx.chatAction = 'typing'
     const parameters = this.modelManager.getModelParameters(model)
-    const response = await this.chatCompletion(conversation, model, usesTools, parameters)
+    const response = await this.chatCompletion(conversation, model, ctx, usesTools, parameters)
     if (response.completion) {
       if (model === this.modelsEnum.O1) {
         const msgs = splitTelegramMessage(response.completion.content as string)

src/modules/llms/openaiBot.ts

@@ -73,8 +73,8 @@ export class OpenAIBot extends LlmsBase {
   ): Promise<LlmCompletion> {
     return await streamChatCompletion(
       conversation,
-      ctx,
       model,
+      ctx,
       msgId,
       true, // telegram messages has a character limit
       parameters
@@ -84,10 +84,11 @@ export class OpenAIBot extends LlmsBase {
   async chatCompletion (
     conversation: ChatConversation[],
     model: ModelVersion,
+    ctx: OnMessageContext | OnCallBackQueryData,
     usesTools: boolean,
     parameters?: ModelParameters
   ): Promise<LlmCompletion> {
-    return await chatCompletion(conversation, model, model !== this.modelsEnum.O1, parameters) // limitTokens doesn't apply for o1-preview
+    return await chatCompletion(conversation, model, ctx, model !== this.modelsEnum.O1, parameters) // limitTokens doesn't apply for o1-preview
   }
 
   hasPrefix (prompt: string): string {

src/modules/llms/utils/llmsData.ts

@@ -223,25 +223,25 @@ export const llmData: LLMData = {
     },
     claude: {
       defaultParameters: {
-        system: config.openAi.chatGpt.chatCompletionContext,
+        // system: config.openAi.chatGpt.chatCompletionContext,
         max_tokens: +config.openAi.chatGpt.maxTokens
       }
     },
     xai: {
       defaultParameters: {
-        system: config.openAi.chatGpt.chatCompletionContext,
+        // system: config.openAi.chatGpt.chatCompletionContext,
         max_tokens: +config.openAi.chatGpt.maxTokens
       }
     },
     vertex: {
       defaultParameters: {
-        system: config.openAi.chatGpt.chatCompletionContext,
+        // system: config.openAi.chatGpt.chatCompletionContext,
         max_tokens: +config.openAi.chatGpt.maxTokens
       }
     },
     luma: {
       defaultParameters: {
-        system: config.openAi.chatGpt.chatCompletionContext,
+        // system: config.openAi.chatGpt.chatCompletionContext,
         max_tokens: +config.openAi.chatGpt.maxTokens
       }
     }

src/modules/llms/vertexBot.ts

@@ -46,6 +46,9 @@ export class VertexBot extends LlmsBase {
     msgId: number,
     limitTokens: boolean,
     parameters?: ModelParameters): Promise<LlmCompletion> {
+    if (parameters) {
+      parameters.system = ctx.session.currentPrompt
+    }
     return await vertexStreamCompletion(conversation,
       model,
       ctx,
@@ -58,10 +61,14 @@ export class VertexBot extends LlmsBase {
   async chatCompletion (
     conversation: ChatConversation[],
     model: ModelVersion,
+    ctx: OnMessageContext | OnCallBackQueryData,
     usesTools: boolean,
     parameters?: ModelParameters
   ): Promise<LlmCompletion> {
-    return await vertexCompletion(conversation, model, parameters)
+    if (parameters) {
+      parameters.system = ctx.session.currentPrompt
+    }
+    return await vertexCompletion(conversation, model, ctx, parameters)
   }
 
   public async onEvent (ctx: OnMessageContext | OnCallBackQueryData): Promise<void> {

src/modules/llms/xaiBot.ts

@@ -52,10 +52,14 @@ export class XaiBot extends LlmsBase {
   async chatCompletion (
     conversation: ChatConversation[],
     model: ModelVersion,
+    ctx: OnMessageContext | OnCallBackQueryData,
     hasTools: boolean,
     parameters?: ModelParameters
   ): Promise<LlmCompletion> {
-    return await xaiCompletion(conversation, model, parameters)
+    if (parameters) {
+      parameters.system = ctx.session.currentPrompt
+    }
+    return await xaiCompletion(conversation, model, ctx, parameters)
   }
 
   public async onEvent (ctx: OnMessageContext | OnCallBackQueryData): Promise<void> {

src/modules/types.ts

@@ -187,6 +187,7 @@ export interface BotSessionData {
   subagents: SubagentSessionData
   dalle: ImageGenSessionData
   currentModel: ModelVersion
+  currentPrompt: string
   lastBroadcast: string
   voiceMemo: VoiceMemoSessionData
 }

src/modules/voice-to-voice-gpt/index.ts

@@ -64,7 +64,7 @@ export class VoiceToVoiceGPTBot implements PayableBot {
     fs.rmSync(filename)
 
     const conversation = [{ role: 'user', content: resultText, timestamp: Date.now() }]
-    const response = await chatCompletion(conversation, LlmModelsEnum.GPT_35_TURBO)
+    const response = await chatCompletion(conversation, LlmModelsEnum.GPT_35_TURBO, ctx)
 
     const voiceResult = await generateVoiceFromText(response.completion?.content as string)
     // const voiceResult = await gcTextToSpeedClient.ssmlTextToSpeech({ text: response.completion, ssmlGender: 'MALE', languageCode: 'en-US' })