Fixed LLM multi message responses

package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@ztimson/ai-utils",
-	"version": "0.1.19",
+	"version": "0.1.22",
 	"description": "AI Utility library",
 	"author": "Zak Timson",
 	"license": "MIT",

src/antrhopic.ts

@@ -1,5 +1,5 @@
 import {Anthropic as anthropic} from '@anthropic-ai/sdk';
-import {findByProp, objectMap, JSONSanitize, JSONAttemptParse} from '@ztimson/utils';
+import {findByProp, objectMap, JSONSanitize, JSONAttemptParse, deepCopy} from '@ztimson/utils';
 import {Ai} from './ai.ts';
 import {LLMMessage, LLMRequest} from './llm.ts';
 import {AbortablePromise, LLMProvider} from './provider.ts';
@@ -52,6 +52,7 @@ export class Anthropic extends LLMProvider {
 		const controller = new AbortController();
 		const response = new Promise<any>(async (res, rej) => {
 			let history = this.fromStandard([...options.history || [], {role: 'user', content: message, timestamp: Date.now()}]);
+			const original = deepCopy(history);
 			if(options.compress) history = await this.ai.llm.compress(<any>history, options.compress.max, options.compress.min, options);
 			const requestParams: any = {
 				model: options.model || this.model,
@@ -72,12 +73,15 @@ export class Anthropic extends LLMProvider {
 				stream: !!options.stream,
 			};
 
-			let resp: any;
-			const loopMessages: any[] = [];
+			let resp: any, isFirstMessage = true;
+			const assistantMessages: string[] = [];
 			do {
 				resp = await this.client.messages.create(requestParams);
+
+				// Streaming mode
 				if(options.stream) {
-					if(loopMessages.length) options.stream({text: '\n\n'});
+					if(!isFirstMessage) options.stream({text: '\n\n'});
+					else isFirstMessage = false;
 					resp.content = [];
 					for await (const chunk of resp) {
 						if(controller.signal.aborted) break;
@@ -104,10 +108,11 @@ export class Anthropic extends LLMProvider {
 					}
 				}
 
-				loopMessages.push({role: 'assistant', content: resp.content, timestamp: Date.now()});
+				// Run tools
 				const toolCalls = resp.content.filter((c: any) => c.type === 'tool_use');
 				if(toolCalls.length && !controller.signal.aborted) {
 					history.push({role: 'assistant', content: resp.content});
+					original.push({role: 'assistant', content: resp.content});
 					const results = await Promise.all(toolCalls.map(async (toolCall: any) => {
 						const tool = options.tools?.find(findByProp('name', toolCall.name));
 						if(!tool) return {tool_use_id: toolCall.id, is_error: true, content: 'Tool not found'};
@@ -118,18 +123,13 @@ export class Anthropic extends LLMProvider {
 							return {type: 'tool_result', tool_use_id: toolCall.id, is_error: true, content: err?.message || err?.toString() || 'Unknown'};
 						}
 					}));
-					const userMsg = {role: 'user', content: results};
-					history.push(userMsg);
-					loopMessages.push({...userMsg, timestamp: Date.now()});
+					history.push({role: 'user', content: results});
 					requestParams.messages = history;
 				}
 			} while (!controller.signal.aborted && resp.content.some((c: any) => c.type === 'tool_use'));
 
-			const combinedContent = loopMessages.filter(m => m.role === 'assistant')
-				.map(m => m.content.filter((c: any) => c.type == 'text').map((c: any) => c.text).join('\n\n'))
-				.filter(c => c).join('\n\n');
 			if(options.stream) options.stream({done: true});
-			res(this.toStandard([...history, {role: 'assistant', content: combinedContent, timestamp: Date.now()}]));
+			res(this.toStandard([...history, {role: 'assistant', content: resp.content.filter((c: any) => c.type == 'text').map((c: any) => c.text).join('\n\n')}]));
 		});
 
 		return Object.assign(response, {abort: () => controller.abort()});

src/ollama.ts

@@ -72,12 +72,12 @@ export class Ollama extends LLMProvider {
 				}))
 			}
 
-			let resp: any;
-			const loopMessages: any[] = [];
+			let resp: any, isFirstMessage = true;
 			do {
 				resp = await this.client.chat(requestParams);
 				if(options.stream) {
-					if(loopMessages.length) options.stream({text: '\n\n'});
+					if(!isFirstMessage) options.stream({text: '\n\n'});
+					else isFirstMessage = false;
 					resp.message = {role: 'assistant', content: '', tool_calls: []};
 					for await (const chunk of resp) {
 						if(controller.signal.aborted) break;
@@ -90,7 +90,6 @@ export class Ollama extends LLMProvider {
 					}
 				}
 
-				loopMessages.push({role: 'assistant', content: resp.message?.content, timestamp: Date.now()});
 				if(resp.message?.tool_calls?.length && !controller.signal.aborted) {
 					history.push(resp.message);
 					const results = await Promise.all(resp.message.tool_calls.map(async (toolCall: any) => {
@@ -105,15 +104,12 @@ export class Ollama extends LLMProvider {
 						}
 					}));
 					history.push(...results);
-					loopMessages.push(...results.map(r => ({...r, timestamp: Date.now()})));
 					requestParams.messages = history;
 				}
 			} while (!controller.signal.aborted && resp.message?.tool_calls?.length);
 
-			const combinedContent = loopMessages.filter(m => m.role === 'assistant')
-				.map(m => m.content).filter(c => c).join('\n\n');
 			if(options.stream) options.stream({done: true});
-			res(this.toStandard([...history, {role: 'assistant', content: combinedContent, timestamp: Date.now()}]));
+			res(this.toStandard([...history, {role: 'assistant', content: resp.message?.content}]));
 		});
 
 		return Object.assign(response, {abort: () => controller.abort()});

src/open-ai.ts

@@ -87,12 +87,12 @@ export class OpenAi extends LLMProvider {
 				}))
 			};
 
-			let resp: any;
-			const loopMessages: any[] = [];
+			let resp: any, isFirstMessage = true;
 			do {
 				resp = await this.client.chat.completions.create(requestParams);
 				if(options.stream) {
-					if(loopMessages.length) options.stream({text: '\n\n'});
+					if(!isFirstMessage) options.stream({text: '\n\n'});
+					else isFirstMessage = false;
 					resp.choices = [{message: {content: '', tool_calls: []}}];
 					for await (const chunk of resp) {
 						if(controller.signal.aborted) break;
@@ -106,8 +106,6 @@ export class OpenAi extends LLMProvider {
 					}
 				}
 
-				loopMessages.push({role: 'assistant', content: resp.choices[0].message.content || '', timestamp: Date.now()});
-
 				const toolCalls = resp.choices[0].message.tool_calls || [];
 				if(toolCalls.length && !controller.signal.aborted) {
 					history.push(resp.choices[0].message);
@@ -123,15 +121,12 @@ export class OpenAi extends LLMProvider {
 						}
 					}));
 					history.push(...results);
-					loopMessages.push(...results.map(r => ({...r, timestamp: Date.now()})));
 					requestParams.messages = history;
 				}
 			} while (!controller.signal.aborted && resp.choices?.[0]?.message?.tool_calls?.length);
 
-			const combinedContent = loopMessages.filter(m => m.role === 'assistant')
-				.map(m => m.content).filter(c => c).join('\n\n');
 			if(options.stream) options.stream({done: true});
-			res(this.toStandard([...history, {role: 'assistant', content: combinedContent, timestamp: Date.now()}]));
+			res(this.toStandard([...history, {role: 'assistant', content: resp.choices[0].message.content || ''}]));
 		});
 		return Object.assign(response, {abort: () => controller.abort()});
 	}