adai/MemoryBear
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(multimodal_service): add '文档内容：' prefix to document text and simplify image placeholder text

Browse Source
This commit is contained in:
Timebomb2018
2026-04-27 12:25:27 +08:00
parent 6aef8227b1
commit a268d0f7f1
1 changed files with 4 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
api/app/services/multimodal_service.py
View File

@@ -95,7 +95,7 @@ class DashScopeFormatStrategy(MultimodalFormatStrategy):
"""通义千问文档格式""" """通义千问文档格式"""
return True, { return True, {
"type": "text", "type": "text",
"text": f"<document name=\"{file_name}\">\n{text}\n</document>" "text": f"<document name=\"{file_name}\">\n文档内容：\n{text}\n</document>"
} }
async def format_audio( async def format_audio(
@@ -167,6 +167,7 @@ class BedrockFormatStrategy(MultimodalFormatStrategy):
async def format_document(self, file_name: str, text: str) -> tuple[bool, Dict[str, Any]]: async def format_document(self, file_name: str, text: str) -> tuple[bool, Dict[str, Any]]:
"""Bedrock/Anthropic 文档格式（需要 base64 编码）""" """Bedrock/Anthropic 文档格式（需要 base64 编码）"""
# Bedrock 文档需要 base64 编码 # Bedrock 文档需要 base64 编码
text = f"文档内容：\n{text}\n"
text_bytes = text.encode('utf-8') text_bytes = text.encode('utf-8')
base64_text = base64.b64encode(text_bytes).decode('utf-8') base64_text = base64.b64encode(text_bytes).decode('utf-8')
@@ -223,7 +224,7 @@ class OpenAIFormatStrategy(MultimodalFormatStrategy):
"""OpenAI 文档格式""" """OpenAI 文档格式"""
return True, { return True, {
"type": "text", "type": "text",
"text": f"<document name=\"{file_name}\">\n{text}\n</document>" "text": f"<document name=\"{file_name}\">\n文档内容：\n{text}\n</document>"
} }
async def format_audio( async def format_audio(
@@ -395,7 +396,7 @@ class MultimodalService:
ext = img_info.get("ext", "png") ext = img_info.get("ext", "png")
try: try:
_, img_url = await self._save_doc_image_to_storage(img_info["bytes"], ext, tenant_id, workspace_id) _, img_url = await self._save_doc_image_to_storage(img_info["bytes"], ext, tenant_id, workspace_id)
placeholder = f"第{page}页 第{index + 1}张图片" if page > 0 else f"第{index + 1}张图片" placeholder = f"第{page}页 第{index + 1}张" if page > 0 else f"第{index + 1}张"
# 在文本内容中追加图片位置标记 # 在文本内容中追加图片位置标记
if result and result[-1].get("type") in ("text", "document"): if result and result[-1].get("type") in ("text", "document"):
key = "text" if "text" in result[-1] else list(result[-1].keys())[-1] key = "text" if "text" in result[-1] else list(result[-1].keys())[-1]