Improve Claude stream completion handling

src/_locales/de/main.json

@@ -87,6 +87,10 @@
   "Clear Conversation": "Konversation löschen",
   "Retry": "Erneut versuchen",
   "Exceeded maximum context length": "Maximale Kontextlänge überschritten, bitte Konversation löschen und erneut versuchen",
+  "Claude reached the response token limit. Increase Max Response Token Length and try again.": "Claude hat das Antwort-Token-Limit erreicht. Erhöhen Sie die maximale Antwort-Token-Länge und versuchen Sie es erneut.",
+  "Claude reached the model context window limit. Clear the conversation and try again.": "Claude hat das Kontextfenster des Modells ausgeschöpft. Löschen Sie die Konversation und versuchen Sie es erneut.",
+  "Claude declined to respond to this request.": "Claude hat die Antwort auf diese Anfrage abgelehnt.",
+  "Claude response stream ended before completion.": "Der Claude-Antwortstream wurde vorzeitig beendet.",
   "Regenerate the answer after switching model": "Antwort nach dem Wechseln des Modells neu generieren",
   "Pin": "Anheften",
   "Unpin": "Loslösen",

src/_locales/en/main.json

@@ -94,6 +94,10 @@
   "Clear Conversation": "Clear Conversation",
   "Retry": "Retry",
   "Exceeded maximum context length": "Exceeded maximum context length, please clear the conversation and try again",
+  "Claude reached the response token limit. Increase Max Response Token Length and try again.": "Claude reached the response token limit. Increase Max Response Token Length and try again.",
+  "Claude reached the model context window limit. Clear the conversation and try again.": "Claude reached the model context window limit. Clear the conversation and try again.",
+  "Claude declined to respond to this request.": "Claude declined to respond to this request.",
+  "Claude response stream ended before completion.": "Claude response stream ended before completion.",
   "Regenerate the answer after switching model": "Regenerate the answer after switching model",
   "Pin": "Pin",
   "Unpin": "Unpin",

src/_locales/es/main.json

@@ -85,6 +85,10 @@
   "Clear Conversation": "Borrar conversación",
   "Retry": "Reintentar",
   "Exceeded maximum context length": "Se superó la longitud máxima del contexto, borre la conversación y vuelva a intentarlo",
+  "Claude reached the response token limit. Increase Max Response Token Length and try again.": "Claude alcanzó el límite de tokens de respuesta. Aumente la longitud máxima de tokens de respuesta e inténtelo de nuevo.",
+  "Claude reached the model context window limit. Clear the conversation and try again.": "Claude alcanzó el límite de la ventana de contexto del modelo. Borre la conversación e inténtelo de nuevo.",
+  "Claude declined to respond to this request.": "Claude se negó a responder a esta solicitud.",
+  "Claude response stream ended before completion.": "El flujo de respuesta de Claude terminó antes de completarse.",
   "Regenerate the answer after switching model": "Regenerar la respuesta después de cambiar el modelo",
   "Pin": "Fijar",
   "Unpin": "Desfijar",

src/_locales/fr/main.json

@@ -85,6 +85,10 @@
   "Clear Conversation": "Effacer la conversation",
   "Retry": "Réessayer",
   "Exceeded maximum context length": "Dépassement de la longueur maximale de contexte, veuillez effacer la conversation et réessayer",
+  "Claude reached the response token limit. Increase Max Response Token Length and try again.": "Claude a atteint la limite de jetons de réponse. Augmentez la longueur maximale des jetons de réponse, puis réessayez.",
+  "Claude reached the model context window limit. Clear the conversation and try again.": "Claude a atteint la limite de la fenêtre de contexte du modèle. Effacez la conversation, puis réessayez.",
+  "Claude declined to respond to this request.": "Claude a refusé de répondre à cette requête.",
+  "Claude response stream ended before completion.": "Le flux de réponse de Claude s’est terminé avant la fin.",
   "Regenerate the answer after switching model": "Régénérer la réponse après avoir changé de modèle",
   "Pin": "Épingler",
   "Unpin": "Détacher",

src/_locales/in/main.json

@@ -85,6 +85,10 @@
   "Clear Conversation": "Bersihkan Percakapan",
   "Retry": "Coba Lagi",
   "Exceeded maximum context length": "Melampaui batas maksimum panjang konteks, harap bersihkan percakapan dan coba lagi",
+  "Claude reached the response token limit. Increase Max Response Token Length and try again.": "Claude mencapai batas token respons. Tingkatkan panjang token respons maksimum lalu coba lagi.",
+  "Claude reached the model context window limit. Clear the conversation and try again.": "Claude mencapai batas jendela konteks model. Hapus percakapan lalu coba lagi.",
+  "Claude declined to respond to this request.": "Claude menolak menanggapi permintaan ini.",
+  "Claude response stream ended before completion.": "Aliran respons Claude berakhir sebelum selesai.",
   "Regenerate the answer after switching model": "Hasilkan kembali jawaban setelah beralih ke model lain",
   "Pin": "Sematkan",
   "Unpin": "Lepas Sematan",

src/_locales/it/main.json

@@ -85,6 +85,10 @@
   "Clear Conversation": "Pulisci la conversazione",
   "Retry": "Riprova",
   "Exceeded maximum context length": "Lunghezza massima del contesto superata, si prega di pulire la conversazione e riprovare",
+  "Claude reached the response token limit. Increase Max Response Token Length and try again.": "Claude ha raggiunto il limite di token della risposta. Aumenta la lunghezza massima dei token della risposta e riprova.",
+  "Claude reached the model context window limit. Clear the conversation and try again.": "Claude ha raggiunto il limite della finestra di contesto del modello. Cancella la conversazione e riprova.",
+  "Claude declined to respond to this request.": "Claude ha rifiutato di rispondere a questa richiesta.",
+  "Claude response stream ended before completion.": "Il flusso di risposta di Claude è terminato prima del completamento.",
   "Regenerate the answer after switching model": "Rigenerare la risposta dopo aver cambiato il modello",
   "Pin": "Fissa",
   "Unpin": "Sblocca",

src/_locales/ja/main.json

@@ -85,6 +85,10 @@
   "Clear Conversation": "会話をクリア",
   "Retry": "再試行",
   "Exceeded maximum context length": "最大コンテキスト長を超えました。会話をクリアして再試行してください",
+  "Claude reached the response token limit. Increase Max Response Token Length and try again.": "Claude が応答トークンの上限に達しました。最大応答トークン長を増やして、もう一度お試しください。",
+  "Claude reached the model context window limit. Clear the conversation and try again.": "Claude がモデルのコンテキストウィンドウ上限に達しました。会話をクリアして、もう一度お試しください。",
+  "Claude declined to respond to this request.": "Claude はこのリクエストへの応答を拒否しました。",
+  "Claude response stream ended before completion.": "Claude の応答ストリームが完了前に終了しました。",
   "Regenerate the answer after switching model": "モデルを切り替えた後に回答を再生成",
   "Pin": "ピン留め",
   "Unpin": "ピン留め解除",

src/_locales/ko/main.json

@@ -85,6 +85,10 @@
   "Clear Conversation": "대화 내용 지우기",
   "Retry": "재시도",
   "Exceeded maximum context length": "최대 컨텍스트 길이를 초과하였습니다. 대화 내용을 지우고 다시 시도해주세요.",
+  "Claude reached the response token limit. Increase Max Response Token Length and try again.": "Claude가 응답 토큰 한도에 도달했습니다. 최대 응답 토큰 길이를 늘린 후 다시 시도하세요.",
+  "Claude reached the model context window limit. Clear the conversation and try again.": "Claude가 모델의 컨텍스트 창 한도에 도달했습니다. 대화 내용을 지운 후 다시 시도하세요.",
+  "Claude declined to respond to this request.": "Claude가 이 요청에 대한 응답을 거부했습니다.",
+  "Claude response stream ended before completion.": "Claude 응답 스트림이 완료되기 전에 종료되었습니다.",
   "Regenerate the answer after switching model": "모델 전환 후 대답 다시 생성",
   "Pin": "고정",
   "Unpin": "고정 해제",

src/_locales/pt/main.json

@@ -85,6 +85,10 @@
   "Clear Conversation": "Limpar Conversa",
   "Retry": "Tentar novamente",
   "Exceeded maximum context length": "Ultrapassou o comprimento máximo do contexto. Limpe a conversa e tente novamente",
+  "Claude reached the response token limit. Increase Max Response Token Length and try again.": "O Claude atingiu o limite de tokens da resposta. Aumente o comprimento máximo dos tokens da resposta e tente novamente.",
+  "Claude reached the model context window limit. Clear the conversation and try again.": "O Claude atingiu o limite da janela de contexto do modelo. Limpe a conversa e tente novamente.",
+  "Claude declined to respond to this request.": "O Claude recusou responder a este pedido.",
+  "Claude response stream ended before completion.": "O fluxo de resposta do Claude terminou antes de ser concluído.",
   "Regenerate the answer after switching model": "Regenerar a resposta após trocar o modelo",
   "Pin": "Fixar",
   "Unpin": "Desafixar",

src/_locales/ru/main.json

@@ -93,6 +93,10 @@
   "Clear Conversation": "Очистить беседу",
   "Retry": "Повторить",
   "Exceeded maximum context length": "Превышена максимальная длина контекста, очистите беседу и повторите попытку",
+  "Claude reached the response token limit. Increase Max Response Token Length and try again.": "Claude достиг лимита токенов ответа. Увеличьте максимальную длину ответа в токенах и повторите попытку.",
+  "Claude reached the model context window limit. Clear the conversation and try again.": "Claude достиг лимита контекстного окна модели. Очистите беседу и повторите попытку.",
+  "Claude declined to respond to this request.": "Claude отказался отвечать на этот запрос.",
+  "Claude response stream ended before completion.": "Поток ответа Claude завершился до окончания ответа.",
   "Regenerate the answer after switching model": "Генерировать ответ после смены модели",
   "Pin": "Закрепить",
   "Unpin": "Открепить",

src/_locales/tr/main.json

@@ -85,6 +85,10 @@
   "Clear Conversation": "Konuşmayı Temizle",
   "Retry": "Tekrar Dene",
   "Exceeded maximum context length": "Maksimum bağlam uzunluğu aşıldı",
+  "Claude reached the response token limit. Increase Max Response Token Length and try again.": "Claude yanıt belirteci sınırına ulaştı. Maksimum yanıt belirteci uzunluğunu artırıp yeniden deneyin.",
+  "Claude reached the model context window limit. Clear the conversation and try again.": "Claude modelin bağlam penceresi sınırına ulaştı. Konuşmayı temizleyip yeniden deneyin.",
+  "Claude declined to respond to this request.": "Claude bu isteğe yanıt vermeyi reddetti.",
+  "Claude response stream ended before completion.": "Claude yanıt akışı tamamlanmadan sona erdi.",
   "Regenerate the answer after switching model": "Modeli değiştirdikten sonra cevabı yeniden oluştur",
   "Pin": "Sabitle",
   "Unpin": "Sabitlemeyi Kaldır",

src/_locales/zh-hans/main.json

@@ -88,6 +88,10 @@
   "Clear Conversation": "清理对话",
   "Retry": "重试",
   "Exceeded maximum context length": "超出最大上下文长度, 请清理对话并重试",
+  "Claude reached the response token limit. Increase Max Response Token Length and try again.": "Claude 已达到响应 token 上限。请增大最大响应 token 长度，然后重试。",
+  "Claude reached the model context window limit. Clear the conversation and try again.": "Claude 已达到模型上下文窗口上限。请清理对话，然后重试。",
+  "Claude declined to respond to this request.": "Claude 拒绝响应此请求。",
+  "Claude response stream ended before completion.": "Claude 响应流在完成前已结束。",
   "Regenerate the answer after switching model": "快捷切换模型时自动重新生成回答",
   "Pin": "固定侧边",
   "Unpin": "收缩侧边",

src/_locales/zh-hant/main.json

@@ -88,6 +88,10 @@
   "Clear Conversation": "清除對話",
   "Retry": "重試",
   "Exceeded maximum context length": "超出最大上下文長度，請清除對話並重試",
+  "Claude reached the response token limit. Increase Max Response Token Length and try again.": "Claude 已達回應 token 上限。請提高 [回應的最大 token 長度]，然後再試一次。",
+  "Claude reached the model context window limit. Clear the conversation and try again.": "Claude 已達模型的上下文視窗上限。請清除對話後再試一次。",
+  "Claude declined to respond to this request.": "Claude 拒絕回應這項要求。",
+  "Claude response stream ended before completion.": "Claude 回應串流尚未完成便已結束。",
   "Regenerate the answer after switching model": "切換模型後自動重新產生回答",
   "Pin": "固定側邊",
   "Unpin": "取消固定側邊",

src/services/apis/claude-api.mjs

@@ -1,6 +1,6 @@
 import { getUserConfig } from '../../config/index.mjs'
 import { pushRecord, setAbortController } from './shared.mjs'
-import { fetchSSE } from '../../utils/fetch-sse.mjs'
+import { FETCH_RESPONSE_STREAM_FAILED, fetchSSE } from '../../utils/fetch-sse.mjs'
 import { isEmpty } from 'lodash-es'
 import { getConversationPairs } from '../../utils/get-conversation-pairs.mjs'
 import { getModelValue } from '../../utils/model-name-convert.mjs'
@@ -49,7 +49,11 @@ export async function generateAnswersWithClaudeApi(port, question, session) {
   }
 
   let answer = ''
-  await fetchSSE(`${apiUrl}/v1/messages`, {
+  let stopReason = ''
+  let completionError
+  let wasAborted = false
+  let completedSuccessfully = false
+  const streamError = await fetchSSE(`${apiUrl}/v1/messages`, {
     method: 'POST',
     signal: controller.signal,
     headers: {
@@ -69,10 +73,42 @@ export async function generateAnswersWithClaudeApi(port, question, session) {
         console.debug('json error', error)
         return
       }
+      if (data?.type === 'error') {
+        controller.abort()
+        const error = new Error(JSON.stringify(data))
+        if (completedSuccessfully) error.code = FETCH_RESPONSE_STREAM_FAILED
+        throw error
+      }
+      if (completedSuccessfully) return
+      if (data?.type === 'message_delta') {
+        stopReason = data?.delta?.stop_reason || stopReason
+        return
+      }
       if (data?.type === 'message_stop') {
+        if (stopReason === 'max_tokens') {
+          completionError = new Error(
+            'Claude reached the response token limit. Increase Max Response Token Length and try again.',
+          )
+        }
+        if (stopReason === 'model_context_window_exceeded') {
+          completionError = new Error(
+            'Claude reached the model context window limit. Clear the conversation and try again.',
+          )
+        }
+        if (stopReason === 'refusal') {
+          completionError = new Error('Claude declined to respond to this request.')
+        }
+        if (!completionError && !['end_turn', 'stop_sequence'].includes(stopReason)) {
+          completionError = new Error('Claude response stream ended before completion.')
+        }
+        if (completionError) {
+          controller.abort()
+          throw completionError
+        }
         pushRecord(session, question, answer)
         console.debug('conversation history', { content: session.conversationRecords })
         port.postMessage({ answer: null, done: true, session: session })
+        completedSuccessfully = true
         return
       }
 
@@ -84,12 +120,10 @@ export async function generateAnswersWithClaudeApi(port, question, session) {
     },
     async onStart() {},
     async onEnd(aborted) {
+      wasAborted = aborted
       try {
-        if (!aborted) {
-          port.postMessage({ done: true })
-        }
-      } finally {
         port.onMessage.removeListener(messageListener)
+      } finally {
         port.onDisconnect.removeListener(disconnectListener)
       }
     },
@@ -100,5 +134,16 @@ export async function generateAnswersWithClaudeApi(port, question, session) {
       const error = await resp.json().catch(() => ({}))
       throw new Error(!isEmpty(error) ? JSON.stringify(error) : `${resp.status} ${resp.statusText}`)
     },
-  })
+  }).catch((error) => error)
+  if (wasAborted) return
+  if (completionError) throw completionError
+  if (
+    streamError &&
+    (!completedSuccessfully || streamError.code !== FETCH_RESPONSE_STREAM_FAILED)
+  ) {
+    throw streamError
+  }
+  if (!completedSuccessfully) {
+    throw new Error('Claude response stream ended before completion.')
+  }
 }