图片识别

yuthon_ai_agent/controllers/main.py

@@ -247,25 +247,42 @@ class AiChatController(http.Controller):
         # OCR 提取文字
         ocr_text = ''
         ocr_error = ''
+        ocr_diag = {}  # 诊断信息
         if provider and provider.ocr_enabled:
             from ..models.ocr_provider import get_ocr_instance
             ocr, init_err = get_ocr_instance(provider)
+            ocr_diag['provider'] = provider.name
+            ocr_diag['ocr_enabled'] = True
             if ocr is None:
                 ocr_error = init_err or 'OCR 实例创建失败'
                 _logger.error('[AI-OCR] get_ocr_instance 失败: %s', ocr_error)
+                ocr_diag['init_error'] = ocr_error
             else:
-                result = ocr.smart_recognize(raw_bytes, mimetype)
+                ocr_diag['init_ok'] = True
+                # 直接用通用文字识别（不先走票据识别，避免不必要的 API 调用）
+                _logger.info('[AI-OCR] 开始识别 %s (%d bytes, %s)', filename, len(raw_bytes), mimetype)
+                result = ocr.recognize_general(raw_bytes, mimetype)
+                ocr_diag['method'] = 'GeneralBasicOCR'
+                ocr_diag['success'] = result.get('success')
                 if result.get('success'):
                     ocr_text = result.get('text') or ''
+                    ocr_diag['text_len'] = len(ocr_text)
                     if ocr_text:
                         document.sudo().write({'ai_ocr_content': ocr_text})
-                        _logger.info('[AI-OCR] 识别成功，%d 字符', len(ocr_text))
+                        _logger.info('[AI-OCR] ✅ 识别成功，%d 字符: %s', len(ocr_text), ocr_text[:200])
                     else:
                         ocr_error = '图片中未识别到文字'
-                        _logger.info('[AI-OCR] 识别完成但无文字')
+                        _logger.info('[AI-OCR] ⚠️ 识别完成但无文字 — 图片中可能确实没有印刷文字')
                 else:
                     ocr_error = result.get('error', 'OCR 调用失败')
-                    _logger.error('[AI-OCR] %s', ocr_error)
+                    ocr_diag['error'] = ocr_error
+                    _logger.error('[AI-OCR] ❌ %s', ocr_error)
+        else:
+            ocr_diag['ocr_enabled'] = False
+            if not provider:
+                ocr_diag['reason'] = '未找到 AI 服务商（conversation_id 可能无效）'
+            else:
+                ocr_diag['reason'] = 'OCR 未启用'
 
         return request.make_json_response({
             'success': True,
@@ -274,4 +291,5 @@ class AiChatController(http.Controller):
             'mimetype': mimetype,
             'ocr_text': ocr_text,
             'ocr_error': ocr_error,
+            '_ocr_diag': ocr_diag,  # 前端 console 可查看
         })

yuthon_ai_agent/models/ai_agent.py

@@ -769,6 +769,7 @@ class AiConversation(models.Model):
         ocr_result = ''
         attachment_note = ''
         msg_attachment_ids = []
+        no_text_count = 0  # 无文字附件数量
         if attachment_ids:
             attachments = self.env['documents.document'].sudo().browse(attachment_ids).exists()
             if attachments:
@@ -789,6 +790,7 @@ class AiConversation(models.Model):
                     ocr_result = '\n\n'.join(ocr_parts) + '\n\n'
                 # 有附件但 OCR 无文字 → 告知 AI 它是文本模型，无法“看”图片
                 if no_text_names:
+                    no_text_count = len(no_text_names)
                     file_list = '、'.join(no_text_names)
                     attachment_note = (
                         f'[系统提示：用户上传了附件（{file_list}），'
@@ -802,6 +804,15 @@ class AiConversation(models.Model):
         # 构建最终发送给 AI 的用户消息内容
         full_content = attachment_note + ocr_result + content if (ocr_result or attachment_note) else content
 
+        _logger.info(
+            '[AI-MSG] send_message | attachments=%d | ocr_chars=%d | no_text=%d | '
+            'full_content_head=%s',
+            len(msg_attachment_ids) if msg_attachment_ids else 0,
+            len(ocr_result) if ocr_result else 0,
+            no_text_count,
+            (full_content or '')[:300],
+        )
+
         # 1. 保存用户消息（content 存原始用户文本，ocr_content 存 OCR 结果，
         #    full_content 带 OCR 的版本发给 AI，前端通过 attachments 字段渲染附件预览）
         msg_vals = {

yuthon_ai_agent/static/src/js/ai_chat_action.js

@@ -138,8 +138,12 @@ export class AiChatAction extends Component {
             }
 
             if (data && data.success) {
+                console.log("[AI] 上传成功:", data.filename, "| OCR文字:", (data.ocr_text || '(无)').substring(0, 100));
+                if (data._ocr_diag) {
+                    console.log("[AI] OCR诊断:", JSON.stringify(data._ocr_diag));
+                }
                 if (data.ocr_error) {
-                    console.warn("[AI] OCR:", data.ocr_error);
+                    console.warn("[AI] OCR错误:", data.ocr_error);
                 }
                 this.state.attachments = [
                     ...this.state.attachments,

yuthon_ai_agent/static/src/js/ai_floating.js

@@ -112,8 +112,13 @@ async function uploadFile(file) {
         }
 
         if (data && data.success) {
+            // 诊断日志：查看 OCR 各环节状态
+            console.log("[AI] 上传成功:", data.filename, "| OCR文字:", (data.ocr_text || '(无)').substring(0, 100));
+            if (data._ocr_diag) {
+                console.log("[AI] OCR诊断:", JSON.stringify(data._ocr_diag));
+            }
             if (data.ocr_error) {
-                console.warn("[AI] OCR:", data.ocr_error);
+                console.warn("[AI] OCR错误:", data.ocr_error);
             }
             pendingAttachments.push({
                 attachment_id: data.attachment_id,