duxingchen/WQ_GUI
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(report): 支持 Minimax AI 后端 + 统一 AI 配置对话框，修复 figure_counter 返回值断链 Bug

Browse Source
This commit is contained in:
DXC
2026-06-08 14:58:16 +08:00
parent d5dd2ba1da
commit e57fdb4f75
4 changed files with 469 additions and 114 deletions
Download Patch File Download Diff File Expand all files Collapse all files

218
src/postprocessing/report_word.py
View File

@ -63,14 +63,23 @@ class _SimpleProgress:
@dataclass
class ReportGenerationConfig:
"""
报告生成与 Ollama AI 分析的可选配置。
未设置的字段沿用环境变量（OLLAMA_*、ENABLE_AI_ANALYSIS）或生成器默认值。
报告生成与 AI 分析的可选配置。
支持 Ollama 和 Minimax 两种后端，通过 AI_PROVIDER 环境变量切换。
未设置的字段沿用环境变量或生成器默认值。
"""
# 通用
ai_provider: Optional[str] = None # "ollama" | "minimax"，默认 "minimax"
enable_ai_analysis: Optional[bool] = None
# Ollama 专属
ollama_base_url: Optional[str] = None
ollama_vision_model: Optional[str] = None
ollama_text_model: Optional[str] = None
ollama_timeout_s: Optional[int] = None
enable_ai_analysis: Optional[bool] = None
# Minimax 专属
minimax_api_key: Optional[str] = None
minimax_vision_model: Optional[str] = None
minimax_text_model: Optional[str] = None
minimax_timeout_s: Optional[int] = None
class WaterQualityReportGenerator:
@ -105,7 +114,14 @@ class WaterQualityReportGenerator:
self.english_font = 'Times New Roman' # 英文
cfg = ai_config
# Ollama：显式 ai_config 优先，否则环境变量
# AI Provider 选择：默认 "minimax"
self.ai_provider = (
cfg.ai_provider
if cfg and cfg.ai_provider
else os.environ.get("AI_PROVIDER", "minimax").lower()
)
# Ollama 配置
default_url = os.environ.get("OLLAMA_URL", "http://localhost:11434").rstrip("/")
self.ollama_base_url = (
cfg.ollama_base_url.rstrip("/")
@ -127,6 +143,33 @@ class WaterQualityReportGenerator:
if cfg and cfg.ollama_timeout_s is not None
else int(os.environ.get("OLLAMA_TIMEOUT_S", "120"))
)
# Minimax 配置
self.minimax_api_key = (
cfg.minimax_api_key
if cfg and cfg.minimax_api_key
else os.environ.get("MINIMAX_API_KEY", "")
)
self.minimax_base_url = (
os.environ.get("MINIMAX_BASE_URL", "https://api.minimaxi.com/v1/text/chatcompletion_v2").rstrip("/")
)
self.minimax_vision_model = (
cfg.minimax_vision_model
if cfg and cfg.minimax_vision_model
else os.environ.get("MINIMAX_VISION_MODEL", "abab6.5s-chat")
)
self.minimax_text_model = (
cfg.minimax_text_model
if cfg and cfg.minimax_text_model
else os.environ.get("MINIMAX_TEXT_MODEL", "abab6.5s-chat")
)
self.minimax_timeout_s = (
int(cfg.minimax_timeout_s)
if cfg and cfg.minimax_timeout_s is not None
else int(os.environ.get("MINIMAX_TIMEOUT_S", "120"))
)
# 通用配置
if cfg and cfg.enable_ai_analysis is not None:
self.enable_ai_analysis = bool(cfg.enable_ai_analysis)
else:
@ -262,8 +305,10 @@ class WaterQualityReportGenerator:
}
def apply_ai_config(self, ai_config: ReportGenerationConfig) -> None:
"""在已创建的生成器上更新 AI 相关设置（下次 _ollama_chat 生效）。"""
"""在已创建的生成器上更新 AI 相关设置（下次 _ai_chat 生效）。"""
cfg = ai_config
if cfg.ai_provider:
self.ai_provider = cfg.ai_provider.lower()
if cfg.ollama_base_url:
self.ollama_base_url = cfg.ollama_base_url.rstrip("/")
if cfg.ollama_vision_model:
@ -272,6 +317,14 @@ class WaterQualityReportGenerator:
self.ollama_text_model = cfg.ollama_text_model
if cfg.ollama_timeout_s is not None:
self.ollama_timeout_s = int(cfg.ollama_timeout_s)
if cfg.minimax_api_key:
self.minimax_api_key = cfg.minimax_api_key
if cfg.minimax_vision_model:
self.minimax_vision_model = cfg.minimax_vision_model
if cfg.minimax_text_model:
self.minimax_text_model = cfg.minimax_text_model
if cfg.minimax_timeout_s is not None:
self.minimax_timeout_s = int(cfg.minimax_timeout_s)
if cfg.enable_ai_analysis is not None:
self.enable_ai_analysis = bool(cfg.enable_ai_analysis)
@ -337,6 +390,133 @@ class WaterQualityReportGenerator:
except Exception as e:
return f"（Ollama解析失败：{e}）"
def _call_minimax_text(self, system_prompt: str, user_prompt: str) -> str:
"""调用 Minimax 文本模型 /v1/text/chatcompletion_v2。"""
if not self.minimax_api_key:
return "（Minimax API Key 未配置，请设置 MINIMAX_API_KEY 环境变量）"
payload: Dict[str, Any] = {
"model": self.minimax_text_model,
"messages": [
{"role": "system", "content": system_prompt},
{"role": "user", "content": user_prompt},
],
}
data = json.dumps(payload, ensure_ascii=False).encode("utf-8")
req = Request(
url=self.minimax_base_url,
data=data,
headers={
"Authorization": f"Bearer {self.minimax_api_key}",
"Content-Type": "application/json",
},
method="POST",
)
try:
with urlopen(req, timeout=self.minimax_timeout_s) as resp:
raw = resp.read().decode("utf-8", errors="ignore")
obj = json.loads(raw)
return (
obj.get("choices", [{}])[0]
.get("message", {})
.get("content", "")
.strip()
or "（模型未返回内容）"
)
except HTTPError as e:
body = e.read().decode("utf-8", errors="ignore")
print(f"[Minimax HTTP {e.code}] {body}")
return f"（Minimax调用失败 HTTP {e.code}：{e.reason}）"
except (URLError, TimeoutError) as e:
return f"（Minimax调用失败：{e}）"
except Exception as e:
return f"（Minimax解析失败：{e}）"
def _call_minimax_vision(self, system_prompt: str, user_prompt: str, image_path: Path) -> str:
"""调用 Minimax 视觉模型（多模态），图片转为 base64 后通过 image_url 传入。"""
if not self.minimax_api_key:
return "（Minimax API Key 未配置，请设置 MINIMAX_API_KEY 环境变量）"
try:
img_bytes = image_path.read_bytes()
img_b64 = base64.b64encode(img_bytes).decode("utf-8")
except Exception as e:
return f"（读取图片失败：{e}）"
payload: Dict[str, Any] = {
"model": self.minimax_vision_model,
"messages": [
{
"role": "user",
"content": [
{"type": "text", "text": user_prompt},
{
"type": "image_url",
"image_url": {"url": f"data:image/jpeg;base64,{img_b64}"},
},
],
}
],
}
if system_prompt:
payload["messages"].insert(
0,
{"role": "system", "content": system_prompt},
)
data = json.dumps(payload, ensure_ascii=False).encode("utf-8")
req = Request(
url=self.minimax_base_url,
data=data,
headers={
"Authorization": f"Bearer {self.minimax_api_key}",
"Content-Type": "application/json",
},
method="POST",
)
try:
with urlopen(req, timeout=self.minimax_timeout_s) as resp:
raw = resp.read().decode("utf-8", errors="ignore")
obj = json.loads(raw)
return (
obj.get("choices", [{}])[0]
.get("message", {})
.get("content", "")
.strip()
or "（模型未返回内容）"
)
except HTTPError as e:
body = e.read().decode("utf-8", errors="ignore")
print(f"[Minimax Vision HTTP {e.code}] {body}")
return f"（Minimax Vision调用失败 HTTP {e.code}：{e.reason}）"
except (URLError, TimeoutError) as e:
return f"（Minimax Vision调用失败：{e}）"
except Exception as e:
return f"（Minimax Vision解析失败：{e}）"
def _ai_chat(
self,
model: str,
system_prompt: str,
user_prompt: str,
image_path: Optional[Path] = None,
) -> str:
"""
统一 AI 调用入口。根据 self.ai_provider 路由到不同后端实现。
model 参数在 ollama 模式下直接使用；在 minimax 模式下忽略（使用类级别配置的模型）。
"""
if self.ai_provider == "minimax":
if image_path is not None:
return self._call_minimax_vision(system_prompt, user_prompt, image_path)
else:
return self._call_minimax_text(system_prompt, user_prompt)
else:
return self._ollama_chat(model, system_prompt, user_prompt, image_path)
def _get_prompt_for_image(self, image_type: str, param: str, figure_num: int) -> Dict[str, str]:
"""按图片类型返回 system/user 提示词，带防幻觉约束。"""
system = (
@ -545,7 +725,7 @@ class WaterQualityReportGenerator:
return str(cache[cache_key])
prompts = self._get_prompt_for_image(image_type=image_type, param=param, figure_num=figure_num)
text = self._ollama_chat(
text = self._ai_chat(
model=self.ollama_vision_model,
system_prompt=prompts["system"],
user_prompt=prompts["user"],
@ -585,7 +765,7 @@ class WaterQualityReportGenerator:
输出格式：数据特征分析（变异程度、数值范围等）结论与数据质量评估"""
return self._ollama_chat(self.ollama_text_model, system, user, image_path=None)
return self._ai_chat(self.ollama_text_model, system, user, image_path=None)
def generate_report(self,
@ -662,7 +842,7 @@ class WaterQualityReportGenerator:
base_section_num = 5
last_param_section_num = base_section_num + len(parameters) - 1
for section_num, param in enumerate(parameters, base_section_num):
self._add_parameter_section(
figure_counter = self._add_parameter_section(
doc,
param,
vis_dir,
@ -671,7 +851,6 @@ class WaterQualityReportGenerator:
all_image_analyses,
progress=progress,
)
figure_counter += len(self.parameter_images.get(param, []))
if section_num != last_param_section_num:
doc.add_page_break()
@ -700,7 +879,7 @@ class WaterQualityReportGenerator:
"- 不要编造具体数值、地名、日期\n\n"
f"{analyses_text}"
)
summary_text = self._ollama_chat(self.ollama_text_model, system, user, image_path=None)
summary_text = self._ai_chat(self.ollama_text_model, system, user, image_path=None)
para = doc.add_paragraph(summary_text)
para.paragraph_format.first_line_indent = Pt(24)
para.paragraph_format.line_spacing = 1.5
@ -741,7 +920,7 @@ class WaterQualityReportGenerator:
"""为单个参数添加报告章节（带编号和规范中英文图题）"""
if param not in self.parameter_descriptions:
print(f"警告: 参数 {param} 没有预定义的描述")
return
return start_figure_num
# 添加带编号的参数标题
heading = doc.add_heading(f"{param_index}. {param} 参数分析", level=1)
@ -851,6 +1030,7 @@ class WaterQualityReportGenerator:
pass
doc.add_paragraph() # 章节结束空行
return start_figure_num + len(image_list)
def _add_cover_page(self, doc):
"""添加专业的封面页 - 优化后的布局"""
@ -1188,9 +1368,9 @@ class WaterQualityReportGenerator:
请用专业且简洁的语言描述，控制在150字以内。"""
if glint_img_path and Path(glint_img_path).exists():
return self._ollama_chat(self.ollama_vision_model, "你是一个专业的水质遥感分析专家。", analysis_prompt, Path(glint_img_path))
return self._ai_chat(self.ollama_vision_model, "你是一个专业的水质遥感分析专家。", analysis_prompt, Path(glint_img_path))
elif original_img_path and Path(original_img_path).exists():
return self._ollama_chat(self.ollama_vision_model, "你是一个专业的水质遥感分析专家。", analysis_prompt, Path(original_img_path))
return self._ai_chat(self.ollama_vision_model, "你是一个专业的水质遥感分析专家。", analysis_prompt, Path(original_img_path))
else:
return "基于影像分析，耀斑主要分布在水体表面强反射区域，对水质参数反演有一定影响，建议在数据处理时重点关注这些区域。"
@ -1231,7 +1411,7 @@ class WaterQualityReportGenerator:
...
各架次轨迹分布合理，覆盖了目标水体区域。"""
result = self._ollama_chat(
result = self._ai_chat(
self.ollama_vision_model,
"你是一位专业的航空摄影测量和遥感专家，擅长分析航线规划图。",
analysis_prompt,
@ -1283,7 +1463,7 @@ class WaterQualityReportGenerator:
【示例输出】
水体面积25.60 km² ，占比: 42.3% ，形态: 扇形分叉。入库方向：西北角和东北角各有狭窄水道汇入，为主要入库河流。出水/大坝方向：南侧水体最窄处。流向推断：水体从西北和东北两个方向汇入，向南侧大坝方向流动。补充描述：水库整体呈扇形，库区宽阔，有两个明显入库分支，符合山区水库典型特征"""
result = self._ollama_chat(
result = self._ai_chat(
self.ollama_vision_model,
"你是一位专业的水体遥感分析专家，擅长解读水体掩膜图和水域分布特征。",
analysis_prompt,
@ -1337,7 +1517,7 @@ class WaterQualityReportGenerator:
请根据图像内容给出专业分析。"""
result = self._ollama_chat(
result = self._ai_chat(
self.ollama_vision_model,
"你是一位专业的水质采样设计专家，擅长评估采样点布局的合理性和代表性。",
analysis_prompt,
@ -1456,13 +1636,13 @@ class WaterQualityReportGenerator:
if not processed_data_dir.exists():
doc.add_paragraph(f"未找到数据处理目录: {processed_data_dir}")
doc.add_page_break()
return
return start_figure_num
csv_files = list(processed_data_dir.glob("*.csv"))
if not csv_files:
doc.add_paragraph(f"在 {processed_data_dir} 目录下未找到CSV统计数据文件。")
doc.add_page_break()
return
return start_figure_num
csv_path = csv_files[0] # 使用找到的第一个CSV文件