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

189
src/gui/dialogs.py
View File

@ -6,6 +6,7 @@
与 water_quality_gui.py 保持 1:1 风格（中文注释 / 顶部 encoding 声明）。
"""
import os
from PyQt5.QtCore import Qt, QTimer
from PyQt5.QtGui import QFont
from PyQt5.QtWidgets import (
@ -17,6 +18,8 @@ from PyQt5.QtWidgets import (
QHBoxLayout,
QDialogButtonBox,
QWidget,
QComboBox,
QLineEdit,
)
@ -145,3 +148,189 @@ class BandConfirmDialog(QDialog):
"""点"取消运行"触发：停表 + 关闭，调用方需中止流程"""
self._timer.stop()
super().reject()
# ─────────────────────────────────────────────────────────────────────────────
# AI 引擎设置对话框
# ─────────────────────────────────────────────────────────────────────────────
from PyQt5.QtCore import QSettings
AI_SETTINGS_ORG = "IrisWaterQuality"
AI_SETTINGS_APP = "WQ_GUI"
AI_DEFAULTS = {
"ollama": {
"api_base_url": "http://localhost:11434",
"vision_model": "qwen3-vl:8b",
"text_model": "qwen3-vl:8b",
},
"minimax": {
"api_base_url": "https://api.minimaxi.com/v1/text/chatcompletion_v2",
"vision_model": "abab6.5s-chat",
"text_model": "abab6.5s-chat",
},
}
class AISettingsDialog(QDialog):
"""AI 引擎可视化配置弹窗，配置持久化到 QSettings。"""
def __init__(self, parent=None):
super().__init__(parent)
self.setWindowTitle("AI 引擎配置")
self.setModal(True)
self.setMinimumWidth(520)
self._load_settings()
self._init_ui()
def _load_settings(self):
"""从 QSettings 读取已有配置；无记录则回退到环境变量或默认值。"""
s = QSettings(AI_SETTINGS_ORG, AI_SETTINGS_APP)
self._provider = s.value("ai_provider", "minimax", type=str)
# API Key 不设默认值（敏感信息，首次必须由用户输入）
self._api_key = s.value("minimax_api_key", "", type=str)
# 已保存的 URL 和模型；若 QSettings 无记录则读环境变量
if self._provider == "ollama":
self._api_base_url = (
s.value("api_base_url", "")
or os.environ.get("OLLAMA_URL", AI_DEFAULTS["ollama"]["api_base_url"])
)
self._vision_model = (
s.value("vision_model", "")
or os.environ.get("OLLAMA_VISION_MODEL", AI_DEFAULTS["ollama"]["vision_model"])
)
self._text_model = (
s.value("text_model", "")
or os.environ.get("OLLAMA_TEXT_MODEL", AI_DEFAULTS["ollama"]["text_model"])
)
else:
self._api_base_url = (
s.value("api_base_url", "")
or os.environ.get("MINIMAX_BASE_URL", AI_DEFAULTS["minimax"]["api_base_url"])
)
self._vision_model = (
s.value("vision_model", "")
or os.environ.get("MINIMAX_VISION_MODEL", AI_DEFAULTS["minimax"]["vision_model"])
)
self._text_model = (
s.value("text_model", "")
or os.environ.get("MINIMAX_TEXT_MODEL", AI_DEFAULTS["minimax"]["text_model"])
)
self._timeout = s.value("timeout_s", 120, type=int)
def _init_ui(self):
layout = QVBoxLayout(self)
layout.setSpacing(12)
# ── Provider ──────────────────────────────────────────────────────────
provider_row = QHBoxLayout()
provider_row.addWidget(QLabel("AI 引擎提供商:"))
self._provider_combo = QComboBox()
self._provider_combo.addItems(["Ollama", "Minimax"])
self._provider_combo.setCurrentText("Ollama" if self._provider == "ollama" else "Minimax")
self._provider_combo.currentIndexChanged.connect(self._on_provider_changed)
provider_row.addWidget(self._provider_combo, 1)
provider_row.addStretch(1)
layout.addLayout(provider_row)
# ── API Base URL ───────────────────────────────────────────────────────
url_row = QHBoxLayout()
url_row.addWidget(QLabel("API Base URL:"))
self._url_edit = QLineEdit(self._api_base_url)
self._url_edit.setPlaceholderText("例如: http://localhost:11434")
url_row.addWidget(self._url_edit, 1)
layout.addLayout(url_row)
# ── API Key ───────────────────────────────────────────────────────────
key_row = QHBoxLayout()
key_row.addWidget(QLabel("API Key:"))
self._key_edit = QLineEdit(self._api_key)
self._key_edit.setPlaceholderText("输入 API Key（敏感信息，已加密存储）")
self._key_edit.setEchoMode(QLineEdit.Password)
key_row.addWidget(self._key_edit, 1)
layout.addLayout(key_row)
# ── 模型名称 ───────────────────────────────────────────────────────────
model_row = QHBoxLayout()
model_row.addWidget(QLabel("视觉模型:"))
self._vision_edit = QLineEdit(self._vision_model)
model_row.addWidget(self._vision_edit, 1)
model_row.addSpacing(12)
model_row.addWidget(QLabel("文本模型:"))
self._text_edit = QLineEdit(self._text_model)
model_row.addWidget(self._text_edit, 1)
layout.addLayout(model_row)
# ── 超时 ──────────────────────────────────────────────────────────────
timeout_row = QHBoxLayout()
timeout_row.addWidget(QLabel("请求超时(秒):"))
self._timeout_spin = QSpinBox()
self._timeout_spin.setRange(30, 3600)
self._timeout_spin.setSingleStep(30)
self._timeout_spin.setValue(self._timeout)
timeout_row.addWidget(self._timeout_spin)
timeout_row.addStretch(1)
layout.addLayout(timeout_row)
# ── 说明 ──────────────────────────────────────────────────────────────
hint = QLabel(
"提示：切换引擎后将自动填充推荐默认值（可手动修改）。"
"API Key 仅本地加密存储，不会明文暴露。"
)
hint.setStyleSheet("color: #888; font-size: 10px;")
hint.setWordWrap(True)
layout.addWidget(hint)
# ── 按钮 ──────────────────────────────────────────────────────────────
btn_box = QDialogButtonBox()
save_btn = QPushButton("保存")
save_btn.setDefault(True)
save_btn.clicked.connect(self._save_and_close)
cancel_btn = QPushButton("取消")
cancel_btn.clicked.connect(self.reject)
btn_box.addButton(save_btn, QDialogButtonBox.AcceptRole)
btn_box.addButton(cancel_btn, QDialogButtonBox.RejectRole)
layout.addWidget(btn_box)
def _on_provider_changed(self):
"""切换 Provider 时自动填充推荐默认值。"""
provider = self._provider_combo.currentText().lower()
defaults = AI_DEFAULTS.get(provider, AI_DEFAULTS["minimax"])
self._url_edit.setText(defaults["api_base_url"])
self._vision_edit.setText(defaults["vision_model"])
self._text_edit.setText(defaults["text_model"])
def _save_and_close(self):
"""持久化到 QSettings 并关闭。"""
s = QSettings(AI_SETTINGS_ORG, AI_SETTINGS_APP)
provider = self._provider_combo.currentText().lower()
s.setValue("ai_provider", provider)
s.setValue("api_base_url", self._url_edit.text().strip())
s.setValue("api_key", self._key_edit.text().strip())
s.setValue("vision_model", self._vision_edit.text().strip())
s.setValue("text_model", self._text_edit.text().strip())
s.setValue("timeout_s", self._timeout_spin.value())
s.sync()
self.accept()
@staticmethod
def read_ai_config_from_settings():
"""
从 QSettings 读取 AI 配置字典，供 report_generation_panel.py 等处使用。
返回键：ai_provider / api_base_url / api_key / vision_model / text_model / timeout_s
"""
s = QSettings(AI_SETTINGS_ORG, AI_SETTINGS_APP)
provider = s.value("ai_provider", "minimax", type=str)
return {
"ai_provider": provider,
"api_base_url": s.value("api_base_url", "", type=str),
"api_key": s.value("api_key", "", type=str),
"vision_model": s.value("vision_model", "", type=str),
"text_model": s.value("text_model", "", type=str),
"timeout_s": s.value("timeout_s", 120, type=int),
}

162
src/gui/panels/report_generation_panel.py
View File

@ -9,14 +9,15 @@ import traceback
from pathlib import Path
from typing import Optional
from PyQt5.QtCore import Qt, QThread, pyqtSignal
from PyQt5.QtCore import Qt, QThread, pyqtSignal, QSettings
from PyQt5.QtWidgets import (
QWidget, QVBoxLayout, QHBoxLayout, QGroupBox, QFormLayout,
QLabel, QCheckBox, QPushButton, QLineEdit, QSpinBox,
QLabel, QCheckBox, QPushButton, QLineEdit,
QMessageBox, QFileDialog,
)
from src.gui.styles import ModernStylesheet
from src.gui.dialogs import AISettingsDialog, AI_SETTINGS_ORG, AI_SETTINGS_APP
class ReportGenerateThread(QThread):
@ -25,35 +26,43 @@ class ReportGenerateThread(QThread):
failed = pyqtSignal(str)
log_message = pyqtSignal(str, str)
def __init__(self, work_dir: str, output_dir: Optional[str], report_title: str, options: dict):
def __init__(self, work_dir: str, output_dir: Optional[str], report_title: str, enable_ai: bool):
super().__init__()
self.work_dir = work_dir
self.output_dir = output_dir
self.report_title = report_title
self.options = options
self.enable_ai = enable_ai
def run(self):
try:
from src.postprocessing.report_word import WaterQualityReportGenerator, ReportGenerationConfig
url = (self.options.get("ollama_url") or "").strip() or None
vision = (self.options.get("ollama_vision_model") or "").strip() or None
text = (self.options.get("ollama_text_model") or "").strip() or None
if self.options.get("text_same_as_vision"):
text = vision
timeout = self.options.get("ollama_timeout_s")
enable_ai = self.options.get("enable_ai_analysis")
# 唯一数据源：直接从 QSettings 读取 AI 配置
s = QSettings(AI_SETTINGS_ORG, AI_SETTINGS_APP)
provider = s.value("ai_provider", "minimax", type=str)
timeout = int(s.value("timeout_s", 120, type=int))
if provider == "ollama":
ai_cfg = ReportGenerationConfig(
ai_provider="ollama",
ollama_base_url=s.value("api_base_url", "", type=str) or None,
ollama_vision_model=s.value("vision_model", "", type=str) or None,
ollama_text_model=s.value("text_model", "", type=str) or None,
ollama_timeout_s=timeout,
enable_ai_analysis=self.enable_ai,
)
else:
ai_cfg = ReportGenerationConfig(
ai_provider="minimax",
minimax_api_key=s.value("api_key", "", type=str) or "",
minimax_vision_model=s.value("vision_model", "", type=str) or None,
minimax_text_model=s.value("text_model", "", type=str) or None,
minimax_timeout_s=timeout,
enable_ai_analysis=self.enable_ai,
)
ai_cfg = ReportGenerationConfig(
ollama_base_url=url,
ollama_vision_model=vision,
ollama_text_model=text,
ollama_timeout_s=int(timeout) if timeout is not None else None,
enable_ai_analysis=bool(enable_ai),
)
self.log_message.emit(
f"报告生成：工作目录={self.work_dir}，AI={'开' if enable_ai else '关'}，"
f"模型URL={url or '(环境变量 OLLAMA_URL)'}",
f"报告生成：工作目录={self.work_dir}，AI={'开' if self.enable_ai else '关'}，Provider={provider}",
"info",
)
gen = WaterQualityReportGenerator(
@ -71,12 +80,13 @@ class ReportGenerateThread(QThread):
class ReportGenerationPanel(QWidget):
"""Word 报告生成：工作目录、输出目录、Ollama URL/模型、是否启用 AI 等。"""
"""Word 报告生成面板。AI 配置统一由 AISettingsDialog 管理，本面板不持有配置状态。"""
def __init__(self, main_window=None, parent=None):
super().__init__(parent)
self.main_window = main_window
self._report_thread = None
self._ai_label = None
self.init_ui()
def init_ui(self):
@ -86,7 +96,7 @@ class ReportGenerationPanel(QWidget):
intro = QLabel(
"根据工作目录下的可视化结果（14_visualization 等）生成 Word 分析报告。"
"需已存在可视化图表；AI 分析通过 Ollama /api/chat 调用本地或远程服务。"
"需已存在可视化图表；AI 分析通过 Ollama 或 Minimax 调用云端/本地服务。"
)
intro.setWordWrap(True)
intro.setStyleSheet(
@ -94,6 +104,7 @@ class ReportGenerationPanel(QWidget):
)
layout.addWidget(intro)
# ── 路径 ──────────────────────────────────────────────────────────────
path_group = QGroupBox("路径")
path_form = QFormLayout()
@ -125,52 +136,32 @@ class ReportGenerationPanel(QWidget):
path_group.setLayout(path_form)
layout.addWidget(path_group)
ai_group = QGroupBox("AI 分析（Ollama）")
ai_form = QFormLayout()
# ── AI 分析 ───────────────────────────────────────────────────────────
ai_group = QGroupBox("AI 分析")
ai_layout = QVBoxLayout()
self.enable_ai_cb = QCheckBox("启用 AI 图表解读与综合总结")
self.enable_ai_cb.setChecked(
os.environ.get("ENABLE_AI_ANALYSIS", "1") not in {"0", "false", "False"}
)
ai_form.addRow(self.enable_ai_cb)
ai_layout.addWidget(self.enable_ai_cb)
self.ollama_url_edit = QLineEdit()
self.ollama_url_edit.setText(
os.environ.get("OLLAMA_URL", "http://localhost:11434").rstrip("/")
)
ai_form.addRow("服务 URL:", self.ollama_url_edit)
# 只读提示行：当前引擎 + 配置按钮
ai_status_row = QHBoxLayout()
ai_status_row.addWidget(QLabel("当前 AI 引擎:"))
self._ai_label = QLabel()
self._ai_label.setStyleSheet("color: #0078d4; font-weight: bold;")
ai_status_row.addWidget(self._ai_label)
ai_status_row.addStretch(1)
open_settings_btn = QPushButton("高级配置...")
open_settings_btn.clicked.connect(self._open_ai_settings)
ai_status_row.addWidget(open_settings_btn)
ai_layout.addLayout(ai_status_row)
self.vision_model_edit = QLineEdit()
self.vision_model_edit.setText(
os.environ.get("OLLAMA_VISION_MODEL", "qwen3-vl:8b")
)
ai_form.addRow("视觉模型:", self.vision_model_edit)
self.same_text_model_cb = QCheckBox("文本总结与视觉使用同一模型")
self.same_text_model_cb.setChecked(True)
ai_form.addRow(self.same_text_model_cb)
self.text_model_edit = QLineEdit()
self.text_model_edit.setText(
os.environ.get(
"OLLAMA_TEXT_MODEL",
self.vision_model_edit.text() or "qwen3-vl:8b"
)
)
self.text_model_edit.setEnabled(False)
self.same_text_model_cb.toggled.connect(self._on_same_text_toggled)
self.vision_model_edit.textChanged.connect(self._sync_text_model_if_linked)
ai_form.addRow("文本模型:", self.text_model_edit)
self.timeout_spin = QSpinBox()
self.timeout_spin.setRange(30, 3600)
self.timeout_spin.setSingleStep(30)
self.timeout_spin.setValue(int(os.environ.get("OLLAMA_TIMEOUT_S", "120")))
ai_form.addRow("请求超时(秒):", self.timeout_spin)
ai_group.setLayout(ai_form)
ai_group.setLayout(ai_layout)
layout.addWidget(ai_group)
# ── 按钮 ──────────────────────────────────────────────────────────────
btn_row = QHBoxLayout()
self.generate_btn = QPushButton("生成 Word 报告")
self.generate_btn.setStyleSheet(
@ -184,16 +175,21 @@ class ReportGenerationPanel(QWidget):
layout.addStretch()
self.setLayout(layout)
def _on_same_text_toggled(self, checked: bool):
self.text_model_edit.setEnabled(not checked)
if checked:
self.text_model_edit.setText(self.vision_model_edit.text())
# 刷新引擎提示文字
self._refresh_ai_label()
def _sync_text_model_if_linked(self, _t=None):
if self.same_text_model_cb.isChecked():
self.text_model_edit.blockSignals(True)
self.text_model_edit.setText(self.vision_model_edit.text())
self.text_model_edit.blockSignals(False)
def _refresh_ai_label(self):
"""从 QSettings 读取当前 Provider 并更新只读标签。"""
s = QSettings(AI_SETTINGS_ORG, AI_SETTINGS_APP)
provider = s.value("ai_provider", "minimax", type=str)
label_map = {"ollama": "Ollama (本地)", "minimax": "Minimax (云端)"}
self._ai_label.setText(label_map.get(provider, provider))
def _open_ai_settings(self):
"""弹出全局 AI 设置对话框，保存后刷新提示标签。"""
dlg = AISettingsDialog(self)
if dlg.exec_() == dlg.Accepted:
self._refresh_ai_label()
def _get_default_work_dir(self):
"""获取 work_dir，优先用主窗口缓存的 work_dir"""
@ -227,15 +223,11 @@ class ReportGenerationPanel(QWidget):
self.work_dir_edit.setText(str(work_dir))
def get_config(self):
"""返回路径和标题配置（AI 配置不由本面板持有）。"""
return {
"work_dir": self.work_dir_edit.text().strip() or None,
"output_dir": self.output_dir_edit.text().strip() or None,
"report_title": self.report_title_edit.text().strip() or "水质参数反演分析报告",
"ollama_url": self.ollama_url_edit.text().strip(),
"ollama_vision_model": self.vision_model_edit.text().strip(),
"ollama_text_model": self.text_model_edit.text().strip(),
"text_same_as_vision": self.same_text_model_cb.isChecked(),
"ollama_timeout_s": self.timeout_spin.value(),
"enable_ai_analysis": self.enable_ai_cb.isChecked(),
}
@ -248,16 +240,6 @@ class ReportGenerationPanel(QWidget):
self.output_dir_edit.setText(str(config["output_dir"] or ""))
if config.get("report_title"):
self.report_title_edit.setText(str(config["report_title"]))
if config.get("ollama_url"):
self.ollama_url_edit.setText(str(config["ollama_url"]))
if config.get("ollama_vision_model"):
self.vision_model_edit.setText(str(config["ollama_vision_model"]))
if "text_same_as_vision" in config:
self.same_text_model_cb.setChecked(bool(config["text_same_as_vision"]))
if config.get("ollama_text_model"):
self.text_model_edit.setText(str(config["ollama_text_model"]))
if config.get("ollama_timeout_s") is not None:
self.timeout_spin.setValue(int(config["ollama_timeout_s"]))
if "enable_ai_analysis" in config:
self.enable_ai_cb.setChecked(bool(config["enable_ai_analysis"]))
@ -280,16 +262,10 @@ class ReportGenerationPanel(QWidget):
out = self.output_dir_edit.text().strip() or None
title = self.report_title_edit.text().strip() or "水质参数反演分析报告"
opts = {
"ollama_url": self.ollama_url_edit.text().strip(),
"ollama_vision_model": self.vision_model_edit.text().strip(),
"ollama_text_model": self.text_model_edit.text().strip(),
"text_same_as_vision": self.same_text_model_cb.isChecked(),
"ollama_timeout_s": self.timeout_spin.value(),
"enable_ai_analysis": self.enable_ai_cb.isChecked(),
}
enable_ai = self.enable_ai_cb.isChecked()
self.generate_btn.setEnabled(False)
self._report_thread = ReportGenerateThread(wd, out, title, opts)
self._report_thread = ReportGenerateThread(wd, out, title, enable_ai)
self._report_thread.log_message.connect(self._forward_log, Qt.QueuedConnection)
self._report_thread.finished_ok.connect(self._on_report_ok, Qt.QueuedConnection)
self._report_thread.failed.connect(self._on_report_fail, Qt.QueuedConnection)
@ -312,4 +288,4 @@ class ReportGenerationPanel(QWidget):
def _on_report_fail(self, err: str):
QMessageBox.critical(self, "失败", f"报告生成失败：\n{err[:800]}")
self._forward_log(err, "error")
self._forward_log(err, "error")

14
src/gui/water_quality_gui.py
View File

@ -128,7 +128,7 @@ from src.gui.panels.step8_panel import Step8Panel
from src.gui.panels.step8_5_panel import Step8_5Panel
from src.gui.panels.step8_75_panel import Step8_75Panel
from src.gui.panels.step9_panel import Step9Panel
from src.gui.dialogs import BandConfirmDialog
from src.gui.dialogs import BandConfirmDialog, AISettingsDialog
from src.gui.panels.visualization_panel import VisualizationPanel
from src.gui.panels.report_generation_panel import ReportGenerationPanel
@ -1684,7 +1684,12 @@ class WaterQualityGUI(QMainWindow):
open_dir_action.triggered.connect(self.open_work_directory)
tools_menu.addSeparator()
ai_config_action = tools_menu.addAction("AI 引擎配置...")
ai_config_action.triggered.connect(self._show_ai_settings)
tools_menu.addSeparator()
# 添加自动填充功能
auto_fill_action = tools_menu.addAction("自动填充所有输入路径")
auto_fill_action.triggered.connect(self.auto_populate_all_steps)
@ -2745,6 +2750,11 @@ class WaterQualityGUI(QMainWindow):
"邮箱：hanshanlong@iris-rs.cn\n"
)
def _show_ai_settings(self):
"""弹出 AI 引擎配置对话框。"""
dlg = AISettingsDialog(self)
dlg.exec_()
def _precheck_step3_bands(self) -> bool:
"""步骤 3 波段越界预检（主线程同步执行，避多线程弹窗坑）

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