大更新，架构调整，数据分析能力提升，

tests/test_phase3.py

@@ -0,0 +1,233 @@
+# -*- coding: utf-8 -*-
+"""
+Unit tests for Phase 3: Agent Changes
+
+Run: python -m pytest tests/test_phase3.py -v
+"""
+
+import os
+import sys
+import json
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+import pytest
+from data_analysis_agent import DataAnalysisAgent
+from prompts import data_analysis_system_prompt, final_report_system_prompt
+
+
+# ===========================================================================
+# Task 8.1: _summarize_result
+# ===========================================================================
+
+class TestSummarizeResult:
+    @pytest.fixture
+    def agent(self):
+        """Create a minimal DataAnalysisAgent for testing."""
+        agent = DataAnalysisAgent.__new__(DataAnalysisAgent)
+        agent._session_ref = None
+        return agent
+
+    def test_success_with_evidence_rows(self, agent):
+        """8.1: Success with evidence rows produces DataFrame summary."""
+        result = {
+            "success": True,
+            "evidence_rows": [{"a": 1, "b": 2}, {"a": 3, "b": 4}],
+            "auto_exported_files": [{"variable_name": "df", "filename": "df.csv", "rows": 150, "cols": 8, "columns": []}],
+        }
+        summary = agent._summarize_result(result)
+        assert "执行成功" in summary
+        assert "DataFrame" in summary
+        assert "150" in summary
+        assert "8" in summary
+
+    def test_success_with_evidence_no_auto_files(self, agent):
+        """8.1: Success with evidence but no auto_exported_files uses evidence length."""
+        result = {
+            "success": True,
+            "evidence_rows": [{"x": 1}, {"x": 2}, {"x": 3}],
+            "auto_exported_files": [],
+        }
+        summary = agent._summarize_result(result)
+        assert "执行成功" in summary
+        assert "DataFrame" in summary
+
+    def test_success_with_output(self, agent):
+        """8.1: Success with output but no evidence shows first line."""
+        result = {
+            "success": True,
+            "evidence_rows": [],
+            "output": "Hello World\nSecond line",
+        }
+        summary = agent._summarize_result(result)
+        assert "执行成功" in summary
+        assert "Hello World" in summary
+
+    def test_success_no_output(self, agent):
+        """8.1: Success with no output or evidence."""
+        result = {"success": True, "evidence_rows": [], "output": ""}
+        summary = agent._summarize_result(result)
+        assert summary == "执行成功"
+
+    def test_failure_short_error(self, agent):
+        """8.1: Failure with short error message."""
+        result = {"success": False, "error": "KeyError: 'col_x'"}
+        summary = agent._summarize_result(result)
+        assert "执行失败" in summary
+        assert "KeyError" in summary
+
+    def test_failure_long_error_truncated(self, agent):
+        """8.1: Failure with long error is truncated to 100 chars."""
+        long_error = "A" * 200
+        result = {"success": False, "error": long_error}
+        summary = agent._summarize_result(result)
+        assert "执行失败" in summary
+        assert "..." in summary
+        # The error portion should be at most 103 chars (100 + "...")
+        error_part = summary.split("执行失败: ")[1]
+        assert len(error_part) <= 104
+
+    def test_failure_no_error_field(self, agent):
+        """8.1: Failure with missing error field."""
+        result = {"success": False}
+        summary = agent._summarize_result(result)
+        assert "执行失败" in summary
+
+
+# ===========================================================================
+# Task 8.2-8.4: Round_Data construction and session integration
+# ===========================================================================
+
+class TestRoundDataConstruction:
+    def test_handle_generate_code_returns_reasoning(self):
+        """8.2: _handle_generate_code returns reasoning from yaml_data."""
+        agent = DataAnalysisAgent.__new__(DataAnalysisAgent)
+        agent._session_ref = None
+        # We need a minimal executor mock
+        from unittest.mock import MagicMock
+        agent.executor = MagicMock()
+        agent.executor.execute_code.return_value = {
+            "success": True, "output": "ok", "error": "",
+            "variables": {}, "evidence_rows": [],
+            "auto_exported_files": [], "prompt_saved_files": [],
+        }
+        yaml_data = {"code": "x = 1", "reasoning": "Testing reasoning field"}
+        result = agent._handle_generate_code("response text", yaml_data)
+        assert result["reasoning"] == "Testing reasoning field"
+
+    def test_handle_generate_code_empty_reasoning(self):
+        """8.2: _handle_generate_code returns empty reasoning when not in yaml_data."""
+        agent = DataAnalysisAgent.__new__(DataAnalysisAgent)
+        agent._session_ref = None
+        from unittest.mock import MagicMock
+        agent.executor = MagicMock()
+        agent.executor.execute_code.return_value = {
+            "success": True, "output": "", "error": "",
+            "variables": {}, "evidence_rows": [],
+            "auto_exported_files": [], "prompt_saved_files": [],
+        }
+        yaml_data = {"code": "x = 1"}
+        result = agent._handle_generate_code("response text", yaml_data)
+        assert result["reasoning"] == ""
+
+
+# ===========================================================================
+# Task 8.3: set_session_ref
+# ===========================================================================
+
+class TestSetSessionRef:
+    def test_session_ref_default_none(self):
+        """8.3: _session_ref defaults to None."""
+        agent = DataAnalysisAgent()
+        assert agent._session_ref is None
+
+    def test_set_session_ref(self):
+        """8.3: set_session_ref stores the session reference."""
+        agent = DataAnalysisAgent()
+
+        class FakeSession:
+            rounds = []
+            data_files = []
+
+        session = FakeSession()
+        agent.set_session_ref(session)
+        assert agent._session_ref is session
+
+
+# ===========================================================================
+# Task 9.1: Prompt - intermediate data saving instructions
+# ===========================================================================
+
+class TestPromptDataSaving:
+    def test_data_saving_instructions_in_system_prompt(self):
+        """9.1: data_analysis_system_prompt contains DATA_FILE_SAVED instructions."""
+        assert "[DATA_FILE_SAVED]" in data_analysis_system_prompt
+        assert "中间数据保存规则" in data_analysis_system_prompt
+
+    def test_data_saving_example_in_prompt(self):
+        """9.1: Prompt contains example of saving and printing marker."""
+        assert "to_csv" in data_analysis_system_prompt
+        assert "session_output_dir" in data_analysis_system_prompt
+
+
+# ===========================================================================
+# Task 9.2: Prompt - evidence annotation instructions
+# ===========================================================================
+
+class TestPromptEvidenceAnnotation:
+    def test_evidence_annotation_in_report_prompt(self):
+        """9.2: final_report_system_prompt contains evidence annotation instructions."""
+        assert "evidence:round_" in final_report_system_prompt
+        assert "证据标注规则" in final_report_system_prompt
+
+    def test_evidence_annotation_example(self):
+        """9.2: Prompt contains example of evidence annotation."""
+        assert "<!-- evidence:round_3 -->" in final_report_system_prompt
+
+
+# ===========================================================================
+# Task 9.3: _build_final_report_prompt includes evidence
+# ===========================================================================
+
+class TestBuildFinalReportPromptEvidence:
+    def test_evidence_included_when_session_has_rounds(self):
+        """9.3: _build_final_report_prompt includes evidence data when rounds exist."""
+        agent = DataAnalysisAgent.__new__(DataAnalysisAgent)
+        agent.analysis_results = []
+        agent.current_round = 2
+        agent.session_output_dir = "/tmp/test"
+        agent.data_profile = "test profile"
+
+        class FakeSession:
+            rounds = [
+                {
+                    "round": 1,
+                    "reasoning": "分析车型分布",
+                    "result_summary": "执行成功，输出 DataFrame (10行×3列)",
+                    "evidence_rows": [{"车型": "A", "数量": 42}],
+                },
+                {
+                    "round": 2,
+                    "reasoning": "分析模块分布",
+                    "result_summary": "执行成功",
+                    "evidence_rows": [],
+                },
+            ]
+
+        agent._session_ref = FakeSession()
+        prompt = agent._build_final_report_prompt([])
+        assert "各轮次分析证据数据" in prompt
+        assert "第1轮" in prompt
+        assert "第2轮" in prompt
+        assert "车型" in prompt
+
+    def test_no_evidence_when_no_session_ref(self):
+        """9.3: _build_final_report_prompt works without session ref."""
+        agent = DataAnalysisAgent.__new__(DataAnalysisAgent)
+        agent.analysis_results = []
+        agent.current_round = 1
+        agent.session_output_dir = "/tmp/test"
+        agent.data_profile = "test profile"
+        agent._session_ref = None
+        prompt = agent._build_final_report_prompt([])
+        assert "各轮次分析证据数据" not in prompt