feat: Add a user-facing report index with search and filtering #179

README.md

@@ -375,6 +375,33 @@ mode, the summary additionally reports pending candidates; a missing, corrupt,
 or non-terminal validation is pending instead of disappearing from the totals.
 Default-mode summary behavior is unchanged.
 
+#### Interactive Report Index (`fm_agent/report.html`)
+
+A single self-contained `report.html` is generated automatically at the end of
+every full pipeline run (spec-only mode included), aggregating both the bug
+validation results and the per-function logic verification results into one
+browseable page. It is rendered purely from the run artifacts — no LLM calls,
+no network, no extra dependencies — and is byte-deterministic: the same
+artifacts always produce the same page.
+
+Each row shows the report title, a status badge, the source file (linked back
+to the original source), the function name, and a code location. The page
+supports:
+
+- **Search** across title, source file, function name, and code location
+- **Filter** by status (`confirmed` / `not_confirmed` / `error` / `pending`,
+  or `MATCH` / `MISMATCH` / `ERROR` / `SKIPPED`) and by source file
+- **Sort** by status, source file, function name, or code location
+- **Expand / collapse** individual reports or all reports at once, plus a
+  reset button
+
+The page is written to `fm_agent/report.html`; open it in a browser. It can
+also be regenerated on demand from any existing run's artifacts with
+`uv run python report.py <proj_dir>` (or pass an `fm_agent/` workspace or an
+archived-workspace directory directly). The incremental pipeline
+(`--incremental`) does not run the auto-generation hook — regenerate the page
+manually with the same command.
+
 ## Important Notes
 
 1. FM-Agent will create an `fm_agent/` directory under your codebase directory. Make sure there is no name conflict.

README_zh.md

@@ -302,6 +302,19 @@ FM-Agent 会在代码库目录下创建 `fm_agent/` 目录，主要输出内容
 
 `fm_agent/bug_validation/` 目录下的 `summary.json` 文件汇总了所有 Bug 结果，包括报告的 Bug 总数、已确认 Bug 数、未确认 Bug 数。在 `--all-bugs` 模式下，汇总还会报告待验证候选；缺失、损坏或尚未进入终态的验证结果会显示为 `pending`，而不会从统计中消失。默认模式的汇总行为保持不变。
 
+#### 交互式报告索引（`fm_agent/report.html`）
+
+每次完整流水线运行结束后（包括仅生成规约的 `--only-spec` 模式），会自动生成一个自包含的 `report.html` 索引页，将 Bug 验证结果与逐函数逻辑验证结果汇总到同一页面。该页面完全由运行产物渲染而成——不调用 LLM、不联网、无额外依赖，且为字节级确定：相同的产物始终生成相同的页面。
+
+每一行展示报告标题、状态徽章、源文件（可点击跳转到原始源码）、函数名与代码位置。页面支持：
+
+- **搜索**：按标题、源文件、函数名、代码位置搜索
+- **筛选**：按状态（`confirmed` / `not_confirmed` / `error` / `pending`，或 `MATCH` / `MISMATCH` / `ERROR` / `SKIPPED`）与源文件筛选
+- **排序**：按状态、源文件、函数名或代码位置排序
+- **展开/折叠**：单个报告与全部报告一键展开/折叠，以及重置按钮
+
+页面写入 `fm_agent/report.html`，用浏览器打开即可浏览。也可随时基于任意已有运行的产物重新生成：`uv run python report.py <项目目录>`（也可直接传入 `fm_agent/` 工作目录或归档的工作区目录）。增量模式（`--incremental`）不会自动生成该页面，需手动运行上述命令。
+
 #### 日志文件（`fm_agent/fm_agent.log`）
 
 单一日志文件记录完整的流水线执行过程，包括文件提取进度、推理任务的提交与完成情况、网络错误与重试，以及最终的推理统计摘要。日志级别为 `INFO`，格式为 `%(asctime)s [%(levelname)s] %(message)s`。

main.py

@@ -472,6 +472,16 @@ def run_pipeline(
             confirmed = summary.get("total_confirmed", 0)
             print(f"[Pipeline] Confirmed bugs: {confirmed}")
 
+    # Regenerate the static HTML report index from the completed artifacts.
+    # Runs in every completed mode (including --only-spec, which yields an
+    # empty-state page); a rendering problem must never fail the run.
+    try:
+        from report import generate_report
+        index_path = generate_report(work_dir)
+        print(f"[Pipeline] Report index written to {os.path.relpath(index_path, proj_dir)}")
+    except Exception as exc:
+        logging.warning("[Pipeline] report.html generation skipped: %s", exc)
+
     if only_spec:
         print("[Pipeline] Done (specs only; reasoning & bug validation skipped).")
     else: