feat(auto-reply): 接入万川平台模型配置 + 各模型独立网关回退
万川平台对接 - 新增 wanchuan_proxy.go：WanchuanLogin/WanchuanGetModel 代理登录与按 code 拉取模型， 日志对 password/token/apiKey 打码（含 encryptedConfig 二次解析） - 新增 PlatformConfig（baseUrl/username/password）及 Get/SavePlatformConfig 持久化 - 前端万川卡片：登录→拉取 chat/vision/embedding/rerank/voice→回填 form 并保存→必要时重建向量索引 各模型独立网关（url+key），留空回退聊天网关 - RetrievalConfig 新增 embeddingBaseUrl/embeddingApiKey、rerankBaseUrl/rerankApiKey - embeddingRequestConfig/rerankRequestConfig：优先独立网关，未配置回退 AI.BaseURL/APIKey - vision/audio 同模式：非 DashScope 网关下视觉/语音模型留空时不再锁死或强写 DashScope， 运行期由 fallbackString(VisionModel, Model) 动态复用聊天模型 陈旧向量空间防护 - loadEmbeddingIndex 检测磁盘索引与当前 embedding 模型/维度不一致时清空向量、回退关键词检索， 并提示重建（embeddingIndexStaleReason，兼容旧版无模型名索引） UI 状态修复 - 登录拉模型期间统一置全局 busy，禁用闸门收敛为 busy（与刷新联系人等按钮同范式）， platformBusy 仅保留用于按钮「处理中…」文案，杜绝并发读写 form 与反向可点洞 其他 - 删除遗留 helper/auto_reply_ai.go.bak - 补充 config/helper 单元测试（视觉回退分支、陈旧索引判定）
This commit is contained in:
@@ -1,902 +0,0 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"context"
|
||||
"encoding/json"
|
||||
"fmt"
|
||||
"io"
|
||||
"mime/multipart"
|
||||
"net/http"
|
||||
"net/url"
|
||||
"os"
|
||||
"path/filepath"
|
||||
"strings"
|
||||
"time"
|
||||
|
||||
"qiweimanager/config"
|
||||
)
|
||||
|
||||
type AIResult struct {
|
||||
Answer string `json:"answer"`
|
||||
RawSummary string `json:"rawSummary"`
|
||||
DurationMS int64 `json:"durationMs"`
|
||||
}
|
||||
|
||||
const (
|
||||
aiPromptMaxHits = 8 // 增加到8个片段，提供更多上下文
|
||||
aiPromptMaxChunkRunes = 1200 // 增加到1200字，保留更多细节
|
||||
aiPromptMaxContextRune = 8000 // 增加到8000字，支持更长的知识库内容
|
||||
defaultAudioModel = "qwen3-asr-flash"
|
||||
audioModeAuto = "auto"
|
||||
audioModeOpenAIChat = "openai_audio_chat"
|
||||
audioModeParaformer = "dashscope_paraformer"
|
||||
audioModeTranscription = "local_openai_transcription"
|
||||
audioModeCustomHTTP = "custom_http"
|
||||
)
|
||||
|
||||
func (e *AutoReplyEngine) getConfig() config.AutoReplyConfig {
|
||||
e.mu.Lock()
|
||||
defer e.mu.Unlock()
|
||||
cfg := e.config
|
||||
if cfg.AI.TimeoutSeconds <= 0 {
|
||||
cfg.AI.TimeoutSeconds = 20
|
||||
}
|
||||
if cfg.AI.MaxTokens <= 0 {
|
||||
cfg.AI.MaxTokens = 700
|
||||
}
|
||||
if strings.TrimSpace(cfg.AI.ReplyDetail) == "" {
|
||||
cfg.AI.ReplyDetail = "detailed"
|
||||
}
|
||||
if cfg.Knowledge.TopK <= 0 {
|
||||
cfg.Knowledge.TopK = 3
|
||||
}
|
||||
if cfg.Knowledge.MinScore <= 0 {
|
||||
cfg.Knowledge.MinScore = 0.40
|
||||
}
|
||||
if cfg.ReplyPolicy.UnknownAnswerToken == "" {
|
||||
cfg.ReplyPolicy.UnknownAnswerToken = "NO_ANSWER"
|
||||
}
|
||||
return cfg
|
||||
}
|
||||
|
||||
func (e *AutoReplyEngine) askAI(question string, hits []KnowledgeChunk, msg autoReplyMessage) (*AIResult, error) {
|
||||
cfg := e.getConfig()
|
||||
if strings.TrimSpace(cfg.AI.BaseURL) == "" {
|
||||
return nil, fmt.Errorf("AI Base URL未配置")
|
||||
}
|
||||
if strings.TrimSpace(cfg.AI.Model) == "" {
|
||||
return nil, fmt.Errorf("AI模型未配置")
|
||||
}
|
||||
systemPrompt := buildAutoReplySystemPrompt(cfg)
|
||||
msg.ContextText = e.recentContextPrompt(msg, 6)
|
||||
userPrompt := buildAutoReplyUserPrompt(question, hits, msg, cfg.ReplyPolicy.UnknownAnswerToken)
|
||||
switch strings.ToLower(strings.TrimSpace(cfg.AI.Provider)) {
|
||||
case "local", "ollama":
|
||||
return callOllamaChat(cfg.AI, systemPrompt, userPrompt)
|
||||
default:
|
||||
return callOpenAICompatibleChat(cfg.AI, systemPrompt, userPrompt)
|
||||
}
|
||||
}
|
||||
|
||||
func (e *AutoReplyEngine) askGeneralAI(question string, msg autoReplyMessage) (*AIResult, error) {
|
||||
cfg := e.getConfig()
|
||||
if strings.TrimSpace(cfg.AI.BaseURL) == "" {
|
||||
return nil, fmt.Errorf("AI Base URL未配置")
|
||||
}
|
||||
if strings.TrimSpace(cfg.AI.Model) == "" {
|
||||
return nil, fmt.Errorf("AI模型未配置")
|
||||
}
|
||||
systemPrompt := buildGeneralAutoReplySystemPrompt(cfg)
|
||||
msg.ContextText = e.recentContextPrompt(msg, 6)
|
||||
userPrompt := buildGeneralAutoReplyUserPrompt(question, msg)
|
||||
switch strings.ToLower(strings.TrimSpace(cfg.AI.Provider)) {
|
||||
case "local", "ollama":
|
||||
return callOllamaChat(cfg.AI, systemPrompt, userPrompt)
|
||||
default:
|
||||
return callOpenAICompatibleChat(cfg.AI, systemPrompt, userPrompt)
|
||||
}
|
||||
}
|
||||
|
||||
func (e *AutoReplyEngine) askNonTextAI(msg autoReplyMessage) (*AIResult, error) {
|
||||
cfg := e.getConfig()
|
||||
if strings.TrimSpace(cfg.AI.BaseURL) == "" {
|
||||
return nil, fmt.Errorf("AI Base URL未配置")
|
||||
}
|
||||
if strings.TrimSpace(cfg.AI.Model) == "" {
|
||||
return nil, fmt.Errorf("AI模型未配置")
|
||||
}
|
||||
systemPrompt := buildNonTextAutoReplySystemPrompt(cfg)
|
||||
userPrompt := buildNonTextAutoReplyUserPrompt(msg)
|
||||
switch strings.ToLower(strings.TrimSpace(cfg.AI.Provider)) {
|
||||
case "local", "ollama":
|
||||
return callOllamaChat(cfg.AI, systemPrompt, userPrompt)
|
||||
default:
|
||||
if mediaURL := strings.TrimSpace(msg.MediaURL); mediaURL != "" {
|
||||
return callOpenAICompatibleVisionChat(cfg.AI, systemPrompt, userPrompt, mediaURL)
|
||||
}
|
||||
return callOpenAICompatibleChat(cfg.AI, systemPrompt, userPrompt)
|
||||
}
|
||||
}
|
||||
|
||||
func (e *AutoReplyEngine) testAIConnection() (*AIResult, error) {
|
||||
testMsg := autoReplyMessage{
|
||||
FromNickName: "测试客户",
|
||||
ConversationID: "test",
|
||||
}
|
||||
hits := []KnowledgeChunk{{
|
||||
Source: "test.md",
|
||||
Content: "测试知识：自动客服连接测试时，请回复“连接正常”。",
|
||||
Score: 1,
|
||||
}}
|
||||
return e.askAI("请回复连接正常", hits, testMsg)
|
||||
}
|
||||
|
||||
func buildAutoReplySystemPrompt(cfg config.AutoReplyConfig) string {
|
||||
token := cfg.ReplyPolicy.UnknownAnswerToken
|
||||
if token == "" {
|
||||
token = "NO_ANSWER"
|
||||
}
|
||||
return prependAISystemPrompt(cfg, "你是企业微信售后客服助手。只能根据提供的知识库片段回答客户问题。"+replyDetailInstruction(cfg)+"知识库不足以确定答案时，只输出 "+token+"。不要编造政策、价格、承诺、库存或物流时效。客户要求人工、投诉、退款、合同、发票、赔偿或价格特殊审批时，也只输出 "+token+"。")
|
||||
}
|
||||
|
||||
func buildGeneralAutoReplySystemPrompt(cfg config.AutoReplyConfig) string {
|
||||
token := cfg.ReplyPolicy.UnknownAnswerToken
|
||||
if token == "" {
|
||||
token = "NO_ANSWER"
|
||||
}
|
||||
return prependAISystemPrompt(cfg, "你是企业微信智能客服助手。请用中文自然、和蔼地回答普通问候、身份介绍和日常沟通问题。"+replyDetailInstruction(cfg)+"不要冒充真人，不要编造产品参数、价格、政策、库存、物流、合同、发票或售后结论。遇到需要公司专有资料、知识库、人工审批或无法确认的信息时，不要硬编，可以温和说明会按资料核对或请客户补充具体问题。不要输出 "+token+"，除非客户明确要求停止回复。")
|
||||
}
|
||||
|
||||
func buildNonTextAutoReplySystemPrompt(cfg config.AutoReplyConfig) string {
|
||||
return prependAISystemPrompt(cfg, "你是企业微信客服岗位助手。用户发来非文本消息时，请根据消息类型和文字描述判断是否属于客服岗位可处理范围。范围内包括产品咨询、订单、售后、方案资料、使用问题、客户服务沟通；可回复时要自然、和蔼。"+replyDetailInstruction(cfg)+"不要编造图片里不存在的信息。若无法判断图片/表情内容，礼貌请客户补充文字说明。若明显超出客服岗位范围，只能回复：抱歉，你这问题超出我的岗位认知了，回答不了。不要主动转人工，除非客户明确要求人工。")
|
||||
}
|
||||
|
||||
func buildVisionRecognitionSystemPrompt(cfg config.AutoReplyConfig) string {
|
||||
return prependAISystemPrompt(cfg, "你是企业微信客服岗位的图片识别助手。请识别客户发来的图片/表情/封面中与客服沟通有关的内容，输出一句简洁中文描述；如果明显不是客服岗位可处理的内容，也请说明其大概内容。不要编造看不见的信息。")
|
||||
}
|
||||
|
||||
func prependAISystemPrompt(cfg config.AutoReplyConfig, base string) string {
|
||||
identity := strings.TrimSpace(cfg.AI.SystemPrompt)
|
||||
if identity == "" {
|
||||
identity = "你是一名企业微信智能客服。"
|
||||
}
|
||||
return identity + "\n" + base
|
||||
}
|
||||
|
||||
func replyDetailInstruction(cfg config.AutoReplyConfig) string {
|
||||
switch strings.ToLower(strings.TrimSpace(cfg.AI.ReplyDetail)) {
|
||||
case "concise":
|
||||
return "回复保持简洁，通常1-2句，约80-140个中文字符；先回答结论，必要时补一句下一步建议。"
|
||||
case "medium":
|
||||
return "回复详细程度适中，通常2-4句，约160-280个中文字符；先回答结论，再说明关键原因或注意事项，最后给出下一步建议。"
|
||||
default:
|
||||
return "回复尽量详细但不要啰嗦，通常3-6句，约280-500个中文字符；先明确回答客户问题，再结合可用资料说明关键点、适用场景或限制，最后给出具体下一步建议。"
|
||||
}
|
||||
}
|
||||
|
||||
func effectiveReplyMaxTokens(cfg config.AIConfig) int {
|
||||
maxTokens := cfg.MaxTokens
|
||||
switch strings.ToLower(strings.TrimSpace(cfg.ReplyDetail)) {
|
||||
case "concise":
|
||||
if maxTokens < 220 {
|
||||
return 220
|
||||
}
|
||||
case "medium":
|
||||
if maxTokens < 450 {
|
||||
return 450
|
||||
}
|
||||
default:
|
||||
if maxTokens < 700 {
|
||||
return 700
|
||||
}
|
||||
}
|
||||
return maxTokens
|
||||
}
|
||||
|
||||
func buildGeneralAutoReplyUserPrompt(question string, msg autoReplyMessage) string {
|
||||
var b strings.Builder
|
||||
b.WriteString("客户昵称：")
|
||||
if msg.FromNickName != "" {
|
||||
b.WriteString(msg.FromNickName)
|
||||
} else {
|
||||
b.WriteString("未知")
|
||||
}
|
||||
b.WriteString("\n客户问题：")
|
||||
b.WriteString(question)
|
||||
if contextText := strings.TrimSpace(msg.ContextText); contextText != "" {
|
||||
b.WriteString("\n\n最近对话上下文：\n")
|
||||
b.WriteString(contextText)
|
||||
}
|
||||
b.WriteString("\n请直接给客户一条友好、可发送的回复。")
|
||||
return b.String()
|
||||
}
|
||||
|
||||
func buildNonTextAutoReplyUserPrompt(msg autoReplyMessage) string {
|
||||
var b strings.Builder
|
||||
b.WriteString("客户昵称：")
|
||||
if msg.FromNickName != "" {
|
||||
b.WriteString(msg.FromNickName)
|
||||
} else {
|
||||
b.WriteString("未知")
|
||||
}
|
||||
b.WriteString("\n消息类型：")
|
||||
b.WriteString(msg.MessageType)
|
||||
b.WriteString("\n原始类型：")
|
||||
b.WriteString(fmt.Sprintf("%d", msg.RawType))
|
||||
b.WriteString("\n消息描述：")
|
||||
if strings.TrimSpace(msg.Content) != "" {
|
||||
b.WriteString(msg.Content)
|
||||
} else {
|
||||
b.WriteString("无文字描述")
|
||||
}
|
||||
if strings.TrimSpace(msg.MediaURL) != "" {
|
||||
b.WriteString("\n媒体地址：")
|
||||
b.WriteString(msg.MediaURL)
|
||||
}
|
||||
b.WriteString("\n请直接给客户一条可发送的回复。")
|
||||
return b.String()
|
||||
}
|
||||
|
||||
func buildAutoReplyUserPrompt(question string, hits []KnowledgeChunk, msg autoReplyMessage, noAnswerToken string) string {
|
||||
noAnswerToken = strings.TrimSpace(noAnswerToken)
|
||||
if noAnswerToken == "" {
|
||||
noAnswerToken = "NO_ANSWER"
|
||||
}
|
||||
var b strings.Builder
|
||||
b.WriteString("客户昵称：")
|
||||
if msg.FromNickName != "" {
|
||||
b.WriteString(msg.FromNickName)
|
||||
} else {
|
||||
b.WriteString("未知")
|
||||
}
|
||||
b.WriteString("\n客户问题：")
|
||||
b.WriteString(question)
|
||||
if contextText := strings.TrimSpace(msg.ContextText); contextText != "" {
|
||||
b.WriteString("\n\n最近对话上下文：\n")
|
||||
b.WriteString(contextText)
|
||||
}
|
||||
b.WriteString("\n\n知识库片段：\n")
|
||||
for i, hit := range compactKnowledgeHitsForAI(hits) {
|
||||
b.WriteString(fmt.Sprintf("[%d] 来源：%s 分数：%.3f\n%s\n\n", i+1, hit.Source, hit.Score, hit.Content))
|
||||
}
|
||||
b.WriteString("\u53ea\u80fd\u4f7f\u7528\u4e0a\u9762\u7247\u6bb5\u4e2d\u660e\u786e\u51fa\u73b0\u7684\u4e8b\u5b9e\u56de\u7b54\uff1b\u5982\u679c\u8be2\u95ee\u90e8\u95e8\u3001\u4f1a\u8bae\u65f6\u95f4\u3001\u6807\u51c6\u6216\u89c4\u5b9a\uff0c\u53ea\u80fd\u5217\u51fa\u7247\u6bb5\u91cc\u76f4\u63a5\u51fa\u73b0\u7684\u503c\uff0c\u4e0d\u5f97\u6839\u636e\u5e38\u8bc6\u8865\u5145\u5176\u4ed6\u90e8\u95e8\u6216\u65f6\u95f4\u3002\n")
|
||||
if isGenericProductQuery(question) {
|
||||
b.WriteString("客户在泛问产品时，请优先按知识库列出具体产品或型号，每项用一句话说明定位，最后询问客户更关注硬件、模型还是AI应用。不要只概括为几大类。无法确认时只输出 ")
|
||||
} else {
|
||||
b.WriteString("请基于知识库片段回答客户。无法确认时只输出 ")
|
||||
}
|
||||
b.WriteString(noAnswerToken)
|
||||
b.WriteString("。")
|
||||
return b.String()
|
||||
}
|
||||
|
||||
func compactKnowledgeHitsForAI(hits []KnowledgeChunk) []KnowledgeChunk {
|
||||
if len(hits) == 0 {
|
||||
return nil
|
||||
}
|
||||
limit := aiPromptMaxHits
|
||||
if len(hits) < limit {
|
||||
limit = len(hits)
|
||||
}
|
||||
result := make([]KnowledgeChunk, 0, limit)
|
||||
totalRunes := 0
|
||||
for i := 0; i < limit; i++ {
|
||||
hit := hits[i]
|
||||
content := strings.TrimSpace(hit.Content)
|
||||
if content == "" {
|
||||
continue
|
||||
}
|
||||
content = truncateTextForPrompt(content, aiPromptMaxChunkRunes)
|
||||
remaining := aiPromptMaxContextRune - totalRunes
|
||||
if remaining <= 0 {
|
||||
break
|
||||
}
|
||||
if len([]rune(content)) > remaining {
|
||||
content = truncateTextForPrompt(content, remaining)
|
||||
}
|
||||
hit.Content = content
|
||||
totalRunes += len([]rune(content))
|
||||
result = append(result, hit)
|
||||
}
|
||||
return result
|
||||
}
|
||||
|
||||
func truncateTextForPrompt(text string, max int) string {
|
||||
if max <= 0 {
|
||||
return ""
|
||||
}
|
||||
runes := []rune(text)
|
||||
if len(runes) <= max {
|
||||
return text
|
||||
}
|
||||
return string(runes[:max])
|
||||
}
|
||||
|
||||
func callOpenAICompatibleChat(cfg config.AIConfig, systemPrompt string, userPrompt string) (*AIResult, error) {
|
||||
url := strings.TrimRight(cfg.BaseURL, "/")
|
||||
if !strings.HasSuffix(url, "/chat/completions") {
|
||||
url += "/chat/completions"
|
||||
}
|
||||
payload := map[string]interface{}{
|
||||
"model": cfg.Model,
|
||||
"temperature": cfg.Temperature,
|
||||
"max_tokens": effectiveReplyMaxTokens(cfg),
|
||||
"enable_thinking": cfg.EnableThinking,
|
||||
"messages": []map[string]string{
|
||||
{"role": "system", "content": systemPrompt},
|
||||
{"role": "user", "content": userPrompt},
|
||||
},
|
||||
}
|
||||
var response struct {
|
||||
Choices []struct {
|
||||
Message struct {
|
||||
Content string `json:"content"`
|
||||
} `json:"message"`
|
||||
} `json:"choices"`
|
||||
Error interface{} `json:"error"`
|
||||
}
|
||||
result, err := doAIJSONRequest(cfg, url, payload, &response)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
if response.Error != nil {
|
||||
return nil, fmt.Errorf("AI返回错误: %v", response.Error)
|
||||
}
|
||||
if len(response.Choices) == 0 {
|
||||
return nil, fmt.Errorf("AI返回空choices")
|
||||
}
|
||||
answer := strings.TrimSpace(response.Choices[0].Message.Content)
|
||||
result.Answer = answer
|
||||
result.RawSummary = truncateText(answer, 160)
|
||||
return result, nil
|
||||
}
|
||||
|
||||
func callOpenAICompatibleVisionChat(cfg config.AIConfig, systemPrompt string, userPrompt string, imageURL string) (*AIResult, error) {
|
||||
visionCfg := visionRequestConfig(cfg)
|
||||
url := strings.TrimRight(visionCfg.BaseURL, "/")
|
||||
if !strings.HasSuffix(url, "/chat/completions") {
|
||||
url += "/chat/completions"
|
||||
}
|
||||
payload := map[string]interface{}{
|
||||
"model": visionCfg.Model,
|
||||
"temperature": visionCfg.Temperature,
|
||||
"max_tokens": visionCfg.MaxTokens,
|
||||
"enable_thinking": visionCfg.EnableThinking,
|
||||
"messages": []map[string]interface{}{
|
||||
{"role": "system", "content": systemPrompt},
|
||||
{
|
||||
"role": "user",
|
||||
"content": []map[string]interface{}{
|
||||
{"type": "text", "text": userPrompt},
|
||||
{"type": "image_url", "image_url": map[string]string{"url": imageURL}},
|
||||
},
|
||||
},
|
||||
},
|
||||
}
|
||||
var response struct {
|
||||
Choices []struct {
|
||||
Message struct {
|
||||
Content string `json:"content"`
|
||||
} `json:"message"`
|
||||
} `json:"choices"`
|
||||
Error interface{} `json:"error"`
|
||||
}
|
||||
result, err := doAIJSONRequest(visionCfg, url, payload, &response)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
if response.Error != nil {
|
||||
return nil, fmt.Errorf("AI返回错误: %v", response.Error)
|
||||
}
|
||||
if len(response.Choices) == 0 {
|
||||
return nil, fmt.Errorf("AI返回空choices")
|
||||
}
|
||||
answer := strings.TrimSpace(response.Choices[0].Message.Content)
|
||||
result.Answer = answer
|
||||
result.RawSummary = truncateText(answer, 160)
|
||||
return result, nil
|
||||
}
|
||||
|
||||
func visionRequestConfig(cfg config.AIConfig) config.AIConfig {
|
||||
visionCfg := cfg
|
||||
visionCfg.Model = fallbackString(cfg.VisionModel, cfg.Model)
|
||||
if strings.TrimSpace(cfg.VisionBaseURL) != "" {
|
||||
visionCfg.BaseURL = strings.TrimSpace(cfg.VisionBaseURL)
|
||||
}
|
||||
visionKey := strings.TrimSpace(cfg.VisionAPIKey)
|
||||
if visionKey != "" && !looksLikeURL(visionKey) {
|
||||
visionCfg.APIKey = visionKey
|
||||
}
|
||||
return visionCfg
|
||||
}
|
||||
|
||||
func callOpenAICompatibleAudioChatTranscription(cfg config.AIConfig, audioPath string) (string, error) {
|
||||
audioCfg := audioRequestConfig(cfg)
|
||||
audioDataURL, err := audioDataURLFromFile(audioPath)
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
url := strings.TrimRight(audioCfg.BaseURL, "/")
|
||||
if !strings.HasSuffix(url, "/chat/completions") {
|
||||
url += "/chat/completions"
|
||||
}
|
||||
model := fallbackString(audioCfg.Model, defaultAudioModel)
|
||||
payload := map[string]interface{}{
|
||||
"model": model,
|
||||
"temperature": 0,
|
||||
"max_tokens": audioCfg.MaxTokens,
|
||||
"enable_thinking": false,
|
||||
"messages": []map[string]interface{}{
|
||||
{
|
||||
"role": "user",
|
||||
"content": audioChatContentForModel(model, audioDataURL),
|
||||
},
|
||||
},
|
||||
}
|
||||
var response struct {
|
||||
Choices []struct {
|
||||
Message struct {
|
||||
Content string `json:"content"`
|
||||
} `json:"message"`
|
||||
} `json:"choices"`
|
||||
Error interface{} `json:"error"`
|
||||
}
|
||||
if _, err := doAIJSONRequest(audioCfg, url, payload, &response); err != nil {
|
||||
return "", fmt.Errorf("audio chat transcription failed (model=%s endpoint=%s): %w", audioCfg.Model, url, err)
|
||||
}
|
||||
if response.Error != nil {
|
||||
return "", fmt.Errorf("audio chat transcription failed (model=%s endpoint=%s): %v", audioCfg.Model, url, response.Error)
|
||||
}
|
||||
if len(response.Choices) == 0 {
|
||||
return "", fmt.Errorf("audio chat transcription failed (model=%s endpoint=%s): empty choices", audioCfg.Model, url)
|
||||
}
|
||||
text := strings.TrimSpace(response.Choices[0].Message.Content)
|
||||
if text == "" {
|
||||
return "", fmt.Errorf("audio chat transcription failed (model=%s endpoint=%s): empty text", audioCfg.Model, url)
|
||||
}
|
||||
return text, nil
|
||||
}
|
||||
|
||||
func audioChatContentForModel(model string, audioDataURL string) []map[string]interface{} {
|
||||
if isQwenASRModel(model) {
|
||||
return []map[string]interface{}{
|
||||
{"type": "input_audio", "input_audio": audioDataURL},
|
||||
}
|
||||
}
|
||||
return []map[string]interface{}{
|
||||
{"type": "text", "text": "请把这段语音转写成简体中文文本，只输出转写内容，不要解释。"},
|
||||
{"type": "input_audio", "input_audio": map[string]interface{}{"data": audioDataURL}},
|
||||
}
|
||||
}
|
||||
|
||||
func isQwenASRModel(model string) bool {
|
||||
name := strings.ToLower(strings.TrimSpace(model))
|
||||
return strings.HasPrefix(name, "qwen3-asr") || strings.HasPrefix(name, "qwen-asr")
|
||||
}
|
||||
|
||||
func audioRequestConfig(cfg config.AIConfig) config.AIConfig {
|
||||
audioCfg := cfg
|
||||
audioCfg.Model = fallbackString(cfg.AudioModel, defaultAudioModel)
|
||||
if strings.TrimSpace(cfg.AudioBaseURL) != "" {
|
||||
audioCfg.BaseURL = strings.TrimSpace(cfg.AudioBaseURL)
|
||||
}
|
||||
audioKey := strings.TrimSpace(cfg.AudioAPIKey)
|
||||
if audioKey != "" && !looksLikeURL(audioKey) {
|
||||
audioCfg.APIKey = audioKey
|
||||
}
|
||||
audioCfg.EnableThinking = false
|
||||
audioCfg.Temperature = 0
|
||||
return audioCfg
|
||||
}
|
||||
|
||||
func audioConfigWarning(cfg config.AIConfig) string {
|
||||
if looksLikeURL(strings.TrimSpace(cfg.AudioAPIKey)) {
|
||||
return "语音 API Key 误填为 URL，已忽略该值并复用主 API Key"
|
||||
}
|
||||
return ""
|
||||
}
|
||||
|
||||
func inferAudioMode(cfg config.AIConfig) string {
|
||||
mode := normalizeAudioMode(cfg.AudioMode)
|
||||
if mode != audioModeAuto {
|
||||
return mode
|
||||
}
|
||||
provider := normalizeAudioMode(cfg.AudioProvider)
|
||||
if provider != audioModeAuto {
|
||||
return provider
|
||||
}
|
||||
model := strings.ToLower(strings.TrimSpace(cfg.AudioModel))
|
||||
if strings.HasPrefix(model, "paraformer") {
|
||||
return audioModeParaformer
|
||||
}
|
||||
if strings.Contains(model, "whisper") || strings.Contains(model, "transcribe") {
|
||||
return audioModeTranscription
|
||||
}
|
||||
return audioModeOpenAIChat
|
||||
}
|
||||
|
||||
func normalizeAudioMode(value string) string {
|
||||
switch strings.ToLower(strings.TrimSpace(value)) {
|
||||
case "", audioModeAuto:
|
||||
return audioModeAuto
|
||||
case "openai", "openai_chat", "audio_chat", "qwen_audio", "qwen3_asr", audioModeOpenAIChat:
|
||||
return audioModeOpenAIChat
|
||||
case "dashscope", "paraformer", audioModeParaformer:
|
||||
return audioModeParaformer
|
||||
case "transcription", "openai_transcription", "local", "local_asr", audioModeTranscription:
|
||||
return audioModeTranscription
|
||||
case "custom", audioModeCustomHTTP:
|
||||
return audioModeCustomHTTP
|
||||
default:
|
||||
return audioModeAuto
|
||||
}
|
||||
}
|
||||
|
||||
func looksLikeURL(value string) bool {
|
||||
value = strings.TrimSpace(value)
|
||||
return strings.HasPrefix(strings.ToLower(value), "http://") || strings.HasPrefix(strings.ToLower(value), "https://")
|
||||
}
|
||||
|
||||
func supportsSilkDirectly(cfg config.AIConfig) bool {
|
||||
model := strings.ToLower(strings.TrimSpace(cfg.AudioModel))
|
||||
mode := inferAudioMode(cfg)
|
||||
if mode == audioModeParaformer || mode == audioModeTranscription || mode == audioModeCustomHTTP {
|
||||
return false
|
||||
}
|
||||
return strings.Contains(model, "silk")
|
||||
}
|
||||
|
||||
func dashScopeAPIBaseURL(cfg config.AIConfig) string {
|
||||
base := strings.TrimSpace(cfg.AudioBaseURL)
|
||||
if base == "" {
|
||||
base = strings.TrimSpace(cfg.BaseURL)
|
||||
}
|
||||
if base == "" || strings.Contains(base, "/compatible-mode/") {
|
||||
return "https://dashscope.aliyuncs.com/api/v1"
|
||||
}
|
||||
base = strings.TrimRight(base, "/")
|
||||
if strings.HasSuffix(base, "/services/audio/asr/transcription") {
|
||||
return strings.TrimSuffix(base, "/services/audio/asr/transcription")
|
||||
}
|
||||
if strings.Contains(base, "/api/v1/") {
|
||||
return strings.Split(base, "/api/v1/")[0] + "/api/v1"
|
||||
}
|
||||
if strings.HasSuffix(base, "/api/v1") {
|
||||
return base
|
||||
}
|
||||
return base
|
||||
}
|
||||
|
||||
func callOpenAICompatibleAudioTranscription(cfg config.AIConfig, audioPath string) (string, error) {
|
||||
cfg = audioRequestConfig(cfg)
|
||||
url := strings.TrimRight(cfg.BaseURL, "/")
|
||||
if !strings.HasSuffix(url, "/audio/transcriptions") {
|
||||
url += "/audio/transcriptions"
|
||||
}
|
||||
timeout := time.Duration(cfg.TimeoutSeconds) * time.Second
|
||||
if timeout <= 0 {
|
||||
timeout = 20 * time.Second
|
||||
}
|
||||
file, err := os.Open(audioPath)
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
defer file.Close()
|
||||
body := &bytes.Buffer{}
|
||||
writer := multipart.NewWriter(body)
|
||||
if err := writer.WriteField("model", cfg.Model); err != nil {
|
||||
return "", err
|
||||
}
|
||||
part, err := writer.CreateFormFile("file", filepath.Base(audioPath))
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
if _, err := io.Copy(part, file); err != nil {
|
||||
return "", err
|
||||
}
|
||||
if err := writer.Close(); err != nil {
|
||||
return "", err
|
||||
}
|
||||
ctx, cancel := context.WithTimeout(context.Background(), timeout)
|
||||
defer cancel()
|
||||
req, err := http.NewRequestWithContext(ctx, "POST", url, body)
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
req.Header.Set("Content-Type", writer.FormDataContentType())
|
||||
if strings.TrimSpace(cfg.APIKey) != "" {
|
||||
req.Header.Set("Authorization", "Bearer "+strings.TrimSpace(cfg.APIKey))
|
||||
}
|
||||
resp, err := (&http.Client{Timeout: timeout}).Do(req)
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
defer resp.Body.Close()
|
||||
respBody, err := io.ReadAll(resp.Body)
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
if resp.StatusCode < 200 || resp.StatusCode >= 300 {
|
||||
return "", fmt.Errorf("audio transcription failed (model=%s endpoint=%s): HTTP status %d, body=%s", cfg.Model, url, resp.StatusCode, truncateText(string(respBody), 240))
|
||||
}
|
||||
var parsed struct {
|
||||
Text string `json:"text"`
|
||||
Error interface{} `json:"error"`
|
||||
}
|
||||
if err := json.Unmarshal(respBody, &parsed); err != nil {
|
||||
return "", fmt.Errorf("parse audio transcription failed (model=%s endpoint=%s): %v, body=%s", cfg.Model, url, err, truncateText(string(respBody), 240))
|
||||
}
|
||||
if parsed.Error != nil {
|
||||
return "", fmt.Errorf("audio transcription failed (model=%s endpoint=%s): %v", cfg.Model, url, parsed.Error)
|
||||
}
|
||||
text := strings.TrimSpace(parsed.Text)
|
||||
if text == "" {
|
||||
return "", fmt.Errorf("audio transcription failed (model=%s endpoint=%s): empty text", cfg.Model, url)
|
||||
}
|
||||
return text, nil
|
||||
}
|
||||
|
||||
func callDashScopeParaformerTranscription(cfg config.AIConfig, fileURL string) (string, error) {
|
||||
cfg = audioRequestConfig(cfg)
|
||||
fileURL = strings.TrimSpace(fileURL)
|
||||
if fileURL == "" {
|
||||
return "", fmt.Errorf("paraformer transcription failed (model=%s): 需要公网可访问的音频 URL，本地文件不能直接提交给 Paraformer RESTful 接口", cfg.Model)
|
||||
}
|
||||
parsedURL, err := url.Parse(fileURL)
|
||||
if err != nil || (parsedURL.Scheme != "http" && parsedURL.Scheme != "https" && parsedURL.Scheme != "oss") {
|
||||
return "", fmt.Errorf("paraformer transcription failed (model=%s): 音频 URL 无效", cfg.Model)
|
||||
}
|
||||
base := dashScopeAPIBaseURL(cfg)
|
||||
submitURL := strings.TrimRight(base, "/") + "/services/audio/asr/transcription"
|
||||
payload := map[string]interface{}{
|
||||
"model": fallbackString(cfg.Model, "paraformer-v2"),
|
||||
"input": map[string]interface{}{
|
||||
"file_urls": []string{fileURL},
|
||||
},
|
||||
"parameters": map[string]interface{}{
|
||||
"channel_id": []int{0},
|
||||
"language_hints": []string{"zh", "en"},
|
||||
},
|
||||
}
|
||||
var submitResp struct {
|
||||
Output struct {
|
||||
TaskID string `json:"task_id"`
|
||||
TaskStatus string `json:"task_status"`
|
||||
} `json:"output"`
|
||||
Code string `json:"code"`
|
||||
Message string `json:"message"`
|
||||
}
|
||||
if err := doDashScopeJSONRequest(cfg, submitURL, "POST", payload, true, &submitResp); err != nil {
|
||||
return "", fmt.Errorf("paraformer transcription submit failed (model=%s endpoint=%s): %w", cfg.Model, submitURL, err)
|
||||
}
|
||||
if submitResp.Code != "" || submitResp.Message != "" {
|
||||
return "", fmt.Errorf("paraformer transcription submit failed (model=%s endpoint=%s): %s %s", cfg.Model, submitURL, submitResp.Code, submitResp.Message)
|
||||
}
|
||||
taskID := strings.TrimSpace(submitResp.Output.TaskID)
|
||||
if taskID == "" {
|
||||
return "", fmt.Errorf("paraformer transcription submit failed (model=%s endpoint=%s): empty task_id", cfg.Model, submitURL)
|
||||
}
|
||||
return waitDashScopeParaformerTask(cfg, base, taskID)
|
||||
}
|
||||
|
||||
func waitDashScopeParaformerTask(cfg config.AIConfig, base string, taskID string) (string, error) {
|
||||
timeout := time.Duration(cfg.TimeoutSeconds) * time.Second
|
||||
if timeout <= 0 {
|
||||
timeout = 20 * time.Second
|
||||
}
|
||||
deadline := time.Now().Add(timeout)
|
||||
queryURL := strings.TrimRight(base, "/") + "/tasks/" + url.PathEscape(taskID)
|
||||
var lastStatus string
|
||||
for time.Now().Before(deadline) {
|
||||
var queryResp struct {
|
||||
Output struct {
|
||||
TaskStatus string `json:"task_status"`
|
||||
Results []struct {
|
||||
FileURL string `json:"file_url"`
|
||||
TranscriptionURL string `json:"transcription_url"`
|
||||
SubtaskStatus string `json:"subtask_status"`
|
||||
Code string `json:"code"`
|
||||
Message string `json:"message"`
|
||||
} `json:"results"`
|
||||
} `json:"output"`
|
||||
Code string `json:"code"`
|
||||
Message string `json:"message"`
|
||||
}
|
||||
if err := doDashScopeJSONRequest(cfg, queryURL, "GET", nil, false, &queryResp); err != nil {
|
||||
return "", fmt.Errorf("paraformer transcription query failed (model=%s endpoint=%s task=%s): %w", cfg.Model, queryURL, taskID, err)
|
||||
}
|
||||
if queryResp.Code != "" || queryResp.Message != "" {
|
||||
return "", fmt.Errorf("paraformer transcription query failed (model=%s endpoint=%s task=%s): %s %s", cfg.Model, queryURL, taskID, queryResp.Code, queryResp.Message)
|
||||
}
|
||||
lastStatus = strings.ToUpper(strings.TrimSpace(queryResp.Output.TaskStatus))
|
||||
switch lastStatus {
|
||||
case "SUCCEEDED":
|
||||
for _, result := range queryResp.Output.Results {
|
||||
if strings.EqualFold(result.SubtaskStatus, "SUCCEEDED") && strings.TrimSpace(result.TranscriptionURL) != "" {
|
||||
return downloadDashScopeTranscriptionResult(cfg, result.TranscriptionURL)
|
||||
}
|
||||
if result.Code != "" || result.Message != "" {
|
||||
return "", fmt.Errorf("paraformer transcription subtask failed (model=%s task=%s): %s %s", cfg.Model, taskID, result.Code, result.Message)
|
||||
}
|
||||
}
|
||||
return "", fmt.Errorf("paraformer transcription finished without usable result (model=%s task=%s)", cfg.Model, taskID)
|
||||
case "FAILED", "CANCELED", "UNKNOWN":
|
||||
return "", fmt.Errorf("paraformer transcription task failed (model=%s task=%s status=%s)", cfg.Model, taskID, lastStatus)
|
||||
}
|
||||
time.Sleep(500 * time.Millisecond)
|
||||
}
|
||||
return "", fmt.Errorf("paraformer transcription timed out (model=%s task=%s last_status=%s)", cfg.Model, taskID, lastStatus)
|
||||
}
|
||||
|
||||
func downloadDashScopeTranscriptionResult(cfg config.AIConfig, resultURL string) (string, error) {
|
||||
timeout := time.Duration(cfg.TimeoutSeconds) * time.Second
|
||||
if timeout <= 0 {
|
||||
timeout = 20 * time.Second
|
||||
}
|
||||
ctx, cancel := context.WithTimeout(context.Background(), timeout)
|
||||
defer cancel()
|
||||
req, err := http.NewRequestWithContext(ctx, "GET", resultURL, nil)
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
resp, err := (&http.Client{Timeout: timeout}).Do(req)
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
defer resp.Body.Close()
|
||||
respBody, err := io.ReadAll(resp.Body)
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
if resp.StatusCode < 200 || resp.StatusCode >= 300 {
|
||||
return "", fmt.Errorf("download paraformer result failed: HTTP status %d, body=%s", resp.StatusCode, truncateText(string(respBody), 240))
|
||||
}
|
||||
var parsed struct {
|
||||
Transcripts []struct {
|
||||
Text string `json:"text"`
|
||||
} `json:"transcripts"`
|
||||
}
|
||||
if err := json.Unmarshal(respBody, &parsed); err != nil {
|
||||
return "", fmt.Errorf("parse paraformer result failed: %v, body=%s", err, truncateText(string(respBody), 240))
|
||||
}
|
||||
parts := make([]string, 0, len(parsed.Transcripts))
|
||||
for _, transcript := range parsed.Transcripts {
|
||||
if text := strings.TrimSpace(transcript.Text); text != "" {
|
||||
parts = append(parts, text)
|
||||
}
|
||||
}
|
||||
text := strings.TrimSpace(strings.Join(parts, "\n"))
|
||||
if text == "" {
|
||||
return "", fmt.Errorf("paraformer result returned empty text")
|
||||
}
|
||||
return text, nil
|
||||
}
|
||||
|
||||
func doDashScopeJSONRequest(cfg config.AIConfig, endpoint string, method string, payload interface{}, async bool, out interface{}) error {
|
||||
timeout := time.Duration(cfg.TimeoutSeconds) * time.Second
|
||||
if timeout <= 0 {
|
||||
timeout = 20 * time.Second
|
||||
}
|
||||
var body io.Reader
|
||||
if payload != nil {
|
||||
data, err := json.Marshal(payload)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
body = bytes.NewBuffer(data)
|
||||
}
|
||||
ctx, cancel := context.WithTimeout(context.Background(), timeout)
|
||||
defer cancel()
|
||||
req, err := http.NewRequestWithContext(ctx, method, endpoint, body)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
if payload != nil {
|
||||
req.Header.Set("Content-Type", "application/json")
|
||||
}
|
||||
if async {
|
||||
req.Header.Set("X-DashScope-Async", "enable")
|
||||
}
|
||||
if strings.TrimSpace(cfg.APIKey) != "" {
|
||||
req.Header.Set("Authorization", "Bearer "+strings.TrimSpace(cfg.APIKey))
|
||||
}
|
||||
resp, err := (&http.Client{Timeout: timeout}).Do(req)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
defer resp.Body.Close()
|
||||
respBody, err := io.ReadAll(resp.Body)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
if resp.StatusCode < 200 || resp.StatusCode >= 300 {
|
||||
return fmt.Errorf("HTTP status %d, body=%s", resp.StatusCode, truncateText(string(respBody), 240))
|
||||
}
|
||||
if err := json.Unmarshal(respBody, out); err != nil {
|
||||
return fmt.Errorf("parse response failed: %v, body=%s", err, truncateText(string(respBody), 240))
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
func callOllamaChat(cfg config.AIConfig, systemPrompt string, userPrompt string) (*AIResult, error) {
|
||||
url := strings.TrimRight(cfg.BaseURL, "/")
|
||||
if !strings.HasSuffix(url, "/api/chat") {
|
||||
url += "/api/chat"
|
||||
}
|
||||
payload := map[string]interface{}{
|
||||
"model": cfg.Model,
|
||||
"stream": false,
|
||||
"messages": []map[string]string{
|
||||
{"role": "system", "content": systemPrompt},
|
||||
{"role": "user", "content": userPrompt},
|
||||
},
|
||||
"options": map[string]interface{}{
|
||||
"temperature": cfg.Temperature,
|
||||
"num_predict": effectiveReplyMaxTokens(cfg),
|
||||
},
|
||||
}
|
||||
var response struct {
|
||||
Message struct {
|
||||
Content string `json:"content"`
|
||||
} `json:"message"`
|
||||
Response string `json:"response"`
|
||||
Error string `json:"error"`
|
||||
}
|
||||
result, err := doAIJSONRequest(cfg, url, payload, &response)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
if response.Error != "" {
|
||||
return nil, fmt.Errorf("本地模型返回错误: %s", response.Error)
|
||||
}
|
||||
answer := strings.TrimSpace(response.Message.Content)
|
||||
if answer == "" {
|
||||
answer = strings.TrimSpace(response.Response)
|
||||
}
|
||||
if answer == "" {
|
||||
return nil, fmt.Errorf("本地模型返回空内容")
|
||||
}
|
||||
result.Answer = answer
|
||||
result.RawSummary = truncateText(answer, 160)
|
||||
return result, nil
|
||||
}
|
||||
|
||||
func doAIJSONRequest(cfg config.AIConfig, url string, payload interface{}, out interface{}) (*AIResult, error) {
|
||||
timeout := time.Duration(cfg.TimeoutSeconds) * time.Second
|
||||
if timeout <= 0 {
|
||||
timeout = 20 * time.Second
|
||||
}
|
||||
start := time.Now()
|
||||
ctx, cancel := context.WithTimeout(context.Background(), timeout)
|
||||
defer cancel()
|
||||
body, err := json.Marshal(payload)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(body))
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
req.Header.Set("Content-Type", "application/json")
|
||||
if strings.TrimSpace(cfg.APIKey) != "" {
|
||||
req.Header.Set("Authorization", "Bearer "+strings.TrimSpace(cfg.APIKey))
|
||||
}
|
||||
client := &http.Client{Timeout: timeout}
|
||||
resp, err := client.Do(req)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
defer resp.Body.Close()
|
||||
respBody, err := io.ReadAll(resp.Body)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
if resp.StatusCode < 200 || resp.StatusCode >= 300 {
|
||||
return nil, fmt.Errorf("AI HTTP状态码错误: %d, body=%s", resp.StatusCode, truncateText(string(respBody), 240))
|
||||
}
|
||||
if err := json.Unmarshal(respBody, out); err != nil {
|
||||
return nil, fmt.Errorf("解析AI响应失败: %v, body=%s", err, truncateText(string(respBody), 240))
|
||||
}
|
||||
return &AIResult{DurationMS: time.Since(start).Milliseconds()}, nil
|
||||
}
|
||||
@@ -88,10 +88,43 @@ func (e *AutoReplyEngine) loadEmbeddingIndex() error {
|
||||
if idx.Entries == nil {
|
||||
idx.Entries = make(map[string]EmbeddingEntry)
|
||||
}
|
||||
// 防止"陈旧向量空间"问题：磁盘上的索引若与当前 embedding 模型/维度不一致，
|
||||
// 它处于不同的向量空间，余弦相似度会得到无意义的分数（cosineSimilarity 还会静默截断长度）。
|
||||
// 此时清空向量条目（保留模型/维度元信息），让检索自动回退到关键词，直到用户重建索引。
|
||||
if mismatch := embeddingIndexStaleReason(idx, cfg.Retrieval); mismatch != "" && len(idx.Entries) > 0 {
|
||||
if globalLogger != nil {
|
||||
globalLogger.Warn("[Embedding] 索引与当前配置不一致(%s)，已忽略陈旧向量并回退关键词检索，请重建知识库索引", mismatch)
|
||||
}
|
||||
e.setLastErrorWithScope(autoReplyErrorScopeKnowledge, "Embedding 索引与当前模型/维度不一致("+mismatch+")，已回退关键词检索，请重建知识库索引")
|
||||
idx.Entries = make(map[string]EmbeddingEntry)
|
||||
idx.Model = strings.TrimSpace(cfg.Retrieval.EmbeddingModel)
|
||||
idx.Dimensions = cfg.Retrieval.EmbeddingDimensions
|
||||
}
|
||||
e.updateEmbeddingStatus(&idx)
|
||||
return nil
|
||||
}
|
||||
|
||||
// embeddingIndexStaleReason 返回磁盘索引与当前配置不一致的原因；一致则返回空字符串。
|
||||
// 不区分具体平台（DashScope / 万川 / 其它自建网关），只要模型名或维度对不上就视为陈旧向量空间。
|
||||
func embeddingIndexStaleReason(idx EmbeddingIndex, retrievalCfg config.RetrievalConfig) string {
|
||||
wantModel := strings.TrimSpace(retrievalCfg.EmbeddingModel)
|
||||
gotModel := strings.TrimSpace(idx.Model)
|
||||
if wantModel != "" {
|
||||
if gotModel == "" {
|
||||
// 旧版本生成的索引未记录模型名，无法证明它与当前模型同属一个向量空间，按陈旧处理。
|
||||
return fmt.Sprintf("模型 (未知)→%s", wantModel)
|
||||
}
|
||||
if !strings.EqualFold(wantModel, gotModel) {
|
||||
return fmt.Sprintf("模型 %s→%s", gotModel, wantModel)
|
||||
}
|
||||
}
|
||||
wantDim := retrievalCfg.EmbeddingDimensions
|
||||
if wantDim > 0 && idx.Dimensions > 0 && wantDim != idx.Dimensions {
|
||||
return fmt.Sprintf("维度 %d→%d", idx.Dimensions, wantDim)
|
||||
}
|
||||
return ""
|
||||
}
|
||||
|
||||
func (e *AutoReplyEngine) updateEmbeddingStatus(idx *EmbeddingIndex) {
|
||||
if idx == nil {
|
||||
idx = NewEmbeddingIndex("", 0)
|
||||
@@ -107,9 +140,10 @@ func (e *AutoReplyEngine) updateEmbeddingStatus(idx *EmbeddingIndex) {
|
||||
|
||||
func (e *AutoReplyEngine) rebuildEmbeddingIndex(idx *KnowledgeIndex) error {
|
||||
cfg := e.getConfig()
|
||||
if strings.TrimSpace(cfg.AI.APIKey) == "" || strings.TrimSpace(cfg.AI.BaseURL) == "" {
|
||||
embedCfg := embeddingRequestConfig(cfg.AI, cfg.Retrieval)
|
||||
if strings.TrimSpace(embedCfg.APIKey) == "" || strings.TrimSpace(embedCfg.BaseURL) == "" {
|
||||
e.updateEmbeddingStatus(NewEmbeddingIndex(cfg.Retrieval.EmbeddingModel, cfg.Retrieval.EmbeddingDimensions))
|
||||
return fmt.Errorf("Embedding索引跳过：AI Base URL 或 API Key 未配置")
|
||||
return fmt.Errorf("Embedding索引跳过：Embedding/AI Base URL 或 API Key 未配置")
|
||||
}
|
||||
if idx == nil {
|
||||
return nil
|
||||
@@ -455,8 +489,9 @@ func (e *AutoReplyEngine) searchVectorKnowledge(query string, limit int) ([]Know
|
||||
if idx == nil || embeddingIndex == nil || len(embeddingIndex.Entries) == 0 {
|
||||
return nil, fmt.Errorf("向量索引为空，请先重建知识库索引")
|
||||
}
|
||||
if strings.TrimSpace(cfg.AI.APIKey) == "" || strings.TrimSpace(cfg.AI.BaseURL) == "" {
|
||||
return nil, fmt.Errorf("AI Base URL 或 API Key 未配置")
|
||||
embedCfg := embeddingRequestConfig(cfg.AI, cfg.Retrieval)
|
||||
if strings.TrimSpace(embedCfg.APIKey) == "" || strings.TrimSpace(embedCfg.BaseURL) == "" {
|
||||
return nil, fmt.Errorf("Embedding/AI Base URL 或 API Key 未配置")
|
||||
}
|
||||
vectors, err := callDashScopeEmbeddings(cfg.AI, cfg.Retrieval, []string{query})
|
||||
if err != nil {
|
||||
@@ -492,7 +527,8 @@ func callDashScopeEmbeddings(aiCfg config.AIConfig, retrievalCfg config.Retrieva
|
||||
if len(inputs) == 0 {
|
||||
return nil, nil
|
||||
}
|
||||
url := strings.TrimRight(aiCfg.BaseURL, "/")
|
||||
embedCfg := embeddingRequestConfig(aiCfg, retrievalCfg)
|
||||
url := strings.TrimRight(embedCfg.BaseURL, "/")
|
||||
if !strings.HasSuffix(url, "/embeddings") {
|
||||
url += "/embeddings"
|
||||
}
|
||||
@@ -511,7 +547,7 @@ func callDashScopeEmbeddings(aiCfg config.AIConfig, retrievalCfg config.Retrieva
|
||||
} `json:"data"`
|
||||
Error interface{} `json:"error"`
|
||||
}
|
||||
if err := doRetrievalJSONRequest(aiCfg, url, payload, &response); err != nil {
|
||||
if err := doRetrievalJSONRequest(embedCfg, url, payload, &response); err != nil {
|
||||
// 检测是否是模型配置错误
|
||||
errMsg := err.Error()
|
||||
if strings.Contains(strings.ToLower(errMsg), "unsupported model") &&
|
||||
@@ -562,8 +598,9 @@ func callDashScopeRerank(aiCfg config.AIConfig, retrievalCfg config.RetrievalCon
|
||||
Error interface{} `json:"error"`
|
||||
}
|
||||
var lastErr error
|
||||
for _, url := range dashScopeRerankURLs(aiCfg) {
|
||||
if err := doRetrievalJSONRequest(aiCfg, url, payload, &response); err != nil {
|
||||
rerankCfg := rerankRequestConfig(aiCfg, retrievalCfg)
|
||||
for _, url := range dashScopeRerankURLs(rerankCfg) {
|
||||
if err := doRetrievalJSONRequest(rerankCfg, url, payload, &response); err != nil {
|
||||
lastErr = err
|
||||
continue
|
||||
}
|
||||
@@ -629,6 +666,34 @@ func doRetrievalJSONRequest(aiCfg config.AIConfig, url string, payload interface
|
||||
return nil
|
||||
}
|
||||
|
||||
// embeddingRequestConfig 返回用于向量调用的有效配置：
|
||||
// 优先使用 Retrieval.EmbeddingBaseURL / EmbeddingAPIKey，未配置则回退到聊天网关 AI.BaseURL / APIKey。
|
||||
func embeddingRequestConfig(aiCfg config.AIConfig, retrievalCfg config.RetrievalConfig) config.AIConfig {
|
||||
embedCfg := aiCfg
|
||||
if base := strings.TrimSpace(retrievalCfg.EmbeddingBaseURL); base != "" {
|
||||
embedCfg.BaseURL = base
|
||||
}
|
||||
key := strings.TrimSpace(retrievalCfg.EmbeddingAPIKey)
|
||||
if key != "" && !looksLikeURL(key) {
|
||||
embedCfg.APIKey = key
|
||||
}
|
||||
return embedCfg
|
||||
}
|
||||
|
||||
// rerankRequestConfig 返回用于重排调用的有效配置：
|
||||
// 优先使用 Retrieval.RerankBaseURL / RerankAPIKey，未配置则回退到聊天网关 AI.BaseURL / APIKey。
|
||||
func rerankRequestConfig(aiCfg config.AIConfig, retrievalCfg config.RetrievalConfig) config.AIConfig {
|
||||
rerankCfg := aiCfg
|
||||
if base := strings.TrimSpace(retrievalCfg.RerankBaseURL); base != "" {
|
||||
rerankCfg.BaseURL = base
|
||||
}
|
||||
key := strings.TrimSpace(retrievalCfg.RerankAPIKey)
|
||||
if key != "" && !looksLikeURL(key) {
|
||||
rerankCfg.APIKey = key
|
||||
}
|
||||
return rerankCfg
|
||||
}
|
||||
|
||||
func dashScopeRerankURLs(aiCfg config.AIConfig) []string {
|
||||
baseURL := strings.TrimRight(aiCfg.BaseURL, "/")
|
||||
if strings.Contains(baseURL, "dashscope.aliyuncs.com") {
|
||||
|
||||
@@ -3871,3 +3871,46 @@ func TestLongKnowledgeQueryFindsLateSectionByChinesePhrase(t *testing.T) {
|
||||
t.Fatalf("expected neighbor section to be included, got %s", text)
|
||||
}
|
||||
}
|
||||
|
||||
// 磁盘索引与当前配置不一致（含旧版本无模型名的情况）应判为陈旧，触发清空向量回退关键词。
|
||||
func TestEmbeddingIndexStaleReason(t *testing.T) {
|
||||
cases := []struct {
|
||||
name string
|
||||
idx EmbeddingIndex
|
||||
retrieval config.RetrievalConfig
|
||||
wantStale bool
|
||||
}{
|
||||
{
|
||||
name: "same model and dims",
|
||||
idx: EmbeddingIndex{Model: "text-embedding-v4", Dimensions: 512},
|
||||
retrieval: config.RetrievalConfig{EmbeddingModel: "text-embedding-v4", EmbeddingDimensions: 512},
|
||||
wantStale: false,
|
||||
},
|
||||
{
|
||||
name: "different model",
|
||||
idx: EmbeddingIndex{Model: "text-embedding-v4", Dimensions: 512},
|
||||
retrieval: config.RetrievalConfig{EmbeddingModel: "wanchuan-embed", EmbeddingDimensions: 512},
|
||||
wantStale: true,
|
||||
},
|
||||
{
|
||||
name: "legacy index without model name",
|
||||
idx: EmbeddingIndex{Model: "", Dimensions: 0},
|
||||
retrieval: config.RetrievalConfig{EmbeddingModel: "wanchuan-embed", EmbeddingDimensions: 1024},
|
||||
wantStale: true,
|
||||
},
|
||||
{
|
||||
name: "different dimensions same model",
|
||||
idx: EmbeddingIndex{Model: "text-embedding-v4", Dimensions: 512},
|
||||
retrieval: config.RetrievalConfig{EmbeddingModel: "text-embedding-v4", EmbeddingDimensions: 1024},
|
||||
wantStale: true,
|
||||
},
|
||||
}
|
||||
for _, tc := range cases {
|
||||
t.Run(tc.name, func(t *testing.T) {
|
||||
reason := embeddingIndexStaleReason(tc.idx, tc.retrieval)
|
||||
if (reason != "") != tc.wantStale {
|
||||
t.Fatalf("embeddingIndexStaleReason() = %q, wantStale=%v", reason, tc.wantStale)
|
||||
}
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user