diff --git a/fastdeploy/golang_router/.claude/skills/stat-cache-hitrate/SKILL.md b/fastdeploy/golang_router/.claude/skills/stat-cache-hitrate/SKILL.md index 150a15a4dd7..e07281576a6 100644 --- a/fastdeploy/golang_router/.claude/skills/stat-cache-hitrate/SKILL.md +++ b/fastdeploy/golang_router/.claude/skills/stat-cache-hitrate/SKILL.md @@ -94,7 +94,7 @@ python3 .claude/skills/stat-cache-hitrate/scripts/stat_cache_hitrate.py <日志 - `summary/cache_hitrate_report.md` — Per-Worker 统计 + Fallback 明细 + 详情链接 - `detail/per_window_data.md` — 每5s窗口明细（连续空窗口自动合并为 3 行：起始/合并说明/结束） -- `detail/session_hit_details.md` — 每个 session 的命中明细（Markdown 表格），包含 `session / req_count / first_hit / avg_hit(excl_first) / max_hit / min_hit / all_hits / prefill_urls / switch_req_pairs / sharp_drop_request_ids` +- `detail/session_hit_details.md` — 每个 session（无 session_id 时回退 trace_id）的命中明细（Markdown 表格），包含 `id序号 / req_count / first_hit / avg_hit(excl_first) / max_hit / min_hit / all_hits / prefill_urls`，并附「序号与会话ID映射」「切换 reqid 明细（可跳转）」。 ### 交叉诊断矩阵 diff --git a/fastdeploy/golang_router/.claude/skills/stat-cache-hitrate/scripts/session_analysis.py b/fastdeploy/golang_router/.claude/skills/stat-cache-hitrate/scripts/session_analysis.py index 355ba8fc947..f7b4caed542 100644 --- a/fastdeploy/golang_router/.claude/skills/stat-cache-hitrate/scripts/session_analysis.py +++ b/fastdeploy/golang_router/.claude/skills/stat-cache-hitrate/scripts/session_analysis.py @@ -7,7 +7,7 @@ def compute_session_details(strategies, strip_scheme): - """按 session 统计命中详情。""" + """按 session_id（优先）或 trace_id（兜底）统计命中详情。""" def _req_id_from_tags(tags, fallback): return tags.get("request_id") or tags.get("req_id") or tags.get("trace_id") or fallback @@ -18,12 +18,14 @@ def _req_id_from_tags(tags, fallback): continue tags = rec.get("tags", {}) or {} session_id = tags.get("session_id") - if not session_id: + trace_id = tags.get("trace_id") + identity = session_id or trace_id + if not identity: continue - session_records[session_id].append((idx, rec)) + session_records[identity].append((idx, rec)) rows = [] - for session_id, items in session_records.items(): + for identity, items in session_records.items(): items.sort(key=lambda x: (x[1].get("ts_ms", ""), x[1].get("ts", ""), x[0])) recs = [r for _, r in items] hits = [int(r.get("selected_hitRatio", 0)) for r in recs] @@ -62,7 +64,8 @@ def _req_id_from_tags(tags, fallback): rows.append( { - "session": session_id, + "session": identity, + "id_type": "session_id" if recs[0].get("tags", {}).get("session_id") else "trace_id", "req_count": len(hits), "first_hit": f"{hits[0]}%", "avg_hit(excl_first)": f"{avg_excl_first}%" if avg_excl_first != "-" else "-", diff --git a/fastdeploy/golang_router/.claude/skills/stat-cache-hitrate/scripts/stat_cache_hitrate.py b/fastdeploy/golang_router/.claude/skills/stat-cache-hitrate/scripts/stat_cache_hitrate.py index c338e46030b..b5adcb9bd5f 100644 --- a/fastdeploy/golang_router/.claude/skills/stat-cache-hitrate/scripts/stat_cache_hitrate.py +++ b/fastdeploy/golang_router/.claude/skills/stat-cache-hitrate/scripts/stat_cache_hitrate.py @@ -93,6 +93,10 @@ def _truncate_text(v, limit=72): s = str(v) return s if len(s) <= limit else s[: limit - 1] + "…" + +def _seq_label(n): + return f"S{n:03d}" + # ════════════════════════════════════════════════════════════════ # Phase 1: 日志读取 # ════════════════════════════════════════════════════════════════ @@ -257,7 +261,6 @@ def compute_session_hitrate(stats_recs, inference_count): total_total = sum(r.get("total", 0) for r in stats_recs) session_hr = round(total_hits / total_total * 100, 1) if total_total else 0 - coverage = round(total_total / inference_count * 100, 1) if inference_count else 0 # 趋势：每个窗口的 hits/total trend = time_bucket(stats_recs, "auto", [("hits", "sum"), ("total", "sum")]) @@ -270,7 +273,6 @@ def compute_session_hitrate(stats_recs, inference_count): "rate": session_hr, "hits": total_hits, "total": total_total, - "coverage": coverage, "inference_count": inference_count, "trend": trend, } @@ -453,8 +455,6 @@ def format_full_report(filepath, line_count, prefix_hr, session_hr, per_worker, # 2. Session Hit Rate parts.append("### 2. Session Hit Rate (请求级路由粘性)") parts.append(f' 累计: {session_hr["rate"]}% (hits={session_hr["hits"]} / total={session_hr["total"]})') - parts.append(f' 覆盖率: {session_hr["coverage"]}% 的推理请求带 session_id') - trend_str = _quartile_trend(session_hr["trend"], "value") if trend_str: parts.append(f" 趋势: {trend_str}") @@ -521,10 +521,7 @@ def format_tail_report(filepath, line_count, prefix_hr, session_hr, scheduling): parts.append(f"**File**: {filepath} | **tail {line_count} lines**") parts.append("") parts.append(f' Prefix Hit Ratio: {prefix_hr["mean"]}% (avg) | Cold start: {prefix_hr["cold_start_rate"]}%') - parts.append( - f' Session Hit Rate: {session_hr["rate"]}% (hits={session_hr["hits"]}/total={session_hr["total"]})' - f' | Coverage: {session_hr["coverage"]}%' - ) + parts.append(f' Session Hit Rate: {session_hr["rate"]}% (hits={session_hr["hits"]}/total={session_hr["total"]})') parts.append( f' Strategy: scoring {scheduling["scoring_count"]} ({scheduling["scoring_pct"]}%)' f' | fallback {scheduling["fallback_count"]}' @@ -665,7 +662,6 @@ def save_detailed_report( parts.append("### Session Hit Rate") parts.append(f'- 累计: **{session_hr["rate"]}%** (hits={session_hr["hits"]}/total={session_hr["total"]})') - parts.append(f'- 覆盖率: **{session_hr["coverage"]}%**') trend_str = _quartile_trend(session_hr["trend"], "value") if trend_str: parts.append(f"- 趋势: {trend_str}") @@ -765,14 +761,15 @@ def save_detailed_report( ) session_parts.append("") focus_columns = [ - "session", + "id", "req_count", + "id_type", "sticky", "unique_workers", "avg_hit(excl_first)", + "max_hit", "min_hit", - "switch_req_pairs", - "sharp_drop_request_ids", + "switch_reqids", ] session_parts.append("## 优先排查 Session（Top 20）") prioritized_rows = sorted( @@ -784,46 +781,76 @@ def save_detailed_report( ), )[:20] compact_rows = [] + all_rows_with_seq = [] + for i, r in enumerate(session_rows, start=1): + all_rows_with_seq.append({**r, "id": _seq_label(i)}) + + seq_map = {r["session"]: r["id"] for r in all_rows_with_seq} + for r in prioritized_rows: + sid = seq_map.get(r["session"], "-") compact_rows.append( { - "session": r["session"], + "id": sid, "req_count": r["req_count"], + "id_type": r.get("id_type", "session_id"), "sticky": r["sticky"], "unique_workers": r["unique_workers"], "avg_hit(excl_first)": r["avg_hit(excl_first)"], + "max_hit": r["max_hit"], "min_hit": r["min_hit"], - "switch_req_pairs": _truncate_text(r["switch_req_pairs"]), - "sharp_drop_request_ids": _truncate_text(r["sharp_drop_request_ids"]), + "switch_reqids": f"[查看](#switch-{sid.lower()})" if r["switch_req_pairs"] != "-" else "-", } ) - session_parts.append(_render_markdown_table(compact_rows, focus_columns, align_right={"req_count", "unique_workers"})) + session_parts.append( + _render_markdown_table(compact_rows, focus_columns, align_right={"req_count", "unique_workers"}) + ) session_parts.append("") session_columns = [ - "session", + "id", "req_count", + "id_type", "first_hit", "avg_hit(excl_first)", "max_hit", "min_hit", "all_hits", "prefill_urls", - "switch_req_pairs", - "sharp_drop_request_ids", "sticky", "unique_workers", ] session_parts.append("## 全量明细（Markdown 表格）") session_parts.append( _render_markdown_table( - session_rows, + all_rows_with_seq, session_columns, align_right={"req_count", "unique_workers"}, ) ) session_parts.append("") + session_parts.append("## 序号与会话ID映射") + map_rows = [ + { + "id": r["id"], + "id_type": r.get("id_type", "session_id"), + "session_or_trace_id": r["session"], + } + for r in all_rows_with_seq + ] + session_parts.append(_render_markdown_table(map_rows, ["id", "id_type", "session_or_trace_id"])) + session_parts.append("") + + session_parts.append("## 切换 reqid 明细（可跳转）") + for r in all_rows_with_seq: + session_parts.append(f'### switch-{r["id"].lower()}') + session_parts.append(f'- ID: **{r["id"]}**') + session_parts.append(f'- 会话标识: `{r["session"]}` ({r.get("id_type", "session_id")})') + session_parts.append(f'- switch_req_pairs: {r["switch_req_pairs"]}') + session_parts.append(f'- sharp_drop_request_ids: {r["sharp_drop_request_ids"]}') + session_parts.append("") + session_path = os.path.join(details_dir, "session_hit_details.md") with open(session_path, "w") as f: f.write("\n".join(session_parts))