feat: refactor AI engine logic, remove cross-engine fallback, add retry with backoff, and bump version to 2.80.18

src/agent.ts

@@ -1942,35 +1942,37 @@ export class AgentExecutor {
         temperature: number;
     }): Promise<{ response: Response; engine: 'lmstudio' | 'ollama'; apiUrl: string }> {
         const { baseUrl, modelName, reqMessages, temperature } = params;
-        const primaryEngine = resolveEngine(baseUrl);
-        const engines = primaryEngine === 'lmstudio' ? ['lmstudio', 'ollama'] as const : ['ollama', 'lmstudio'] as const;
+        const engine = resolveEngine(baseUrl);  // 사용자가 설정한 엔진만 사용
+        const apiUrl = buildApiUrl(baseUrl, engine, 'chat');
+        const messageVariants = this.buildEngineMessageVariants(reqMessages, engine);
+        const modelCandidates = this.buildModelCandidates(modelName, engine);
         let lastError: Error | null = null;
 
-        for (const engine of engines) {
-            const apiUrl = buildApiUrl(baseUrl, engine, 'chat');
-            const messageVariants = this.buildEngineMessageVariants(reqMessages, engine);
-            const modelCandidates = this.buildModelCandidates(modelName, engine);
-
-            for (const candidateModel of modelCandidates) {
-                for (const variant of messageVariants) {
-                    const streamBody = {
-                        model: candidateModel,
-                        messages: variant.messages,
-                        stream: true,
-                        ...(engine === 'lmstudio'
-                            ? { max_tokens: 4096, temperature }
-                            : { options: { num_ctx: 32768, num_predict: 4096, temperature } }),
-                    };
+        // 같은 엔진 내에서만 model candidate / message variant retry
+        for (const candidateModel of modelCandidates) {
+            for (const variant of messageVariants) {
+                const streamBody = {
+                    model: candidateModel,
+                    messages: variant.messages,
+                    stream: true,
+                    ...(engine === 'lmstudio'
+                        ? { max_tokens: 4096, temperature }
+                        : { options: { num_ctx: 32768, num_predict: 4096, temperature } }),
+                };
 
+                // 일시적 네트워크 오류용 retry (최대 2회, 지수 backoff)
+                const MAX_RETRIES = 2;
+                for (let attempt = 0; attempt <= MAX_RETRIES; attempt++) {
                     try {
+                        if (attempt > 0) {
+                            const delay = 500 * Math.pow(2, attempt - 1);  // 500ms, 1000ms
+                            await new Promise(r => setTimeout(r, delay));
+                            logInfo('AI streaming request retry.', { engine, attempt, model: candidateModel });
+                        }
                         logInfo('AI streaming request started.', {
-                            engine,
-                            apiUrl,
-                            model: candidateModel,
-                            variant: variant.name,
-                            messageCount: variant.messages.length,
-                            roles: variant.messages.map(message => message.role),
-                            firstUserPreview: summarizeText(String(variant.messages.find(message => message.role === 'user')?.content || ''), 300)
+                            engine, apiUrl, model: candidateModel,
+                            variant: variant.name, messageCount: variant.messages.length,
+                            attempt
                         });
                         const response = await fetch(apiUrl, {
                             method: 'POST',
@@ -1988,7 +1990,12 @@ export class AgentExecutor {
                         if (!response.ok) {
                             const errText = await response.text();
                             lastError = new Error(`AI Engine error (${engine}/${variant.name}): ${response.status} - ${summarizeText(errText, 300)}`);
-                            logError('AI streaming request returned non-OK status.', { engine, variant: variant.name, apiUrl, status: response.status, body: summarizeText(errText, 500) });
+                            logError('AI streaming request returned non-OK status.', {
+                                engine, variant: variant.name, apiUrl,
+                                status: response.status, body: summarizeText(errText, 500)
+                            });
+                            // 4xx는 재시도해도 의미없음. 5xx만 재시도.
+                            if (response.status >= 400 && response.status < 500) break;
                             continue;
                         }
 
@@ -1996,13 +2003,26 @@ export class AgentExecutor {
                         return { response, engine, apiUrl };
                     } catch (error: any) {
                         lastError = error instanceof Error ? error : new Error(String(error));
-                        logError('AI streaming request failed.', { engine, variant: variant.name, apiUrl, model: candidateModel, error: lastError.message });
+                        // AbortError는 사용자가 취소한 것이므로 retry 금지
+                        if (lastError.name === 'AbortError') {
+                            throw lastError;
+                        }
+                        logError('AI streaming request failed.', {
+                            engine, variant: variant.name, apiUrl, model: candidateModel,
+                            attempt, error: lastError.message
+                        });
                     }
                 }
             }
         }
 
-        throw lastError || new Error('Unable to connect to AI engine.');
+        // 명확한 에러 메시지: 어느 엔진이 실패했는지 사용자에게 알림
+        const engineLabel = engine === 'lmstudio' ? 'LM Studio' : 'Ollama';
+        throw new Error(
+            `${engineLabel} 엔진에 연결할 수 없습니다. ` +
+            `${engineLabel}가 실행 중이고 모델 '${modelName}'이 로드되어 있는지 확인하세요. ` +
+            `(원인: ${lastError?.message || 'unknown'})`
+        );
     }
 
     private normalizeMessages(messages: ChatMessage[]) {