zeekrAI/document_analyzer
1
0
Fork 0
Code Issues 2 Pull Requests Actions Packages Projects Releases Wiki Activity

fix: step3 过滤非法 source type + step1 重试质量门控 - Closes #57
CI / test (pull_request) Successful in 11s
Details

Browse Source 
- step3 _normalize_rule: 将 function_unit_description 等非法 source type 标准化为 text
- step1 覆盖反馈重试: 仅纳入实际提升覆盖率的 retry 结果，避免低质量输出稀释 ensemble
- 新增 UT: test_normalize_source_invalid_type

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This commit is contained in:
Peter Zhang
2026-06-02 16:16:47 +08:00
parent 1b8baed542
commit 268520d453
3 changed files with 49 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
skills/ir_generation_skill/step1_semantic_index.py
+27 -10
View File
@@ -880,7 +880,7 @@ def run_ensemble_semantic_index(doc: dict) -> dict:
if v:
print(f" {k}: {len(v)} 个问题")
# Feedback retry: re-run with coverage feedback (up to 2 retries)
# Feedback retry: re-run with coverage feedback (up to 2 retries, quality-gated)
retry_count = 0
while retry_count < 2:
feedback = _build_coverage_feedback(gaps)
@@ -889,6 +889,10 @@ def run_ensemble_semantic_index(doc: dict) -> dict:
retry_count += 1
print(f"\n 覆盖反馈重试 #{retry_count} (feedback长度={len(feedback)}字符)...", flush=True)
try:
# record pre-retry coverage to gate quality
pre_warnings = len(gaps.get("coverage_warnings", []))
pre_missing_rows = len(gaps.get("missing_table_rows", []))
retry_prompt = build_prompt(doc, feedback, all_paths)
print(f" 重试 prompt 长度: {len(retry_prompt)} 字符", flush=True)
retry_result = call_llm(retry_prompt, max_retries=1, temperature=0.3)
@@ -902,15 +906,28 @@ def run_ensemble_semantic_index(doc: dict) -> dict:
if src.get("section"):
retry_sections.add(src["section"])
print(f" 重试新增 sections: {sorted(retry_sections)}", flush=True)
semantic_indices.append(retry_result)
merged = ensemble_merge(semantic_indices)
merged["ensemble_temperatures"] = list(temperatures) + [f"feedback_retry_{retry_count}"]
passed, gaps = _quick_validate(merged, doc, all_paths)
merged["validation_passed"] = passed
merged["validation_gaps"] = {
k: v for k, v in gaps.items() if v
}
print(f" 重试后验证: {'PASS' if passed else 'GAPS FOUND'}", flush=True)
# Quality gate: only include retry if it improves coverage
trial_indices = semantic_indices + [retry_result]
trial_merged = ensemble_merge(trial_indices)
trial_passed, trial_gaps = _quick_validate(trial_merged, doc, all_paths)
trial_warnings = len(trial_gaps.get("coverage_warnings", []))
trial_missing = len(trial_gaps.get("missing_table_rows", []))
if trial_warnings < pre_warnings or trial_missing < pre_missing_rows:
semantic_indices.append(retry_result)
merged = trial_merged
passed, gaps = trial_passed, trial_gaps
merged["ensemble_temperatures"] = list(temperatures) + [f"feedback_retry_{retry_count}"]
merged["validation_passed"] = passed
merged["validation_gaps"] = {
k: v for k, v in gaps.items() if v
}
print(f" 重试后验证 (已采纳): {'PASS' if passed else 'GAPS FOUND'} "
f"(warnings {pre_warnings}→{trial_warnings}, "
f"missing_rows {pre_missing_rows}→{trial_missing})", flush=True)
else:
print(f" 重试结果未提升覆盖率，丢弃 "
f"(warnings {pre_warnings}→{trial_warnings}, "
f"missing_rows {pre_missing_rows}→{trial_missing})", flush=True)
except Exception as e:
print(f" 覆盖反馈重试失败: {e}", flush=True)
import traceback