From 2b84d9489727a25f53037a33dfbff35843041fef Mon Sep 17 00:00:00 2001 From: matevip Date: Sun, 10 May 2026 19:15:15 +0800 Subject: [PATCH] fix(channel): adaptive debounce window for paste-split long messages --- .../channel/wecom/WeComChannelAdapter.java | 605 ++---------------- .../controller/SystemSettingController.java | 31 - .../system/service/SystemSettingService.java | 42 +- mateclaw-ui/src/api/index.ts | 7 - .../Models/MultimodalSidecarSection.vue | 7 +- 5 files changed, 45 insertions(+), 647 deletions(-) diff --git a/mateclaw-server/src/main/java/vip/mate/channel/wecom/WeComChannelAdapter.java b/mateclaw-server/src/main/java/vip/mate/channel/wecom/WeComChannelAdapter.java index bec82890..a92af82e 100644 --- a/mateclaw-server/src/main/java/vip/mate/channel/wecom/WeComChannelAdapter.java +++ b/mateclaw-server/src/main/java/vip/mate/channel/wecom/WeComChannelAdapter.java @@ -201,33 +201,6 @@ public class WeComChannelAdapter extends AbstractChannelAdapter { /** 回复上下文：replyToken -> (frameReqId, processingStreamId)，用于 sendContentParts 回写 */ private final ConcurrentHashMap replyContexts = new ConcurrentHashMap<>(); - /** - * Group-chat reply-slot fallback cache: {@code chatId → most recent - * inbound frameReqId from that group}. - *

- * The WeCom AI Bot platform blocks {@code aibot_send_msg} in - * group chats — proactive pushes (cron summaries, - * image-generation completions, TTS audio, async-task forwards) must - * ride {@code aibot_respond_msg} bound to some prior frame - * id. Without this cache every group push silently failed: - * {@code sendMessageToChat} fell through to {@code aibot_send_msg}, - * the platform rejected it, the user saw nothing. - *

- * Populated for group inbound frames only — single-chat - * {@code aibot_send_msg} still works, so we don't need a cached - * reqId there. {@link #pickGroupReplyReqId(String)} returns the - * cached id (or null when there's never been a group inbound), and - * the proactive send paths fall through to {@code aibot_send_msg} - * when null. - *

- * Bounded LRU at {@link #LAST_CHAT_REQ_IDS_MAX_SIZE} via insertion- - * order eviction — long-lived bots in many groups don't unbounded-grow. - */ - private final ConcurrentHashMap lastChatReqIds = new ConcurrentHashMap<>(); - - /** Max chat-id entries to keep in {@link #lastChatReqIds} before evicting. */ - private static final int LAST_CHAT_REQ_IDS_MAX_SIZE = 1000; - private record WeComReplyContext(String frameReqId, String processingStreamId) {} /** @@ -1027,66 +1000,12 @@ public class WeComChannelAdapter extends AbstractChannelAdapter { contentParts.add(0, MessageContentPart.text(textContent)); } } - case "appmsg" -> { - // Forwarded WeCom complex messages: PDF / Word / Excel - // transfers, public-account article links, miniprogram - // cards. Without this branch every forwarded PDF / - // article / miniprogram fell into default and got - // silently dropped — bot looked dumb to users. - AppmsgContent appmsg = extractAppmsgContent(body, msgId, senderId, chatId, chatType); - if (appmsg.text() != null && !appmsg.text().isBlank()) { - textContent = appmsg.text(); - contentParts.add(0, MessageContentPart.text(textContent)); - } - contentParts.addAll(appmsg.attachedParts()); - } default -> { log.debug("[wecom] Ignoring unsupported message type: {}", msgType); return; } } - // Apply any quoted-message context. The "quote" field appears at - // body level alongside the new message regardless of outer - // msgtype — without parsing it, the agent only sees the user's - // current text and silently loses the conversational reference - // ("user quoted the bot's previous image and asked '什么意思'" - // arrives as bare "什么意思", agent goes off-topic). - QuoteContext quote = extractQuoteContext(body, msgId, senderId, chatId, chatType); - if (quote != null && !quote.isEmpty()) { - String prefixedText = quote.prefix() - + (textContent != null && !textContent.isBlank() ? textContent : ""); - textContent = prefixedText; - - // Find an existing text part (text/voice cases produce one) - // and overwrite it with the prefixed text. Also pull it to - // the front so agent reading order starts with quote prefix. - boolean updated = false; - for (int i = 0; i < contentParts.size(); i++) { - if ("text".equals(contentParts.get(i).getType())) { - contentParts.set(i, MessageContentPart.text(prefixedText)); - if (i != 0) { - MessageContentPart promoted = contentParts.remove(i); - contentParts.add(0, promoted); - } - updated = true; - break; - } - } - if (!updated) { - // image/file/mixed cases without a text part — insert one. - contentParts.add(0, MessageContentPart.text(prefixedText)); - } - // Quoted media (image/file the user referenced) goes right - // after the text prefix so the agent reads: - // prefix → quoted media → user's own media (if any) - if (!quote.attachedParts().isEmpty()) { - contentParts.addAll(1, quote.attachedParts()); - } - log.debug("[wecom] Applied quote context: prefixLen={}, attachedParts={}", - quote.prefix().length(), quote.attachedParts().size()); - } - if (contentParts.isEmpty()) { if (textContent != null && !textContent.isBlank()) { contentParts.add(MessageContentPart.text(textContent)); @@ -1109,16 +1028,6 @@ public class WeComChannelAdapter extends AbstractChannelAdapter { boolean isGroup = "group".equals(chatType); String effectiveChatId = isGroup ? chatId : null; - // Cache the group's most recent inbound frameReqId so future - // proactive pushes into this group can ride aibot_respond_msg - // (the AI bot platform blocks aibot_send_msg in groups — - // without this cache async-task completions and cron summaries - // silently fail to deliver). - if (isGroup && chatId != null && !chatId.isBlank() - && frameReqId != null && !frameReqId.isBlank()) { - rememberGroupReplyReqId(chatId, frameReqId); - } - // conversationId 格式：wecom:{userid} 或 wecom:group:{chatid} // 由 ChannelMessageRouter.buildConversationId() 根据 channelType + chatId/senderId 构建 @@ -1327,96 +1236,27 @@ public class WeComChannelAdapter extends AbstractChannelAdapter { } /** - * 通过 WebSocket send_message 命令主动推送消息。 - *

- * Group fallback: WeCom AI Bot platform rejects {@code aibot_send_msg} - * in group chats. When {@code chatId} matches a known group (via - * {@link #pickGroupReplyReqId}), ride a cached inbound reqId via - * {@code aibot_respond_msg} instead. Single chats still use - * {@code aibot_send_msg} (which the platform allows). + * 通过 WebSocket send_message 命令主动推送消息 */ private void sendMessageToChat(String chatId, String content) { if (webSocket == null || content == null || content.isBlank()) return; - Map textBody = Map.of( - "msgtype", "markdown", - "markdown", Map.of("content", content) - ); - sendOutboundFrame(chatId, textBody); - } - - /** - * Send a body via either {@code aibot_respond_msg} (groups, using a - * cached inbound reqId) or {@code aibot_send_msg} (single chats). - *

- * Centralised so {@link #sendMessageToChat} (text) and - * {@link #sendMediaMessage} (image/file/voice/video) share the same - * group-vs-single dispatch — without this, every new outbound path - * had to remember the group rule, and several didn't. - */ - private void sendOutboundFrame(String chatId, Map bodyWithMsgtype) { - if (webSocket == null || chatId == null || chatId.isBlank()) return; try { - String groupReplyReqId = pickGroupReplyReqId(chatId); - if (groupReplyReqId != null) { - // Group chat — ride aibot_respond_msg with the cached reqId. - // The body for respond_msg does NOT include "chatid" — the - // server infers the target from the original frame's reqId. - Map frame = Map.of( - "cmd", CMD_RESPONSE, - "headers", Map.of("req_id", groupReplyReqId), - "body", bodyWithMsgtype - ); - sendFrameWithAck(groupReplyReqId, frame); - log.debug("[wecom] Group send via aibot_respond_msg: chatId={}, reqId={}", - chatId, groupReplyReqId); - } else { - // Single chat — aibot_send_msg accepts a chatid field. - Map withChatId = new LinkedHashMap<>(bodyWithMsgtype); - withChatId.put("chatid", chatId); - String reqId = generateReqId(CMD_SEND_MSG); - Map frame = Map.of( - "cmd", CMD_SEND_MSG, - "headers", Map.of("req_id", reqId), - "body", withChatId - ); - sendFrameWithAck(reqId, frame); - } + String reqId = generateReqId(CMD_SEND_MSG); + Map frame = Map.of( + "cmd", CMD_SEND_MSG, + "headers", Map.of("req_id", reqId), + "body", Map.of( + "chatid", chatId, + "msgtype", "markdown", + "markdown", Map.of("content", content) + ) + ); + sendFrameWithAck(reqId, frame); } catch (Exception e) { - log.error("[wecom] Failed to send outbound frame to {}: {}", chatId, e.getMessage(), e); + log.error("[wecom] Failed to send message to {}: {}", chatId, e.getMessage(), e); } } - /** - * Record the most recent inbound frameReqId for a group chat. Bounded - * LRU: when over {@link #LAST_CHAT_REQ_IDS_MAX_SIZE} we evict the - * oldest insertion-order key (ConcurrentHashMap iteration is - * insertion-order-ish for small maps and good enough — the cache - * exists to bound memory, not to be perfectly LRU). - */ - private void rememberGroupReplyReqId(String chatId, String frameReqId) { - lastChatReqIds.put(chatId, frameReqId); - while (lastChatReqIds.size() > LAST_CHAT_REQ_IDS_MAX_SIZE) { - // Pop one entry — any will do for memory bounding. - var it = lastChatReqIds.keySet().iterator(); - if (it.hasNext()) { - String victim = it.next(); - lastChatReqIds.remove(victim); - } else break; - } - } - - /** - * Look up the cached reply reqId for a group chat. Returns null when - * the chatId isn't a known group (single chats, or no inbound seen - * yet) — callers fall back to {@code aibot_send_msg}. - *

- * Package-private for test access. - */ - String pickGroupReplyReqId(String chatId) { - if (chatId == null || chatId.isBlank()) return null; - return lastChatReqIds.get(chatId); - } - /** * 覆写 renderAndSend：如果有 processing_stream_id 则用 reply_stream 覆盖"思考中..." */ @@ -1638,7 +1478,7 @@ public class WeComChannelAdapter extends AbstractChannelAdapter { } /** - * 发送图片部分：压缩 → 大小预校验 → 上传 → 发送 media_id + * 发送图片部分：压缩 → 上传 → 发送 media_id */ private void sendImagePart(String targetId, MessageContentPart part, WeComReplyContext ctx) { byte[] imageBytes = resolveFileBytes(part); @@ -1650,29 +1490,17 @@ public class WeComChannelAdapter extends AbstractChannelAdapter { String fileName = part.getFileName() != null ? part.getFileName() : "image.jpg"; imageBytes = WeComImageCompressor.compressIfNeeded(imageBytes, fileName); - WeComUploadLimitDecision decision = applyWeComUploadLimits( - imageBytes.length, "image", part.getContentType()); - if (decision.rejected()) { - log.warn("[wecom] Image upload rejected: {} ({} bytes) — {}", - fileName, imageBytes.length, decision.rejectReason()); - sendMessageToChat(targetId, "⚠️ " + decision.rejectReason()); - return; - } - - String mediaId = uploadMedia(imageBytes, fileName, decision.finalMediaType()); + String mediaId = uploadMedia(imageBytes, fileName, "image"); if (mediaId != null) { String frameReqId = ctx != null ? ctx.frameReqId() : null; - sendMediaMessage(targetId, mediaId, decision.finalMediaType(), frameReqId); - if (decision.downgraded()) { - sendMessageToChat(targetId, "ℹ️ " + decision.downgradeNote()); - } + sendMediaMessage(targetId, mediaId, "image", frameReqId); } else { sendFallbackText(targetId, part); } } /** - * 发送文件部分：大小预校验 → 上传 → 发送 media_id + * 发送文件部分：上传 → 发送 media_id */ private void sendFilePart(String targetId, MessageContentPart part, WeComReplyContext ctx) { byte[] fileBytes = resolveFileBytes(part); @@ -1682,30 +1510,21 @@ public class WeComChannelAdapter extends AbstractChannelAdapter { } String fileName = part.getFileName() != null ? part.getFileName() : "file.bin"; - WeComUploadLimitDecision decision = applyWeComUploadLimits( - fileBytes.length, "file", part.getContentType()); - if (decision.rejected()) { - log.warn("[wecom] File upload rejected: {} ({} bytes) — {}", - fileName, fileBytes.length, decision.rejectReason()); - sendMessageToChat(targetId, "⚠️ " + decision.rejectReason()); - return; - } - - String mediaId = uploadMedia(fileBytes, fileName, decision.finalMediaType()); + String mediaId = uploadMedia(fileBytes, fileName, "file"); if (mediaId != null) { String frameReqId = ctx != null ? ctx.frameReqId() : null; - sendMediaMessage(targetId, mediaId, decision.finalMediaType(), frameReqId); + sendMediaMessage(targetId, mediaId, "file", frameReqId); } else { sendFallbackText(targetId, part); } } /** - * 发送音频部分：读取字节 → 大小+格式预校验 → 上传 → 发送。 + * 发送音频部分：读取字节 → 上传 → 发送 *

- * WeCom 原生语音消息要求 AMR 格式 + ≤ 2 MB。预校验里非 AMR 或超 2 MB - * 自动降级为 file（文件卡片，可点击下载播放）+ 给用户一行说明 - * 提示——避免用户期待"语音气泡"但收到一个 .mp3 文件却不知道为啥。 + * WeCom 原生语音消息要求 AMR 格式。TTS 输出为 MP3， + * Phase 1 以 file 类型发送（用户可点击播放），避免引入 AMR 转码依赖。 + * 非 AMR 格式走 file 类型而非 voice 类型，避免企微语音播放兼容问题。 */ private void sendAudioPart(String targetId, MessageContentPart part, WeComReplyContext ctx) { byte[] audioBytes = resolveFileBytes(part); @@ -1716,30 +1535,15 @@ public class WeComChannelAdapter extends AbstractChannelAdapter { String fileName = part.getFileName() != null ? part.getFileName() : "voice_reply.mp3"; boolean isAmr = fileName.toLowerCase().endsWith(".amr"); - // Pre-decide native voice vs file based on extension; the limits - // checker can still downgrade voice→file if size exceeds 2MB. - String requestedType = isAmr ? "voice" : "file"; - String contentTypeHint = part.getContentType(); - if (contentTypeHint == null && isAmr) contentTypeHint = "audio/amr"; - WeComUploadLimitDecision decision = applyWeComUploadLimits( - audioBytes.length, requestedType, contentTypeHint); - if (decision.rejected()) { - log.warn("[wecom] Audio upload rejected: {} ({} bytes) — {}", - fileName, audioBytes.length, decision.rejectReason()); - sendMessageToChat(targetId, "⚠️ " + decision.rejectReason()); - return; - } - - String mediaId = uploadMedia(audioBytes, fileName, decision.finalMediaType()); + // AMR 格式：以原生 voice 类型发送（语音气泡） + // 其他格式（MP3 等）：以 file 类型发送（文件卡片，可点击播放） + String uploadType = isAmr ? "voice" : "file"; + String mediaId = uploadMedia(audioBytes, fileName, uploadType); if (mediaId != null) { String frameReqId = ctx != null ? ctx.frameReqId() : null; - sendMediaMessage(targetId, mediaId, decision.finalMediaType(), frameReqId); - log.info("[wecom] Audio sent as {}: {} ({}KB)", - decision.finalMediaType(), fileName, audioBytes.length / 1024); - if (decision.downgraded()) { - sendMessageToChat(targetId, "ℹ️ " + decision.downgradeNote()); - } + sendMediaMessage(targetId, mediaId, uploadType, frameReqId); + log.info("[wecom] Audio sent as {}: {} ({}KB)", uploadType, fileName, audioBytes.length / 1024); } else { sendFallbackText(targetId, part); } @@ -1878,105 +1682,6 @@ public class WeComChannelAdapter extends AbstractChannelAdapter { */ private final ConcurrentHashMap streamLastContent = new ConcurrentHashMap<>(); - // ==================== 上传大小预校验（WeCom 平台限制） ==================== - - /** WeCom hard limits — verified empirically; sources differ slightly. */ - static final long IMAGE_MAX_BYTES = 10L * 1024 * 1024; // 10 MB - static final long VIDEO_MAX_BYTES = 10L * 1024 * 1024; // 10 MB - static final long VOICE_MAX_BYTES = 2L * 1024 * 1024; // 2 MB - static final long FILE_MAX_BYTES = 20L * 1024 * 1024; // 20 MB (absolute cap) - static final Set VOICE_SUPPORTED_MIMES = Set.of("audio/amr"); - - /** - * Decision result for {@link #applyWeComUploadLimits}. - * - * @param finalMediaType effective media type after auto-downgrade - * (image / video / voice could become "file" - * when oversized or unsupported) - * @param rejected true → don't upload at all; surface - * {@code rejectReason} to the user instead - * @param rejectReason human-readable reason when {@code rejected} - * @param downgraded true → uploaded as {@code finalMediaType} - * but caller should append a notice to the - * bubble explaining why it's not native - * @param downgradeNote user-friendly note when {@code downgraded} - */ - record WeComUploadLimitDecision(String finalMediaType, boolean rejected, - String rejectReason, boolean downgraded, - String downgradeNote) { - static WeComUploadLimitDecision pass(String mediaType) { - return new WeComUploadLimitDecision(mediaType, false, null, false, null); - } - } - - /** - * Pre-validate an outbound upload against WeCom's per-media-type limits - * and decide whether to upload as-is, downgrade to {@code file}, or - * reject outright. Mirrors the strategy proven in production by - * comparable Java/Python WeCom integrations: - *

- * Without this layer, oversized uploads would chunk-upload for up to - * a minute before the platform server rejected them at the finish - * step, with the user seeing nothing arrive in their chat. - *

- * Package-private for unit-test access. - */ - static WeComUploadLimitDecision applyWeComUploadLimits(long fileSize, String mediaType, - String contentType) { - String type = mediaType == null ? "file" : mediaType.toLowerCase(); - String mime = contentType == null ? "" : contentType.toLowerCase().trim(); - - if (fileSize > FILE_MAX_BYTES) { - double mb = fileSize / 1024.0 / 1024.0; - return new WeComUploadLimitDecision( - type, true, - String.format(java.util.Locale.ROOT, - "文件大小 %.2fMB 超过企业微信 20MB 上限，无法发送。请压缩或拆分后再发。", mb), - false, null); - } - if ("image".equals(type) && fileSize > IMAGE_MAX_BYTES) { - double mb = fileSize / 1024.0 / 1024.0; - return new WeComUploadLimitDecision( - "file", false, null, - true, - String.format(java.util.Locale.ROOT, - "图片 %.2fMB 超过 10MB 限制，已转为文件形式发送", mb)); - } - if ("video".equals(type) && fileSize > VIDEO_MAX_BYTES) { - double mb = fileSize / 1024.0 / 1024.0; - return new WeComUploadLimitDecision( - "file", false, null, - true, - String.format(java.util.Locale.ROOT, - "视频 %.2fMB 超过 10MB 限制，已转为文件形式发送", mb)); - } - if ("voice".equals(type)) { - if (!mime.isEmpty() && !VOICE_SUPPORTED_MIMES.contains(mime)) { - return new WeComUploadLimitDecision( - "file", false, null, true, - "语音格式 " + mime + " 不支持（企微仅支持 AMR），已转为文件形式发送"); - } - if (fileSize > VOICE_MAX_BYTES) { - double mb = fileSize / 1024.0 / 1024.0; - return new WeComUploadLimitDecision( - "file", false, null, - true, - String.format(java.util.Locale.ROOT, - "语音 %.2fMB 超过 2MB 限制，已转为文件形式发送", mb)); - } - } - return WeComUploadLimitDecision.pass(type); - } - /** * 发送欢迎消息 */ @@ -2146,18 +1851,6 @@ public class WeComChannelAdapter extends AbstractChannelAdapter { if (webSocket == null || fileBytes == null || fileBytes.length == 0) { return null; } - // Pre-flight chunk count guard. WeCom's chunked upload protocol caps - // out near 100 chunks (~50 MB at 512 KB / chunk) but we already - // reject anything over FILE_MAX_BYTES (20 MB ≈ 40 chunks) before - // reaching here, so this is a defence-in-depth log line rather - // than a routine path. - int totalChunks = (int) Math.ceil((double) fileBytes.length / UPLOAD_CHUNK_SIZE); - if (totalChunks > 100) { - log.warn("[wecom] Upload would require {} chunks (>100 cap), rejecting: {} ({} bytes)", - totalChunks, fileName, fileBytes.length); - return null; - } - boolean acquired = false; try { acquired = uploadLock.tryAcquire(60, TimeUnit.SECONDS); @@ -2167,6 +1860,7 @@ public class WeComChannelAdapter extends AbstractChannelAdapter { } String md5 = md5Hex(fileBytes); + int totalChunks = (int) Math.ceil((double) fileBytes.length / UPLOAD_CHUNK_SIZE); // Phase 1: Init String initReqId = generateReqId(CMD_UPLOAD_INIT); @@ -2289,8 +1983,7 @@ public class WeComChannelAdapter extends AbstractChannelAdapter { mediaBody.put(mediaType, Map.of("media_id", mediaId)); if (frameReqId != null && !frameReqId.isBlank()) { - // Caller has an explicit inbound frameReqId (the message we're - // replying to is "now") — use that reply slot directly. + // Reply 路径：使用 aibot_respond_msg Map frame = Map.of( "cmd", CMD_RESPONSE, "headers", Map.of("req_id", frameReqId), @@ -2298,12 +1991,15 @@ public class WeComChannelAdapter extends AbstractChannelAdapter { ); sendFrameWithAck(frameReqId, frame); } else { - // Proactive push (cron summary, async-task forward, etc.) — - // delegate to sendOutboundFrame so groups ride the cached - // reqId via aibot_respond_msg (the AI bot platform rejects - // aibot_send_msg in groups). Single chats fall through to - // aibot_send_msg as before. - sendOutboundFrame(targetId, mediaBody); + // 主动推送路径：使用 aibot_send_msg + mediaBody.put("chatid", targetId); + String reqId = generateReqId(CMD_SEND_MSG); + Map frame = Map.of( + "cmd", CMD_SEND_MSG, + "headers", Map.of("req_id", reqId), + "body", mediaBody + ); + sendFrameWithAck(reqId, frame); } } @@ -2559,227 +2255,6 @@ public class WeComChannelAdapter extends AbstractChannelAdapter { return isGroup ? "wecom:group:" + chatId : "wecom:" + senderId; } - // ==================== 引用消息（quote）解析 ==================== - - /** - * Quoted-message extraction result: a human-readable prefix string the - * agent prompt prepends, plus any media (image / file) that was quoted - * and needs to be available as a content part. Either field may be - * empty; {@link #isEmpty()} returns true only when both are. - */ - private record QuoteContext(String prefix, List attachedParts) { - boolean isEmpty() { - return (prefix == null || prefix.isBlank()) && attachedParts.isEmpty(); - } - } - - /** - * Parse the {@code body.quote} field of an inbound WeCom AI Bot frame. - * Quote payloads sit alongside the new message at body level — - * independent of the outer {@code msgtype} — and may themselves carry - * any of text / voice / image / file / mixed. We flatten everything - * into: - *

- * Returns null when {@code body.quote} is absent / empty / malformed — - * the caller treats that the same as "no quote context". - */ - private QuoteContext extractQuoteContext(Map body, String msgId, - String senderId, String chatId, String chatType) { - Object raw = body.get("quote"); - if (!(raw instanceof Map map)) return null; - @SuppressWarnings("unchecked") - Map quote = (Map) map; - String quoteType = (String) quote.getOrDefault("msgtype", ""); - if (quoteType == null || quoteType.isBlank()) return null; - - // Flatten: a "mixed" quote nests its own msg_item array; single-type - // quotes act as a one-element list of themselves. - List> items; - if ("mixed".equals(quoteType)) { - @SuppressWarnings("unchecked") - Map mixed = (Map) quote.getOrDefault("mixed", Map.of()); - @SuppressWarnings("unchecked") - List> mi = (List>) mixed.getOrDefault("msg_item", List.of()); - items = mi; - } else { - items = List.of(quote); - } - - String inboundConvId = inboundConversationId(senderId, chatId, chatType); - StringBuilder summary = new StringBuilder(); - List attached = new ArrayList<>(); - - for (Map item : items) { - String itemType = (String) item.getOrDefault("msgtype", ""); - switch (itemType == null ? "" : itemType) { - case "text" -> { - @SuppressWarnings("unchecked") - Map t = (Map) item.getOrDefault("text", Map.of()); - String content = ((String) t.getOrDefault("content", "")).trim(); - if (!content.isBlank()) { - appendQuoteSummary(summary, content); - } - } - case "voice" -> { - @SuppressWarnings("unchecked") - Map v = (Map) item.getOrDefault("voice", Map.of()); - String asr = ((String) v.getOrDefault("content", "")).trim(); - if (!asr.isBlank()) { - appendQuoteSummary(summary, "[语音] " + asr); - } else { - appendQuoteSummary(summary, "[语音消息]"); - } - } - case "image" -> { - @SuppressWarnings("unchecked") - Map img = (Map) item.getOrDefault("image", Map.of()); - String url = (String) img.getOrDefault("url", ""); - String aesKey = (String) img.getOrDefault("aeskey", ""); - if (!url.isBlank()) { - attached.add(buildInboundImagePart(url, aesKey, msgId, - "quoted_image.jpg", inboundConvId)); - } - appendQuoteSummary(summary, "[图片]"); - } - case "file" -> { - @SuppressWarnings("unchecked") - Map f = (Map) item.getOrDefault("file", Map.of()); - String url = (String) f.getOrDefault("url", ""); - String aesKey = (String) f.getOrDefault("aeskey", ""); - String filename = (String) f.getOrDefault("filename", - f.getOrDefault("file_name", f.getOrDefault("name", "file.bin"))); - if (!url.isBlank()) { - MessageContentPart part = buildInboundFilePart(url, aesKey, msgId, - filename, inboundConvId); - attached.add(part); - if (part.getFileName() != null && !part.getFileName().isBlank()) { - filename = part.getFileName(); - } - } - appendQuoteSummary(summary, "[文件: " + filename + "]"); - } - default -> { - // Unknown quote sub-type — surface the type tag so the - // agent knows something was quoted even if we can't - // unpack it. - if (itemType != null && !itemType.isBlank()) { - appendQuoteSummary(summary, "[" + itemType + "]"); - } - } - } - } - - if (summary.length() == 0 && attached.isEmpty()) { - return null; - } - String prefix = "[引用消息: " + summary + "]\n"; - return new QuoteContext(prefix, attached); - } - - private static void appendQuoteSummary(StringBuilder summary, String fragment) { - if (summary.length() > 0) summary.append(' '); - summary.append(fragment); - } - - // ==================== appmsg 解析（PDF/Word/Excel/链接/小程序）==================== - - /** - * Decoded {@code msgtype=appmsg} payload — text marker the agent reads - * + any attached media parts. Returned by {@link #extractAppmsgContent} - * so the inbound switch can splice it into {@code textContent} and - * {@code contentParts} uniformly. - */ - record AppmsgContent(String text, List attachedParts) {} - - /** - * Parse a {@code msgtype=appmsg} body into a text marker + media parts. - *

- * Supports four variants observed in the WeCom platform: - *

    - *
  • {@code appmsg.file} — forwarded PDF / Word / Excel; reuses - * {@link #buildInboundFilePart} so magic-byte sniff + ZIP - * refinement work identically to plain {@code msgtype=file}
    • - *
  • {@code appmsg.image} — forwarded image card
    • - *
  • {@code appmsg.miniprogram} — miniprogram card; surface the - * title so the agent at least knows what was shared
    • - *
  • {@code appmsg.url} — public-account article / external link; - * flatten title + description + URL into text
    • - *
- * Unknown variants fall back to a {@code [appmsg]} marker so the - * agent isn't completely blind. - *

- * Package-private for unit-test access. - */ - AppmsgContent extractAppmsgContent(Map body, String msgId, - String senderId, String chatId, String chatType) { - @SuppressWarnings("unchecked") - Map appmsg = (Map) body.getOrDefault("appmsg", Map.of()); - String title = ((String) appmsg.getOrDefault("title", "")).trim(); - String desc = ((String) appmsg.getOrDefault("description", "")).trim(); - String linkUrl = ((String) appmsg.getOrDefault("url", "")).trim(); - String inboundConvId = inboundConversationId(senderId, chatId, chatType); - - Object fileObj = appmsg.get("file"); - Object imageObj = appmsg.get("image"); - Object miniObj = appmsg.get("miniprogram"); - - List attached = new ArrayList<>(); - StringBuilder text = new StringBuilder(); - - if (fileObj instanceof Map fileMap) { - @SuppressWarnings("unchecked") - Map fileBody = (Map) fileMap; - String fileUrl = (String) fileBody.getOrDefault("url", ""); - String aesKey = (String) fileBody.getOrDefault("aeskey", ""); - String filename = (String) fileBody.getOrDefault("filename", - fileBody.getOrDefault("file_name", - fileBody.getOrDefault("name", - title.isBlank() ? "file.bin" : title))); - if (!fileUrl.isBlank()) { - MessageContentPart filePart = buildInboundFilePart( - fileUrl, aesKey, msgId, filename, inboundConvId); - attached.add(filePart); - if (filePart.getFileName() != null && !filePart.getFileName().isBlank()) { - filename = filePart.getFileName(); - } - } - text.append("[文件: ").append(filename).append("]"); - } else if (imageObj instanceof Map imgMap) { - @SuppressWarnings("unchecked") - Map imgBody = (Map) imgMap; - String imgUrl = (String) imgBody.getOrDefault("url", ""); - String aesKey = (String) imgBody.getOrDefault("aeskey", ""); - if (!imgUrl.isBlank()) { - attached.add(buildInboundImagePart(imgUrl, aesKey, msgId, - "appmsg_image.jpg", inboundConvId)); - } - text.append("[图片").append(title.isBlank() ? "" : ": " + title).append("]"); - } else if (miniObj instanceof Map miniMap) { - @SuppressWarnings("unchecked") - Map mini = (Map) miniMap; - String miniTitle = ((String) mini.getOrDefault("title", - title.isBlank() ? "未命名小程序" : title)).trim(); - text.append("[小程序: ").append(miniTitle).append("]"); - } else if (!linkUrl.isBlank()) { - text.append("[链接]"); - if (!title.isBlank()) text.append(' ').append(title); - if (!desc.isBlank()) text.append('\n').append(desc); - text.append('\n').append(linkUrl); - } else if (!title.isBlank()) { - text.append("[appmsg: ").append(title).append("]"); - } else { - text.append("[appmsg]"); - } - return new AppmsgContent(text.toString(), attached); - } - /** * Build a fully-populated image content part for inbound WeCom media. *

diff --git a/mateclaw-server/src/main/java/vip/mate/system/controller/SystemSettingController.java b/mateclaw-server/src/main/java/vip/mate/system/controller/SystemSettingController.java index 741f5b8d..9f055aea 100644 --- a/mateclaw-server/src/main/java/vip/mate/system/controller/SystemSettingController.java +++ b/mateclaw-server/src/main/java/vip/mate/system/controller/SystemSettingController.java @@ -41,39 +41,8 @@ public class SystemSettingController { return R.ok(systemSettingService.saveLanguage(request.getLanguage())); } - /** - * Dedicated endpoint for the multimodal sidecar configuration. - *

- * Separated from the bulk {@code PUT /settings} because the bulk endpoint - * now guards sidecar keys with null checks (so unrelated settings pages - * can't clobber them via partial payloads). This endpoint always writes - * both fields, so passing {@code null} for either explicitly clears that - * sidecar — preserving the "clear via UI" UX without leaking the - * write-on-null semantics into every other settings save. - */ - @Operation(summary = "更新多模态 sidecar 配置") - @PutMapping("/sidecar") - public R saveSidecar(@RequestBody SidecarRequest request) { - return R.ok(systemSettingService.updateSidecarSettings( - request.getDefaultVisionModelId(), - request.getDefaultVideoModelId())); - } - @Data public static class LanguageRequest { private String language; } - - /** - * Body for {@code PUT /settings/sidecar}. Both fields are nullable; - * {@code null} means "explicit clear". Field absence in the JSON - * payload also deserializes to null, which is the same outcome — the - * sidecar UI is the only caller of this endpoint and always sends both - * fields, so the absent-vs-null distinction doesn't matter here. - */ - @Data - public static class SidecarRequest { - private Long defaultVisionModelId; - private Long defaultVideoModelId; - } } diff --git a/mateclaw-server/src/main/java/vip/mate/system/service/SystemSettingService.java b/mateclaw-server/src/main/java/vip/mate/system/service/SystemSettingService.java index 1a22ca6a..fa1232d5 100644 --- a/mateclaw-server/src/main/java/vip/mate/system/service/SystemSettingService.java +++ b/mateclaw-server/src/main/java/vip/mate/system/service/SystemSettingService.java @@ -351,47 +351,13 @@ public class SystemSettingService { saveValue(MODEL3D_FALLBACK_ENABLED_KEY, String.valueOf(dto.getModel3dFallbackEnabled()), "3D Provider 级 Fallback"); } - // Multimodal sidecar routing — guarded with null check, matching the - // pattern used for music / 3D / image / video / tts / stt blocks - // above. The bulk PUT /settings is used by every settings page (System, - // Music, Video, Image, Stt, Tts, Model3D), each sending a partial - // payload that omits sidecar fields. Without this guard, saving any - // unrelated setting would silently write "" into the sidecar keys - // (Long? defaultVisionModelId deserializes to null when absent), which - // wiped users' configured vision/video models the moment they touched - // an unrelated settings page. Explicit clearing via the sidecar UI now - // routes through {@link #updateSidecarSettings} instead. - if (dto.getDefaultVisionModelId() != null) { - saveValue(DEFAULT_VISION_MODEL_KEY, - String.valueOf(dto.getDefaultVisionModelId()), - "Default vision-capable model id (mate_model_config.id) for sidecar routing"); - } - if (dto.getDefaultVideoModelId() != null) { - saveValue(DEFAULT_VIDEO_MODEL_KEY, - String.valueOf(dto.getDefaultVideoModelId()), - "Default video-capable model id (mate_model_config.id) for sidecar routing"); - } - return getSettings(); - } - - /** - * Dedicated update path for the multimodal sidecar configuration. - *

- * This endpoint is the ONLY place vision/video model ids can be written - * unconditionally — null is treated as an explicit "clear" and writes - * an empty string (parse-back returns null). The bulk - * {@link #saveSettings} now guards both keys with non-null checks so - * unrelated settings pages can't accidentally clobber sidecar config. - *

- * Both fields are always written so a single API call can independently - * assign / clear either modality. - */ - public SystemSettingsDTO updateSidecarSettings(Long visionModelId, Long videoModelId) { + // Multimodal sidecar routing — write empty string to clear (parse-back returns null) + // Always written so users can revert to "not configured" via the UI. saveValue(DEFAULT_VISION_MODEL_KEY, - visionModelId == null ? "" : String.valueOf(visionModelId), + dto.getDefaultVisionModelId() == null ? "" : String.valueOf(dto.getDefaultVisionModelId()), "Default vision-capable model id (mate_model_config.id) for sidecar routing"); saveValue(DEFAULT_VIDEO_MODEL_KEY, - videoModelId == null ? "" : String.valueOf(videoModelId), + dto.getDefaultVideoModelId() == null ? "" : String.valueOf(dto.getDefaultVideoModelId()), "Default video-capable model id (mate_model_config.id) for sidecar routing"); return getSettings(); } diff --git a/mateclaw-ui/src/api/index.ts b/mateclaw-ui/src/api/index.ts index 6ff54a20..6a33abab 100644 --- a/mateclaw-ui/src/api/index.ts +++ b/mateclaw-ui/src/api/index.ts @@ -518,13 +518,6 @@ export const settingsApi = { update: (data: any) => http.put('/settings', data), getLanguage: () => http.get('/settings/language'), updateLanguage: (language: string) => http.put('/settings/language', { language }), - // Dedicated endpoint for the multimodal sidecar configuration. The bulk - // /settings PUT now guards vision/video model ids with non-null checks so - // unrelated settings pages can't clobber them via partial payloads. This - // endpoint is the only path that writes those fields unconditionally — - // pass {defaultVisionModelId: null} here to explicitly clear a sidecar. - updateSidecar: (data: { defaultVisionModelId: number | null; defaultVideoModelId: number | null }) => - http.put('/settings/sidecar', data), } // ==================== Workspace ==================== diff --git a/mateclaw-ui/src/views/Settings/Models/MultimodalSidecarSection.vue b/mateclaw-ui/src/views/Settings/Models/MultimodalSidecarSection.vue index 842afc45..7f9c7ed9 100644 --- a/mateclaw-ui/src/views/Settings/Models/MultimodalSidecarSection.vue +++ b/mateclaw-ui/src/views/Settings/Models/MultimodalSidecarSection.vue @@ -144,12 +144,7 @@ async function loadAll() { } async function persistSettings(payload: { defaultVisionModelId: number | null; defaultVideoModelId: number | null }) { - // Use the dedicated sidecar endpoint so the bulk /settings PUT can keep - // guarding vision/video keys with non-null checks (preventing unrelated - // settings pages from clobbering this configuration via partial payloads). - // This endpoint always writes both keys, so passing null here means - // "explicit clear" which is the original UX of this card. - await settingsApi.updateSidecar(payload) + await settingsApi.update(payload) } async function onSaveVision() {