Chat Stream Integration (#52)

* rm: axios dependency

* add: stream parsing for normal style

* fix: empty string problem

* add: stream for embedded prompts

* update: version increment

src/utils/messageNormal.ts

@@ -1,6 +1,6 @@
 import { Message } from 'discord.js'
 import { ChatResponse, Ollama } from 'ollama'
-import { UserMessage } from './events.js'
+import { ChatParams, UserMessage, streamResponse, blockResponse } from './index.js'
 import { Queue } from '../queues/queue.js'
 
 /**
@@ -16,38 +16,49 @@ export async function normalMessage(
         channel: string,
         model: string
     },
-    msgHist: Queue<UserMessage>
-) {
+    msgHist: Queue<UserMessage>,
+    stream: boolean
+): Promise<string> {
     // bot's respnse
-    let response: ChatResponse
+    let response: ChatResponse | AsyncGenerator<ChatResponse, any, unknown>
+    let result: string = ''
 
     await message.channel.send('Generating Response . . .').then(async sentMessage => {
         try {
-            // Attempt to query model for message
-            response = await ollama.chat({
+            const params: ChatParams = {
                 model: tokens.model,
-                messages: msgHist.getItems(),
-                options: {
-                    num_thread: 8, // remove if optimization needed further
-                    mirostat: 1,
-                    mirostat_tau: 2.0,
-                    top_k: 70
-                },
-                stream: false
-            })
+                ollama: ollama,
+                msgHist: msgHist.getItems()
+            } 
 
-            // check if message length > discord max for normal messages
-            if (response.message.content.length > 2000) {
-                sentMessage.edit(response.message.content.slice(0, 2000))
-                message.channel.send(response.message.content.slice(2000))
-            } else // edit the 'generic' response to new message
-                sentMessage.edit(response.message.content)            
+            // run query based on stream preference, true = stream, false = block
+            if (stream) {
+                response = await streamResponse(params)
+                for await (const portion of response) {
+                    // append token to message
+                    result += portion.message.content
+
+                    // resent current output, THIS WILL BE SLOW due to discord limits!
+                    sentMessage.edit(result || 'No Content Yet...')
+                }
+            }
+            else {
+                response = await blockResponse(params)
+                result = response.message.content
+
+                // check if message length > discord max for normal messages
+                if (result.length > 2000) {
+                    sentMessage.edit(result.slice(0, 2000))
+                    message.channel.send(result.slice(2000))
+                } else // edit the 'generic' response to new message
+                    sentMessage.edit(result)
+            }            
         } catch(error: any) {
             console.log(`[Util: messageNormal] Error creating message: ${error.message}`)
             sentMessage.edit(`**Response generation failed.**\n\nReason: ${error.message}`)
         }
     })
 
-    // Hope there is a response, force client to believe
-    return response!!
-}
+    // return the string representation of response
+    return result
+}