SQLite 持久连接 — sandbox 不再每次查询开关连接，改为 __init__ 时建连、close() 时释放
Explorer 的 system prompt 明确告知 sandbox 规则 — "每条 SQL 必须包含聚合函数或 LIMIT"，减少 LLM 生成违规 SQL 浪费轮次 LLM 客户端单例 — 所有组件共享一个 openai.OpenAI 实例，不再各建各的 sanitize 顺序修复 — 小样本抑制放在 float round 之前，避免被 round 干扰 quick_detect 从 O(n²) 改为 O(n) — 按列聚合一次，加去重，不再对每行重复算整列统计 历史上下文实际生效 — get_context_for 的结果现在会注入到 Explorer 的初始 prompt 里，多轮分析时 LLM 能看到之前的发现
This commit is contained in:
1
output/__init__.py
Normal file
1
output/__init__.py
Normal file
@@ -0,0 +1 @@
|
||||
"""输出层：报告、图表、整合"""
|
||||
247
output/chart.py
Normal file
247
output/chart.py
Normal file
@@ -0,0 +1,247 @@
|
||||
"""
|
||||
图表生成器 —— 根据探索结果自动生成可视化图表
|
||||
"""
|
||||
import json
|
||||
import os
|
||||
import re
|
||||
from typing import Any
|
||||
|
||||
import matplotlib
|
||||
matplotlib.use("Agg")
|
||||
import matplotlib.pyplot as plt
|
||||
import matplotlib.font_manager as fm
|
||||
|
||||
from core.config import LLM_CONFIG
|
||||
from core.utils import get_llm_client, extract_json_array
|
||||
from layers.explorer import ExplorationStep
|
||||
|
||||
|
||||
def _setup_chinese_font():
|
||||
candidates = [
|
||||
"SimHei", "Microsoft YaHei", "STHeiti", "WenQuanYi Micro Hei",
|
||||
"Noto Sans CJK SC", "PingFang SC", "Source Han Sans CN",
|
||||
]
|
||||
available = {f.name for f in fm.fontManager.ttflist}
|
||||
for font in candidates:
|
||||
if font in available:
|
||||
plt.rcParams["font.sans-serif"] = [font]
|
||||
plt.rcParams["axes.unicode_minus"] = False
|
||||
return font
|
||||
plt.rcParams["axes.unicode_minus"] = False
|
||||
return None
|
||||
|
||||
_setup_chinese_font()
|
||||
|
||||
|
||||
CHART_PLAN_PROMPT = """你是一个数据可视化专家。根据以下分析结果，规划需要生成的图表。
|
||||
|
||||
## 探索结果
|
||||
{exploration_summary}
|
||||
|
||||
## 规划规则
|
||||
1. 每个有意义的查询结果生成 1 张图，最多 5 张
|
||||
2. 图表类型：bar / horizontal_bar / pie / line / stacked_bar
|
||||
3. 跳过数据量太少（<2 行）的结果
|
||||
4. 标题要简洁
|
||||
|
||||
## 输出格式（纯 JSON 数组，不要代码块）
|
||||
[
|
||||
{{
|
||||
"step_index": 0,
|
||||
"chart_type": "bar",
|
||||
"title": "图表标题",
|
||||
"x_column": "分类轴列名",
|
||||
"y_column": "数值轴列名",
|
||||
"y2_column": null,
|
||||
"top_n": 10,
|
||||
"sort_desc": true
|
||||
}}
|
||||
]"""
|
||||
|
||||
|
||||
class ChartGenerator:
|
||||
"""图表生成器"""
|
||||
|
||||
def __init__(self, output_dir: str = "charts"):
|
||||
self.output_dir = output_dir
|
||||
self.client, self.model = get_llm_client(LLM_CONFIG)
|
||||
|
||||
def generate(self, steps: list[ExplorationStep], question: str) -> list[dict]:
|
||||
valid_steps = [(i, s) for i, s in enumerate(steps) if s.success and s.rows and s.row_count >= 2 and s.action != "done"]
|
||||
if not valid_steps:
|
||||
return []
|
||||
|
||||
plans = self._plan_charts(valid_steps, question)
|
||||
if not plans:
|
||||
return []
|
||||
|
||||
self._clean_old_charts()
|
||||
os.makedirs(self.output_dir, exist_ok=True)
|
||||
|
||||
charts = []
|
||||
for i, plan in enumerate(plans):
|
||||
try:
|
||||
path = self._render_chart(plan, steps, i)
|
||||
if path:
|
||||
charts.append({"path": path, "title": plan.get("title", f"图表 {i+1}")})
|
||||
except Exception as e:
|
||||
print(f" ⚠️ 图表生成失败: {e}")
|
||||
return charts
|
||||
|
||||
def _plan_charts(self, valid_steps: list[tuple[int, ExplorationStep]], question: str) -> list[dict]:
|
||||
summary_parts = []
|
||||
for idx, step in valid_steps:
|
||||
summary_parts.append(
|
||||
f"### 步骤 {idx}: {step.purpose}\n列: {step.columns}\n行数: {step.row_count}\n"
|
||||
f"前 5 行: {json.dumps(step.rows[:5], ensure_ascii=False)}"
|
||||
)
|
||||
|
||||
try:
|
||||
response = self.client.chat.completions.create(
|
||||
model=self.model,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是数据可视化专家。只输出纯 JSON 数组，不要 markdown 代码块。"},
|
||||
{"role": "user", "content": CHART_PLAN_PROMPT.format(exploration_summary="\n\n".join(summary_parts))},
|
||||
],
|
||||
temperature=0.1, max_tokens=1024,
|
||||
)
|
||||
plans = extract_json_array(response.choices[0].message.content.strip())
|
||||
return plans if plans else self._fallback_plan(valid_steps)
|
||||
except Exception as e:
|
||||
print(f" ⚠️ 图表规划失败: {e}，使用 fallback")
|
||||
return self._fallback_plan(valid_steps)
|
||||
|
||||
def _fallback_plan(self, valid_steps: list[tuple[int, ExplorationStep]]) -> list[dict]:
|
||||
plans = []
|
||||
for idx, step in valid_steps[:4]:
|
||||
if len(step.columns) < 2 or step.row_count < 2:
|
||||
continue
|
||||
x_col = step.columns[0]
|
||||
y_col = None
|
||||
for col in step.columns[1:]:
|
||||
if isinstance(step.rows[0].get(col), (int, float)):
|
||||
y_col = col
|
||||
break
|
||||
if not y_col:
|
||||
continue
|
||||
|
||||
chart_type = "bar"
|
||||
if any(kw in x_col for kw in ("月", "日期", "时间", "month", "date")):
|
||||
chart_type = "line"
|
||||
elif step.row_count <= 6:
|
||||
chart_type = "pie"
|
||||
elif len(str(step.rows[0].get(x_col, ""))) > 10:
|
||||
chart_type = "horizontal_bar"
|
||||
|
||||
plans.append({
|
||||
"step_index": idx, "chart_type": chart_type,
|
||||
"title": f"各{x_col}的{y_col}",
|
||||
"x_column": x_col, "y_column": y_col,
|
||||
"y2_column": None, "top_n": 10,
|
||||
"sort_desc": chart_type != "line",
|
||||
})
|
||||
return plans
|
||||
|
||||
def _render_chart(self, plan: dict, steps: list[ExplorationStep], chart_idx: int) -> str | None:
|
||||
step_idx = plan.get("step_index", 0)
|
||||
if step_idx >= len(steps):
|
||||
return None
|
||||
step = steps[step_idx]
|
||||
if not step.success or not step.rows:
|
||||
return None
|
||||
|
||||
chart_type = plan.get("chart_type", "bar")
|
||||
title = plan.get("title", f"图表 {chart_idx + 1}")
|
||||
x_col, y_col = plan.get("x_column", ""), plan.get("y_column", "")
|
||||
y2_col = plan.get("y2_column")
|
||||
top_n = plan.get("top_n", 15)
|
||||
sort_desc = plan.get("sort_desc", True)
|
||||
|
||||
rows = step.rows[:top_n] if top_n else step.rows
|
||||
x_vals = [str(r.get(x_col, "")) for r in rows]
|
||||
y_vals = [self._to_number(r.get(y_col, 0)) for r in rows]
|
||||
|
||||
if sort_desc and chart_type not in ("line",):
|
||||
paired = sorted(zip(x_vals, y_vals), key=lambda p: p[1], reverse=True)
|
||||
x_vals, y_vals = zip(*paired) if paired else ([], [])
|
||||
|
||||
if not x_vals or not y_vals:
|
||||
return None
|
||||
|
||||
fig, ax = plt.subplots(figsize=(10, 6))
|
||||
|
||||
if chart_type == "bar":
|
||||
bars = ax.bar(range(len(x_vals)), y_vals, color="#4C78A8")
|
||||
ax.set_xticks(range(len(x_vals)))
|
||||
ax.set_xticklabels(x_vals, rotation=45, ha="right", fontsize=9)
|
||||
self._add_bar_labels(ax, bars)
|
||||
elif chart_type == "horizontal_bar":
|
||||
bars = ax.barh(range(len(x_vals)), y_vals, color="#4C78A8")
|
||||
ax.set_yticks(range(len(x_vals)))
|
||||
ax.set_yticklabels(x_vals, fontsize=9)
|
||||
ax.invert_yaxis()
|
||||
elif chart_type == "pie":
|
||||
filtered = [(x, y) for x, y in zip(x_vals, y_vals) if y > 0]
|
||||
if not filtered:
|
||||
plt.close(fig)
|
||||
return None
|
||||
x_vals, y_vals = zip(*filtered)
|
||||
ax.pie(y_vals, labels=x_vals, autopct="%1.1f%%", startangle=90, textprops={"fontsize": 9})
|
||||
elif chart_type == "line":
|
||||
ax.plot(range(len(x_vals)), y_vals, marker="o", color="#4C78A8", linewidth=2)
|
||||
ax.set_xticks(range(len(x_vals)))
|
||||
ax.set_xticklabels(x_vals, rotation=45, ha="right", fontsize=9)
|
||||
ax.fill_between(range(len(x_vals)), y_vals, alpha=0.1, color="#4C78A8")
|
||||
if y2_col:
|
||||
y2_vals = [self._to_number(r.get(y2_col, 0)) for r in rows]
|
||||
ax2 = ax.twinx()
|
||||
ax2.plot(range(len(x_vals)), y2_vals, marker="s", color="#E45756", linewidth=2, linestyle="--")
|
||||
ax2.set_ylabel(y2_col, fontsize=10, color="#E45756")
|
||||
elif chart_type == "stacked_bar":
|
||||
ax.bar(range(len(x_vals)), y_vals, label=y_col, color="#4C78A8")
|
||||
if y2_col:
|
||||
y2_vals = [self._to_number(r.get(y2_col, 0)) for r in rows]
|
||||
ax.bar(range(len(x_vals)), y2_vals, bottom=y_vals, label=y2_col, color="#E45756")
|
||||
ax.set_xticks(range(len(x_vals)))
|
||||
ax.set_xticklabels(x_vals, rotation=45, ha="right", fontsize=9)
|
||||
ax.legend()
|
||||
|
||||
ax.set_title(title, fontsize=13, fontweight="bold", pad=12)
|
||||
if chart_type not in ("pie",):
|
||||
ax.set_xlabel(x_col, fontsize=10)
|
||||
if chart_type != "horizontal_bar":
|
||||
ax.set_ylabel(y_col, fontsize=10)
|
||||
ax.grid(axis="y", alpha=0.3)
|
||||
|
||||
plt.tight_layout()
|
||||
fname = f"chart_{chart_idx + 1}.png"
|
||||
fpath = os.path.join(self.output_dir, fname)
|
||||
fig.savefig(fpath, dpi=150, bbox_inches="tight")
|
||||
plt.close(fig)
|
||||
return fpath
|
||||
|
||||
def _clean_old_charts(self):
|
||||
if os.path.isdir(self.output_dir):
|
||||
for f in os.listdir(self.output_dir):
|
||||
if f.endswith(".png"):
|
||||
try:
|
||||
os.remove(os.path.join(self.output_dir, f))
|
||||
except OSError:
|
||||
pass
|
||||
|
||||
def _add_bar_labels(self, ax, bars):
|
||||
for bar in bars:
|
||||
h = bar.get_height()
|
||||
if h > 0:
|
||||
label = f"{h:.1f}" if isinstance(h, float) else str(int(h))
|
||||
ax.text(bar.get_x() + bar.get_width() / 2, h, label, ha="center", va="bottom", fontsize=8)
|
||||
|
||||
def _to_number(self, val) -> float:
|
||||
if isinstance(val, (int, float)):
|
||||
return float(val)
|
||||
if isinstance(val, str):
|
||||
try:
|
||||
return float(val.replace("<", "").replace(",", "").strip())
|
||||
except ValueError:
|
||||
return 0.0
|
||||
return 0.0
|
||||
84
output/consolidator.py
Normal file
84
output/consolidator.py
Normal file
@@ -0,0 +1,84 @@
|
||||
"""
|
||||
报告整合器 —— 将多次分析结果合并为一份完整报告
|
||||
"""
|
||||
import json
|
||||
|
||||
from core.config import LLM_CONFIG
|
||||
from core.utils import get_llm_client
|
||||
from layers.context import AnalysisSession
|
||||
|
||||
|
||||
CONSOLIDATE_PROMPT = """你是一个高级数据分析总监。下面是你的团队针对同一份数据做的多次分析，请整合为一份完整的综合报告。
|
||||
|
||||
## 核心问题
|
||||
{question}
|
||||
|
||||
## 各次分析结果
|
||||
{sections}
|
||||
|
||||
## 可用图表
|
||||
{charts_text}
|
||||
|
||||
## 整合要求
|
||||
1. **执行摘要**：3-5 句话概括全局结论
|
||||
2. **核心发现**：从所有分析中提炼最重要的发现，去重，按重要性排列
|
||||
3. **交叉洞察**：不同维度之间的关联
|
||||
4. **图表引用**：用 `` 嵌入相关段落
|
||||
5. **风险与建议**：按优先级排列
|
||||
6. **数据附录**：关键统计数字
|
||||
|
||||
中文，专业简报风格。先结论后细节。"""
|
||||
|
||||
|
||||
class ReportConsolidator:
|
||||
"""报告整合器"""
|
||||
|
||||
def __init__(self):
|
||||
self.client, self.model = get_llm_client(LLM_CONFIG)
|
||||
|
||||
def consolidate(self, sessions: list[AnalysisSession], question: str = "",
|
||||
charts: list[dict] | None = None) -> str:
|
||||
if not sessions:
|
||||
return "（无分析数据可整合）"
|
||||
if not question:
|
||||
question = sessions[0].question
|
||||
|
||||
sections = self._build_sections(sessions)
|
||||
charts_text = "\n".join(f"{i}. {c['title']}: {c['path']}" for i, c in enumerate(charts or [], 1)) or "无图表。"
|
||||
|
||||
try:
|
||||
response = self.client.chat.completions.create(
|
||||
model=self.model,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是高级数据分析总监，整合多维度分析结果。"},
|
||||
{"role": "user", "content": CONSOLIDATE_PROMPT.format(question=question, sections=sections, charts_text=charts_text)},
|
||||
],
|
||||
temperature=0.3, max_tokens=4096,
|
||||
)
|
||||
return response.choices[0].message.content
|
||||
except Exception as e:
|
||||
print(f" ⚠️ LLM 整合失败: {e}，使用拼接模式")
|
||||
return self._fallback_concat(sessions, charts)
|
||||
|
||||
def _build_sections(self, sessions: list[AnalysisSession]) -> str:
|
||||
parts = []
|
||||
for i, session in enumerate(sessions, 1):
|
||||
section = f"### 分析 {i}: {session.question}\n"
|
||||
section += f"类型: {session.plan.get('analysis_type', '未知')}\n\n"
|
||||
for step in session.steps:
|
||||
if not step.success or not step.rows or step.action == "done":
|
||||
continue
|
||||
section += f"- {step.purpose} ({step.row_count} 行)\n"
|
||||
section += f" 数据: {json.dumps(step.rows[:8], ensure_ascii=False)}\n\n"
|
||||
if session.insights:
|
||||
section += "#### 洞察\n" + "\n".join(f"- {i}" for i in session.insights) + "\n"
|
||||
parts.append(section)
|
||||
return "\n---\n".join(parts)
|
||||
|
||||
def _fallback_concat(self, sessions: list[AnalysisSession], charts: list[dict] | None) -> str:
|
||||
parts = ["# 综合分析报告\n"]
|
||||
for i, s in enumerate(sessions, 1):
|
||||
parts.append(f"## 第 {i} 部分: {s.question}\n{s.report}\n")
|
||||
if charts:
|
||||
parts.append("\n## 可视化\n" + "\n".join(f"![{c['title']}]({c['path']})" for c in charts))
|
||||
return "\n".join(parts)
|
||||
84
output/reporter.py
Normal file
84
output/reporter.py
Normal file
@@ -0,0 +1,84 @@
|
||||
"""
|
||||
报告生成器 —— 单次分析报告
|
||||
"""
|
||||
import json
|
||||
from typing import Any
|
||||
|
||||
from core.config import LLM_CONFIG
|
||||
from core.utils import get_llm_client
|
||||
from layers.explorer import ExplorationStep
|
||||
from layers.insights import Insight
|
||||
|
||||
|
||||
REPORT_PROMPT = """你是一个数据分析报告撰写专家。基于以下信息撰写报告。
|
||||
|
||||
## 用户问题
|
||||
{question}
|
||||
|
||||
## 分析计划
|
||||
{plan}
|
||||
|
||||
## 探索过程
|
||||
{exploration}
|
||||
|
||||
## 主动洞察
|
||||
{insights_text}
|
||||
|
||||
## 可用图表
|
||||
{charts_text}
|
||||
|
||||
## 撰写要求
|
||||
1. **开头**：一句话总结核心结论
|
||||
2. **核心发现**：按重要性排列，带具体数字
|
||||
3. **图表引用**：用 `` 嵌入到相关段落
|
||||
4. **深入洞察**：异常、趋势、关联
|
||||
5. **建议**：基于数据的行动建议
|
||||
6. **审计**：末尾附上所有 SQL
|
||||
|
||||
中文，专业简报风格。图表自然嵌入对应段落。"""
|
||||
|
||||
|
||||
class ReportGenerator:
|
||||
"""报告生成器"""
|
||||
|
||||
def __init__(self):
|
||||
self.client, self.model = get_llm_client(LLM_CONFIG)
|
||||
|
||||
def generate(self, question: str, plan: dict, steps: list[ExplorationStep],
|
||||
insights: list[Insight], charts: list[dict] | None = None) -> str:
|
||||
exploration = self._build_exploration(steps)
|
||||
insights_text = "\n".join(str(i) for i in insights) if insights else "未检测到异常。"
|
||||
charts_text = "\n".join(f"{i}. 标题: {c['title']}, 路径: {c['path']}" for i, c in enumerate(charts or [], 1)) or "无图表。"
|
||||
|
||||
prompt = REPORT_PROMPT.format(
|
||||
question=question,
|
||||
plan=json.dumps(plan, ensure_ascii=False, indent=2),
|
||||
exploration=exploration,
|
||||
insights_text=insights_text,
|
||||
charts_text=charts_text,
|
||||
)
|
||||
|
||||
response = self.client.chat.completions.create(
|
||||
model=self.model,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是专业的数据分析师，撰写清晰、有洞察力的分析报告。"},
|
||||
{"role": "user", "content": prompt},
|
||||
],
|
||||
temperature=0.3, max_tokens=4096,
|
||||
)
|
||||
return response.choices[0].message.content
|
||||
|
||||
def _build_exploration(self, steps: list[ExplorationStep]) -> str:
|
||||
parts = []
|
||||
for step in steps:
|
||||
if step.action == "done":
|
||||
parts.append(f"### 结束\n{step.reasoning}")
|
||||
elif step.success:
|
||||
parts.append(
|
||||
f"### 第 {step.round_num} 轮：{step.purpose}\n"
|
||||
f"SQL: `{step.sql}`\n结果 ({step.row_count} 行):\n"
|
||||
f"数据: {json.dumps(step.rows, ensure_ascii=False)}"
|
||||
)
|
||||
else:
|
||||
parts.append(f"### 第 {step.round_num} 轮：{step.purpose}\nSQL: `{step.sql}`\n失败: {step.error}")
|
||||
return "\n\n".join(parts) if parts else "无探索步骤"
|
||||
Reference in New Issue
Block a user