added chat

apps/remix-ide/src/app/plugins/remixAIPlugin.tsx

@@ -43,6 +43,7 @@ export class RemixAIPlugin extends ViewPlugin {
     } else {
       console.log('Activating RemixAIPlugin on browser')
     }
+    this.initialize()
   }
 
   async initialize(model1?:IModel, model2?:IModel, remoteModel?:IRemoteModel){
@@ -66,17 +67,17 @@ export class RemixAIPlugin extends ViewPlugin {
   async code_generation(prompt: string): Promise<any> {
     console.log('code_generation')
     if (this.isOnDesktop) {
-      return this.call(this.remixDesktopPluginName, 'code_generation', prompt)
+      return await this.call(this.remixDesktopPluginName, 'code_generation', prompt)
     } else {
-      return this.remoteInferencer.code_generation(prompt)
+      return await this.remoteInferencer.code_generation(prompt)
     }
   }
 
   async code_completion(prompt: string): Promise<any> {
     if (this.isOnDesktop) {
-      return this.call(this.remixDesktopPluginName, 'code_completion', prompt)
+      return await this.call(this.remixDesktopPluginName, 'code_completion', prompt)
     } else {
-      return this.remoteInferencer.code_completion(prompt)
+      return await this.remoteInferencer.code_completion(prompt)
     }
   }
 
@@ -85,9 +86,9 @@ export class RemixAIPlugin extends ViewPlugin {
 
     let result
     if (this.isOnDesktop) {
-      result = this.call(this.remixDesktopPluginName, 'solidity_answer', prompt)
+      result = await this.call(this.remixDesktopPluginName, 'solidity_answer', prompt)
     } else {
-      result = this.remoteInferencer.solidity_answer(prompt)
+      result = await this.remoteInferencer.solidity_answer(prompt)
     }
     this.call('terminal', 'log', { type: 'aitypewriterwarning', value: result })
 
@@ -101,7 +102,7 @@ export class RemixAIPlugin extends ViewPlugin {
       result = await this.call(this.remixDesktopPluginName, 'code_explaining', prompt)
 
     } else {
-      result = this.remoteInferencer.code_explaining(prompt)
+      result = await this.remoteInferencer.code_explaining(prompt)
     }
     if (result) this.call('terminal', 'log', { type: 'aitypewriterwarning', value: result })
   }
@@ -120,9 +121,9 @@ export class RemixAIPlugin extends ViewPlugin {
 
   async code_insertion(msg_pfx: string, msg_sfx: string): Promise<any> {
     if (this.isOnDesktop) {
-      return this.call(this.remixDesktopPluginName, 'code_insertion', msg_pfx, msg_sfx)
+      return await this.call(this.remixDesktopPluginName, 'code_insertion', msg_pfx, msg_sfx)
     } else {
-      return this.remoteInferencer.code_insertion(msg_pfx, msg_sfx)
+      return await this.remoteInferencer.code_insertion(msg_pfx, msg_sfx)
     }
   }
 

apps/remixdesktop/src/lib/InferenceServerManager.ts

@@ -4,8 +4,9 @@ import fs from 'fs';
 import axios from "axios";
 import { EventEmitter } from 'events';
 import { ICompletions, IModel, IParams, InsertionParams,
-  CompletionParams, GenerationParams, ModelType,
-  IStreamResponse } from "../../../../libs/remix-ai-core/src/index"
+  CompletionParams, GenerationParams, ModelType, AIRequestType,
+  IStreamResponse, ChatHistory,
+  buildSolgptPromt } from "../../../../libs/remix-ai-core/src/index"
 
 class ServerStatusTimer {
   private intervalId: NodeJS.Timeout | null = null;
@@ -127,7 +128,7 @@ export class InferenceManager implements ICompletions {
       console.log('Inference server not running')
       InferenceManager.instance = null
       this.stateTimer.interval += this.stateTimer.interval
-      
+
       if (this.stateTimer.interval >= 60000) {
         // attempt to restart the server
         console.log('Attempting to restart the server')
@@ -257,17 +258,27 @@ export class InferenceManager implements ICompletions {
     }
   }
 
-  private async _makeInferenceRequest(endpoint, payload){
+  private async _makeInferenceRequest(endpoint, payload, rType:AIRequestType){
     try {
       this.event.emit('onInference')
       const options = { headers: { 'Content-Type': 'application/json', } }
       const response = await axios.post(`${this.inferenceURL}/${endpoint}`, payload, options)
+
+      const userPrompt = payload[Object.keys(payload)[0]]
       this.event.emit('onInferenceDone')
 
+      console.log('response', response)
+      console.log('userprompt', userPrompt)
+      console.log('chat history:', ChatHistory.getHistory())
+
       if (response.data?.generatedText) {
+        if (rType === AIRequestType.GENERAL) {
+          ChatHistory.pushHistory(userPrompt, response.data.generatedText)
+        }
         return response.data.generatedText
       } else { return "" }
     } catch (error) {
+      ChatHistory.clearHistory()
       console.error('Error making request to Inference server:', error.message);
     }
   }
@@ -286,22 +297,31 @@ export class InferenceManager implements ICompletions {
         }
         , responseType: 'stream' });
 
+      const userPrompt = payload[Object.keys(payload)[0]]
+      let resultText = ""
       response.data.on('data', (chunk: Buffer) => {
         try {
           const parsedData = JSON.parse(chunk.toString());
           if (parsedData.isGenerating) {
             this.event.emit('onStreamResult', parsedData.generatedText);
+            resultText = resultText + parsedData.generatedText
           } else {
+            resultText = resultText + parsedData.generatedText
+
+            // no additional check for streamed results
+            ChatHistory.pushHistory(userPrompt, resultText)
             return parsedData.generatedText
           }
 
         } catch (error) {
+          ChatHistory.clearHistory()
           console.error('Error parsing JSON:', error);
         }
       });
 
-      return "" // return empty string for now as payload already handled in event
+      return "" // return empty string for now as payload is/will be handled in event
     } catch (error) {
+      ChatHistory.clearHistory()
       console.error('Error making stream request to Inference server:', error.message);
     }
     finally {
@@ -310,6 +330,7 @@ export class InferenceManager implements ICompletions {
   }
 
   private async _makeRequest(endpoint, payload){
+    // make a simple request to the inference server
     try {
       const options = { headers: { 'Content-Type': 'application/json', } }
       const response = await axios.post(`${this.inferenceURL}/${endpoint}`, payload, options)
@@ -329,7 +350,7 @@ export class InferenceManager implements ICompletions {
 
     // as of now no prompt required
     const payload = { context_code: context, ...params }
-    return this._makeInferenceRequest('code_completion', payload)
+    return this._makeInferenceRequest('code_completion', payload, AIRequestType.COMPLETION)
   }
 
   async code_insertion(msg_pfx: string, msg_sfx: string, params:IParams=InsertionParams): Promise<any> {
@@ -338,7 +359,7 @@ export class InferenceManager implements ICompletions {
       return
     }
     const payload = { code_pfx:msg_pfx, code_sfx:msg_sfx, ...params }
-    return this._makeInferenceRequest('code_insertion', payload)
+    return this._makeInferenceRequest('code_insertion', payload, AIRequestType.COMPLETION)
 
   }
 
@@ -347,7 +368,7 @@ export class InferenceManager implements ICompletions {
       console.log('model not ready yet')
       return
     }
-    return this._makeInferenceRequest('code_generation', { prompt, ...params })
+    return this._makeInferenceRequest('code_generation', { prompt, ...params }, AIRequestType.GENERAL)
   }
 
   async code_explaining(code:string, context:string, params:IParams=GenerationParams): Promise<any> {
@@ -358,7 +379,7 @@ export class InferenceManager implements ICompletions {
     if (GenerationParams.stream_result) {
       return this._streamInferenceRequest('code_explaining', { code, context, ...params })
     } else {
-      return this._makeInferenceRequest('code_explaining', { code, context, ...params })
+      return this._makeInferenceRequest('code_explaining', { code, context, ...params }, AIRequestType.GENERAL)
     }
   }
 
@@ -370,19 +391,27 @@ export class InferenceManager implements ICompletions {
     if (GenerationParams.stream_result) {
       return this._streamInferenceRequest('error_explaining', { prompt, ...params })
     } else {
-      return this._makeInferenceRequest('error_explaining', { prompt, ...params })
+      return this._makeInferenceRequest('error_explaining', { prompt, ...params }, AIRequestType.GENERAL)
     }
   }
 
-  async solidity_answer(prompt: string, params:IParams=GenerationParams): Promise<any> {
+  async solidity_answer(userPrompt: string, params:IParams=GenerationParams): Promise<any> {
     if (!this.isReady) {
       console.log('model not ready yet')
       return
     }
+    let modelOP = undefined
+    for (const model of this.selectedModels) {
+      if (model?.modelOP) {
+        modelOP = model.modelOP
+      }
+    }
+    const prompt = buildSolgptPromt(userPrompt, this.selectedModels[0]?.modelOP)
+
     if (GenerationParams.stream_result) {
       return this._streamInferenceRequest('solidity_answer', { prompt, ...params })
     } else {
-      return this._makeInferenceRequest('solidity_answer', { prompt, ...params })
+      return this._makeInferenceRequest('solidity_answer', { prompt, ...params }, AIRequestType.GENERAL)
     }
   }
 

libs/remix-ai-core/src/index.ts

@@ -6,13 +6,14 @@ import { IModel, IModelResponse, IModelRequest, InferenceModel, ICompletions,
 import { ModelType } from './types/constants'
 import { DefaultModels, InsertionParams, CompletionParams, GenerationParams } from './types/models'
 import { getCompletionPrompt, getInsertionPrompt } from './prompts/completionPrompts'
-import { PromptBuilder } from './prompts/promptBuilder'
+import { buildSolgptPromt, PromptBuilder } from './prompts/promptBuilder'
 import { RemoteInferencer } from './inferencers/remote/remoteInference'
+import { ChatHistory } from './prompts/chat'
 
 export {
   IModel, IModelResponse, IModelRequest, InferenceModel,
   ModelType, DefaultModels, ICompletions, IParams, IRemoteModel,
-  getCompletionPrompt, getInsertionPrompt, IStreamResponse,
+  getCompletionPrompt, getInsertionPrompt, IStreamResponse, buildSolgptPromt,
   RemoteInferencer, InsertionParams, CompletionParams, GenerationParams,
-  ChatEntry, AIRequestType, RemoteBackendOPModel, PromptBuilder
+  ChatEntry, AIRequestType, RemoteBackendOPModel, ChatHistory
 }

libs/remix-ai-core/src/inferencers/remote/remoteInference.ts

@@ -1,14 +1,14 @@
-import { ICompletions, IParams, ChatEntry, AIRequestType, RemoteBackendOPModel } from "../../types/types";
-import { PromptBuilder } from "../../prompts/promptBuilder";
+import { ICompletions, IParams, AIRequestType, RemoteBackendOPModel } from "../../types/types";
+import { buildSolgptPromt } from "../../prompts/promptBuilder";
 import axios from "axios";
 import EventEmitter from "events";
+import { ChatHistory } from "../../prompts/chat";
 
 const defaultErrorMessage = `Unable to get a response from AI server`
 
 export class RemoteInferencer implements ICompletions {
   api_url: string
   completion_url: string
-  solgpt_chat_history:ChatEntry[]
   max_history = 7
   model_op = RemoteBackendOPModel.DEEPSEEK
   event: EventEmitter
@@ -16,20 +16,14 @@ export class RemoteInferencer implements ICompletions {
   constructor(apiUrl?:string, completionUrl?:string) {
     this.api_url = apiUrl!==undefined ? apiUrl: "https://solcoder.remixproject.org"
     this.completion_url = completionUrl!==undefined ? completionUrl : "https://completion.remixproject.org"
-    this.solgpt_chat_history = []
     this.event = new EventEmitter()
   }
 
-  private pushChatHistory(prompt, result){
-    const chat:ChatEntry = [prompt, result.data[0]]
-    this.solgpt_chat_history.push(chat)
-    if (this.solgpt_chat_history.length > this.max_history){this.solgpt_chat_history.shift()}
-  }
-
   private async _makeRequest(data, rType:AIRequestType){
     this.event.emit("onInference")
     const requesURL = rType === AIRequestType.COMPLETION ? this.completion_url : this.api_url
-    console.log("requesting on ", requesURL, rType, data.data[1])
+    const userPrompt = data.data[0]
+    console.log('userPrompt reuesting...')
 
     try {
       const result = await axios(requesURL, {
@@ -51,7 +45,7 @@ export class RemoteInferencer implements ICompletions {
       case AIRequestType.GENERAL:
         if (result.statusText === "OK") {
           const resultText = result.data.data[0]
-          this.pushChatHistory(prompt, resultText)
+          ChatHistory.pushHistory(userPrompt, resultText)
           return resultText
         } else {
           return defaultErrorMessage
@@ -59,7 +53,8 @@ export class RemoteInferencer implements ICompletions {
       }
 
     } catch (e) {
-      this.solgpt_chat_history = []
+      ChatHistory.clearHistory()
+      console.error('Error making request to Inference server:', e.message)
       return e
     }
     finally {
@@ -71,32 +66,37 @@ export class RemoteInferencer implements ICompletions {
     try {
       this.event.emit('onInference')
       const requesURL = rType === AIRequestType.COMPLETION ? this.completion_url : this.api_url
-      const options = { headers: { 'Content-Type': 'application/json', "Accept": "text/event-stream" } }
+      const userPrompt = data.data[0]
       const response = await axios({
         method: 'post',
-        url:  rType === AIRequestType.COMPLETION ? this.completion_url : this.api_url,
+        url:  requesURL,
         data: data,
         headers: { 'Content-Type': 'application/json', "Accept": "text/event-stream" },
         responseType: 'stream'
       });
 
+      let resultText = ""
       response.data.on('data', (chunk: Buffer) => {
         try {
           const parsedData = JSON.parse(chunk.toString());
           if (parsedData.isGenerating) {
             this.event.emit('onStreamResult', parsedData.generatedText);
+            resultText = resultText + parsedData.generatedText
           } else {
+            // stream generation is complete
+            resultText = resultText + parsedData.generatedText
+            ChatHistory.pushHistory(userPrompt, resultText)
             return parsedData.generatedText
           }
-
         } catch (error) {
           console.error('Error parsing JSON:', error);
+          ChatHistory.clearHistory()
         }
-
       });
 
       return "" // return empty string for now as handled in event
     } catch (error) {
+      ChatHistory.clearHistory()
       console.error('Error making stream request to Inference server:', error.message);
     }
     finally {
@@ -125,7 +125,7 @@ export class RemoteInferencer implements ICompletions {
   }
 
   async solidity_answer(prompt): Promise<any> {
-    const main_prompt = this._build_solgpt_promt(prompt)
+    const main_prompt = buildSolgptPromt(prompt, this.model_op)
     const payload = { "data":[main_prompt, "solidity_answer", false,2000,0.9,0.8,50]}
     return this._makeRequest(payload, AIRequestType.GENERAL)
   }
@@ -139,19 +139,4 @@ export class RemoteInferencer implements ICompletions {
     const payload = { "data":[prompt, "error_explaining", false,2000,0.9,0.8,50]}
     return this._makeRequest(payload, AIRequestType.GENERAL)
   }
-
-  private _build_solgpt_promt(user_promt:string){
-    if (this.solgpt_chat_history.length === 0){
-      return user_promt
-    } else {
-      let new_promt = ""
-      for (const [question, answer] of this.solgpt_chat_history) {
-        new_promt += PromptBuilder(question.split('sol-gpt')[1], answer, this.model_op)
-      }
-      // finaly
-      new_promt = "sol-gpt " + new_promt + PromptBuilder(user_promt.split('sol-gpt')[1], "", this.model_op)
-      return new_promt
-    }
-  }
-
 }

libs/remix-ai-core/src/prompts/chat.ts

@@ -0,0 +1,21 @@
+import { ChatEntry } from "../types/types"
+
+export abstract class ChatHistory{
+
+  private static chatEntries:ChatEntry[] = []
+  static queuSize:number = 7 // change the queue size wrt the GPU size
+
+  public static pushHistory(prompt, result){
+    const chat:ChatEntry = [prompt, result]
+    this.chatEntries.push(chat)
+    if (this.chatEntries.length > this.queuSize){this.chatEntries.shift()}
+  }
+
+  public static getHistory(){
+    return this.chatEntries
+  }
+
+  public static clearHistory(){
+    this.chatEntries = []
+  }
+}

libs/remix-ai-core/src/prompts/promptBuilder.ts

@@ -1,7 +1,28 @@
 import { RemoteBackendOPModel } from "../types/types"
+import { ChatHistory } from "./chat"
 
 export const PromptBuilder = (inst, answr, modelop) => {
   if (modelop === RemoteBackendOPModel.CODELLAMA) return ""
   if (modelop === RemoteBackendOPModel.DEEPSEEK) return "\n### INSTRUCTION:\n" + inst + "\n### RESPONSE:\n" + answr
   if (modelop === RemoteBackendOPModel.MISTRAL) return ""
+}
+
+export const buildSolgptPromt = (userPrompt:string, modelOP:RemoteBackendOPModel) => {
+  if (modelOP === undefined) {
+    console.log('WARNING: modelOP is undefined. Provide a valide model OP for chat history')
+    return userPrompt
+  }
+  if (ChatHistory.getHistory().length === 0){
+    return userPrompt
+  } else {
+    let newPrompt = ""
+    for (const [question, answer] of ChatHistory.getHistory()) {
+      if (question.startsWith('sol-gpt')) newPrompt += PromptBuilder(question.split('sol-gpt')[1], answer, modelOP)
+      else if (question.startsWith('gpt')) newPrompt += PromptBuilder(question.split('sol-gpt')[1], answer, modelOP)
+      else newPrompt += PromptBuilder(question, answer, modelOP)
+    }
+    // finaly
+    newPrompt = "sol-gpt " + newPrompt + PromptBuilder(userPrompt.split('sol-gpt')[1], "", modelOP)
+    return newPrompt
+  }
 }

libs/remix-ai-core/src/types/models.ts

@@ -2,12 +2,14 @@
 // create a function getModels returning a list of all supported models
 // create a function getModel returning a model by its name
 
-import { IModel, IParams } from './types';
+import { IModel, IParams, RemoteBackendOPModel } from './types';
 import { ModelType } from './constants';
 
+
 const DefaultModels = (): IModel[] => {
   const model1:IModel = {
     name: 'DeepSeek',
+    modelOP: RemoteBackendOPModel.DEEPSEEK,
     task: 'text-generation',
     modelName: 'deepseek-coder-1.3b-instruct.gguf',
     downloadUrl: 'https://huggingface.co/TheBloke/deepseek-coder-1.3b-instruct-GGUF/resolve/main/deepseek-coder-1.3b-instruct.Q4_K_M.gguf?download=true',
@@ -16,6 +18,7 @@ const DefaultModels = (): IModel[] => {
   };
   const model2: IModel = {
     name: 'DeepSeek',
+    modelOP: RemoteBackendOPModel.DEEPSEEK,
     task: 'text-generation',
     modelName: 'deepseek-coder-6.7b-instruct.gguf',
     downloadUrl: 'https://huggingface.co/TheBloke/deepseek-coder-6.7B-instruct-GGUF/resolve/main/deepseek-coder-6.7b-instruct.Q4_K_M.gguf?download=true',
@@ -24,6 +27,7 @@ const DefaultModels = (): IModel[] => {
   };
   const model3: IModel = {
     name: 'DeepSeekTransformer',
+    modelOP: RemoteBackendOPModel.DEEPSEEK,
     task: 'text-generation',
     modelName: 'Xenova/deepseek-coder-1.3b-base',
     downloadUrl: 'Xenova/deepseek-coder-1.3b-base',
@@ -32,6 +36,7 @@ const DefaultModels = (): IModel[] => {
   };
   const model4: IModel = {
     name: 'DeepSeek',
+    modelOP: RemoteBackendOPModel.DEEPSEEK,
     task: 'text-generation',
     modelName: 'deepseek-coder-1.3b-base.gguf',
     downloadUrl: 'https://huggingface.co/TheBloke/deepseek-coder-1.3b-base-GGUF/resolve/main/deepseek-coder-1.3b-base.Q4_K_M.gguf?download=true',
@@ -41,6 +46,7 @@ const DefaultModels = (): IModel[] => {
 
   const model5: IModel = {
     name: 'DeepSeek',
+    modelOP: RemoteBackendOPModel.DEEPSEEK,
     task: 'text-generation',
     modelName: 'deepseek-coder-6.7B-base-GGUF',
     downloadUrl: 'https://huggingface.co/TheBloke/deepseek-coder-6.7B-base-GGUF/resolve/main/deepseek-coder-6.7b-base.Q4_K_M.gguf?download=true',
@@ -50,6 +56,7 @@ const DefaultModels = (): IModel[] => {
 
   const model6: IModel = {
     name: 'DeepSeek',
+    modelOP: RemoteBackendOPModel.DEEPSEEK,
     task: 'text-generation',
     modelName: 'DeepSeek-Coder-V2-Lite-Base.Q2_K.gguf',
     downloadUrl: 'https://huggingface.co/QuantFactory/DeepSeek-Coder-V2-Lite-Base-GGUF/resolve/main/DeepSeek-Coder-V2-Lite-Base.Q2_K.gguf?download=true',
@@ -87,7 +94,7 @@ const GenerationParams:IParams = {
   topK: 40,
   topP: 0.92,
   max_new_tokens: 2000,
-  stream_result: true,
+  stream_result: false,
 }
 
 export { DefaultModels, CompletionParams, InsertionParams, GenerationParams }

libs/remix-ai-core/src/types/types.ts

@@ -18,6 +18,8 @@ export interface IModel {
   modelType: ModelType;
   modelReqs: IModelRequirements;
   downloadPath?: string;
+  modelOP?: RemoteBackendOPModel;
+
 }
 export interface IRemoteModel {
   completionUrl: string;