中期成果提交：PPT自动生成 Agent（感知-规划-行动-记忆闭环 + aura-ppt 渲染引擎 + Web 交互界面 + 测试用例 + AI 使用日志）

tests/bench_models.py

@@ -0,0 +1,80 @@
+# -*- coding: utf-8 -*-
+"""候选免费模型基准测试：大纲任务耗时 + JSON 有效性。"""
+import io
+import json
+import sys
+import time
+import urllib.request
+
+sys.path.insert(0, r"D:\00.张华丹\VS专用\04.PPT自动生成\ppt-agent")
+
+BASE = "http://127.0.0.1:4096"
+SAMPLE = """# 项目背景
+为提升知识周转效率启动平台建设，前期调研两个月
+# 本期成果
+部署完成基础服务，导入文档120份
+混合检索准确率较基线提升明显
+完成三个部门试点接入
+# 下一步
+扩大试点范围，建立运营机制
+"""
+
+PROMPT = """请把下面的内容规划成一份 PPT 的页面大纲。
+要求：
+1. 页面总数 6~9 页（含封面、目录、结尾）
+2. 输出语言：中文
+3. 页面类型只能用：cover / toc / section / content / two_column / end
+4. 第一页 cover，最后一页 end
+5. content 页的 content 数组为要点（每条一句话，4~6 条）
+6. 每页写 notes：2~3 句口语化演讲备注
+7. 标题要有信息量
+输出格式：{"slides": [{"type": "...", "title": "...", "content": ["..."], "notes": "..."}]}
+
+【用户标题】知识平台周报
+【用户内容】
+""" + SAMPLE
+
+
+def post(path, payload, timeout):
+    req = urllib.request.Request(
+        BASE + path, data=json.dumps(payload).encode("utf-8"), method="POST",
+        headers={"Content-Type": "application/json"})
+    with urllib.request.urlopen(req, timeout=timeout) as r:
+        return json.loads(r.read() or b"{}")
+
+
+def run_model(model_id):
+    t0 = time.time()
+    try:
+        sess = post("/session", {"title": "bench"}, 30)
+        sid = sess["id"]
+        provider, _, model = model_id.partition("/")
+        result = post(f"/session/{sid}/message", {
+            "model": {"providerID": provider, "modelID": model},
+            "parts": [{"type": "text", "text": PROMPT}],
+        }, 300)
+        dt = time.time() - t0
+        text = "\n".join(p.get("text", "") for p in result.get("parts", [])
+                         if p.get("type") == "text")
+        s, e = text.find("{"), text.rfind("}")
+        ok = False
+        n = 0
+        if 0 <= s < e:
+            try:
+                data = json.loads(text[s:e + 1])
+                slides = data.get("slides", [])
+                n = len(slides)
+                ok = 4 <= n <= 12 and all("title" in x for x in slides)
+            except json.JSONDecodeError:
+                pass
+        print(f"{model_id:45s} {dt:6.1f}s  pages={n:2d}  valid={ok}")
+        return dt, ok
+    except Exception as exc:  # noqa: BLE001
+        print(f"{model_id:45s} FAIL  {str(exc)[:60]}")
+        return 999, False
+
+
+for m in sys.argv[1:] or ["opencode/nemotron-3.5-lightning-free",
+                          "opencode/mimo-v2.5-free",
+                          "opencode/hy3-free"]:
+    run_model(m)

tests/e2e_chat.py

@@ -0,0 +1,156 @@
+"""对话修改闭环 e2e：生成 → 换色 → 改标题 → 改页内容 → 两版本/下载验证。
+
+用法：python tests/e2e_chat.py
+"""
+
+import json
+import os
+import subprocess
+import sys
+import time
+import urllib.request
+import urllib.error
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT))
+
+BASE = "http://127.0.0.1:5098"
+USER = f"chat_e2e_{int(time.time())}"
+PASS = "chat_e2e_123"
+SAMPLE = "# 项目背景\n为提升内部知识周转效率，启动知识库平台建设\n已完成技术选型与原型验证\n# 本期成果\n部署完成基础服务\n导入文档120份\n"
+
+COOKIE = {"value": ""}
+
+
+def call(method, path, payload=None, raw=False):
+    url = BASE + path
+    data = None if payload is None else json.dumps(payload).encode("utf-8")
+    req = urllib.request.Request(url, data=data, method=method)
+    if payload is not None:
+        req.add_header("Content-Type", "application/json")
+    if COOKIE["value"]:
+        req.add_header("Cookie", COOKIE["value"])
+    try:
+        with urllib.request.urlopen(req, timeout=120) as resp:
+            body = resp.read()
+            sc = resp.headers.get("Set-Cookie")
+            if sc and not COOKIE["value"]:
+                COOKIE["value"] = sc.split(";")[0]
+            return resp.status, (body if raw else json.loads(body or b"{}"))
+    except urllib.error.HTTPError as e:
+        body = e.read()
+        try:
+            parsed = json.loads(body or b"{}")
+        except Exception:
+            parsed = {"raw": body.decode("utf-8", "ignore")}
+        sc = e.headers.get("Set-Cookie")
+        if sc and not COOKIE["value"]:
+            COOKIE["value"] = sc.split(";")[0]
+        return e.code, parsed
+
+
+def main():
+    failures = []
+
+    def check(name, cond, extra=""):
+        print(f"  [{'OK' if cond else 'NG'}] {name} {extra}")
+        if not cond:
+            failures.append(name)
+
+    env = {**os.environ, "PYTHONIOENCODING": "utf-8"}
+    subprocess.run([sys.executable, "-m", "src.db.init"], cwd=str(ROOT),
+                   capture_output=True, env=env)
+    proc = subprocess.Popen(
+        [sys.executable, "-c",
+         f"import sys; sys.path.insert(0, r'{ROOT}');"
+         "from src.web.app import app; app.run(host='127.0.0.1', port=5098)"],
+        cwd=str(ROOT), env=env, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
+    )
+    try:
+        up = False
+        for _ in range(30):
+            time.sleep(0.5)
+            try:
+                call("GET", "/api/me")
+                up = True
+                break
+            except Exception:
+                continue
+        check("server up", up)
+
+        s, b = call("POST", "/api/register",
+                    {"username": USER, "password": PASS, "confirm_password": PASS})
+        check("register", s == 200 and b.get("code") == 0)
+
+        s, b = call("POST", "/api/generate", {
+            "title": "对话修改测试", "content": SAMPLE,
+            "scene": "report", "language": "zh", "canvas": "ppt169",
+            "image_strategy": "off", "color_scheme": "blue",
+            "page_min": 5, "page_max": 8,
+        })
+        check("generate accepted", b.get("code") == 0, str(b)[:60])
+        task_id = b["data"]["task_id"]
+        for _ in range(120):
+            time.sleep(1)
+            s, b = call("GET", f"/api/generate/{task_id}/status")
+            if b["data"]["status"] in ("done", "failed"):
+                break
+        check("generation done", b["data"]["status"] == "done", str(b)[:80])
+        s, b = call("GET", f"/api/generate/{task_id}/result")
+        rid = b["data"]["record"]["id"]
+        check("record ready", bool(rid), f"rid={rid}")
+
+        def chat(msg):
+            return call("POST", "/api/chat", {"record_id": rid, "message": msg})
+
+        s, b = chat("换成绿色")
+        d = b.get("data") or {}
+        check("change color", d.get("changed") is True, str(d.get("reply", ""))[:60])
+        check("record color updated", (d.get("record") or {}).get("color_scheme") == "green")
+
+        s, b = chat('把标题改成"知识平台周报"')
+        d = b.get("data") or {}
+        check("change title", d.get("changed") is True
+              and (d.get("record") or {}).get("title") == "知识平台周报",
+              str(d.get("reply", ""))[:60])
+
+        s, b = chat('把"知识周转"改成"信息流转"')
+        d = b.get("data") or {}
+        check("edit page replace", d.get("changed") is True, str(d.get("reply", ""))[:60])
+
+        s, b = chat("今天天气怎么样")
+        d = b.get("data") or {}
+        check("unknown intent no change", d.get("changed") is False
+              and "不支持" in d.get("reply", ""), str(d.get("reply", ""))[:50])
+
+        s, b = call("GET", f"/api/records/{rid}")
+        versions = b["data"]["versions"]
+        check("two versions kept", len(versions) == 2 and versions[0]["version_no"] == 1,
+              f"n={len(versions)}")
+
+        s, b = call("GET", f"/api/chat/{rid}/messages")
+        msgs = b["data"]["messages"]
+        check("messages logged", len(msgs) == 8, f"n={len(msgs)}")
+
+        s, raw = call("GET", f"/api/files/{rid}/download", raw=True)
+        check("download ok", s == 200 and raw[:2] == b"PK", f"{len(raw)}B")
+        p = ROOT / "tests" / "_chat_download.pptx"
+        p.write_bytes(raw)
+        from pptx import Presentation
+        prs = Presentation(str(p))
+        texts = "\n".join(sh.text_frame.text for sl in prs.slides
+                          for sh in sl.shapes if sh.has_text_frame)
+        check("new title in pptx", "知识平台周报" in texts)
+        check("edited text in pptx", "信息流转" in texts and "知识周转" not in texts)
+        p.unlink()
+    finally:
+        proc.terminate()
+
+    print("\n" + ("CHAT E2E ALL PASSED" if not failures
+                  else "FAILURES: " + "; ".join(failures)))
+    sys.exit(1 if failures else 0)
+
+
+if __name__ == "__main__":
+    main()

tests/e2e_image.py

@@ -0,0 +1,187 @@
+"""网络配图 e2e：image_strategy=web 全链路（test 通道）+ 空图源优雅降级。
+
+用法：python tests/e2e_image.py
+环境：IMAGE_SEARCH_MODE=test + IMAGE_SEARCH_TEST_DIR=<样例图目录>
+"""
+
+import json
+import os
+import subprocess
+import sys
+import time
+import urllib.request
+import urllib.error
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT))
+
+BASE = "http://127.0.0.1:5102"
+USER = f"img_e2e_{int(time.time())}"
+PASS = "img_e2e_123"
+SAMPLE = ("# 配图测试主题\n本季度完成检索服务上线\n准确率达到百分之九十八\n"
+          "# 运营情况\n用户增长明显\n文档数量翻倍\n")
+
+TMP = ROOT / "tests" / "_up"
+
+
+def make_test_images():
+    from PIL import Image
+    TMP.mkdir(exist_ok=True)
+    for i, color in enumerate([(21, 101, 192), (46, 125, 50), (123, 31, 162)]):
+        Image.new("RGB", (800, 500), color).save(str(TMP / f"sample_{i}.jpg"))
+
+
+COOKIE = {"value": ""}
+
+
+def call(method, path, payload=None):
+    data = None if payload is None else json.dumps(payload).encode("utf-8")
+    req = urllib.request.Request(BASE + path, data=data, method=method)
+    if payload is not None:
+        req.add_header("Content-Type", "application/json")
+    if COOKIE["value"]:
+        req.add_header("Cookie", COOKIE["value"])
+    try:
+        with urllib.request.urlopen(req, timeout=180) as resp:
+            sc = resp.headers.get("Set-Cookie")
+            if sc and not COOKIE["value"]:
+                COOKIE["value"] = sc.split(";")[0]
+            return resp.status, json.loads(resp.read() or b"{}")
+    except urllib.error.HTTPError as e:
+        try:
+            return e.code, json.loads(e.read() or b"{}")
+        except Exception:
+            return e.code, {}
+
+
+def generate_and_wait(title):
+    s, b = call("POST", "/api/generate", {
+        "title": title, "content": SAMPLE,
+        "scene": "report", "language": "zh", "canvas": "ppt169",
+        "image_strategy": "web", "color_scheme": "blue",
+        "page_min": 5, "page_max": 7,
+    })
+    if b.get("code") != 0:
+        return None, str(b)[:100]
+    task_id = b["data"]["task_id"]
+    for _ in range(120):
+        time.sleep(1)
+        s, b = call("GET", f"/api/generate/{task_id}/status")
+        st = b.get("data", {}).get("status")
+        if st in ("done", "failed"):
+            return (task_id if st == "done" else None), json.dumps(b.get("data"), ensure_ascii=False)[:120]
+    return None, "timeout"
+
+
+def main():
+    failures = []
+
+    def check(name, cond, extra=""):
+        print(f"  [{'OK' if cond else 'NG'}] {name} {extra}")
+        if not cond:
+            failures.append(name)
+
+    make_test_images()
+    env = {**os.environ, "PYTHONIOENCODING": "utf-8",
+           "LLM_BACKEND": "off",
+           "IMAGE_SEARCH_MODE": "test",
+           "IMAGE_SEARCH_TEST_DIR": str(TMP),
+           "MAX_IMAGE_SLIDES": "3"}
+    subprocess.run([sys.executable, "-m", "src.db.init"], cwd=str(ROOT),
+                   capture_output=True, env=env)
+    proc = subprocess.Popen(
+        [sys.executable, "-c",
+         f"import sys; sys.path.insert(0, r'{ROOT}');"
+         "from src.web.app import app; app.run(host='127.0.0.1', port=5102)"],
+        cwd=str(ROOT), env=env, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
+    )
+    try:
+        up = False
+        for _ in range(30):
+            time.sleep(0.5)
+            try:
+                call("GET", "/api/me")
+                up = True
+                break
+            except Exception:
+                continue
+        check("server up", up)
+
+        s, b = call("POST", "/api/register",
+                    {"username": USER, "password": PASS, "confirm_password": PASS})
+        check("register", s == 200 and b.get("code") == 0)
+
+        # ---- 场景1：配图成功 ----
+        task_id, info = generate_and_wait("配图全链路测试")
+        check("generation done (with images)", task_id is not None, info)
+        if task_id:
+            s, b = call("GET", f"/api/generate/{task_id}/result")
+            rec = b["data"]["record"]
+            project_dir = ROOT / "src" / "web" / "data" / "projects" / rec["project_dir"] \
+                if rec.get("project_dir") else None
+            # 兜底：从 versions 表拿不到目录名时按标题匹配最新目录
+            if not project_dir or not project_dir.exists():
+                dirs = sorted((ROOT / "src" / "web" / "data" / "projects").glob("配图全链路测试_*"))
+                project_dir = dirs[-1]
+            plan = json.loads((project_dir / "plan.json").read_text(encoding="utf-8"))
+            with_img = [sl for sl in plan["slides"] if sl.get("image")]
+            check("plan has image slides", len(with_img) >= 2, f"n={len(with_img)}")
+            check("image files exist", all(Path(sl["image"]).exists() for sl in with_img))
+            check("images inside project dir",
+                  all(str(sl["image"]).startswith(str(project_dir)) for sl in with_img))
+
+            s, raw = call("GET", f"/api/files/{rec['id']}/download", raw=True) \
+                if False else (None, None)
+            # 直接读导出文件验证 PPTX 内嵌图片形状
+            pptx_path = project_dir / "exports" / rec["file_name"] \
+                if rec.get("file_name") else None
+            if not pptx_path or not pptx_path.exists():
+                files = sorted((project_dir / "exports").glob("*.pptx"))
+                pptx_path = files[-1]
+            from pptx import Presentation
+            prs = Presentation(str(pptx_path))
+            pic_slides = [i for i, sl in enumerate(prs.slides)
+                          if any(sh.shape_type == 13 for sh in sl.shapes)]
+            check("pptx contains pictures", len(pic_slides) >= 2, f"slides={pic_slides}")
+            # 保真验证已在生成流程内通过（done 状态），此处确认无 image 键误报的副作用
+            check("no cover/end images", all(
+                sl.get("type") == "content" for sl in with_img))
+
+        # ---- 场景2：空图源 → 优雅降级 ----
+        empty_dir = TMP / "empty_src"
+        empty_dir.mkdir(exist_ok=True)
+        proc.terminate()
+        proc.wait()
+        env2 = {**env, "IMAGE_SEARCH_TEST_DIR": str(empty_dir)}
+        proc = subprocess.Popen(
+            [sys.executable, "-c",
+             f"import sys; sys.path.insert(0, r'{ROOT}');"
+             "from src.web.app import app; app.run(host='127.0.0.1', port=5102)"],
+            cwd=str(ROOT), env=env2, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
+        )
+        for _ in range(30):
+            time.sleep(0.5)
+            try:
+                call("GET", "/api/me")
+                break
+            except Exception:
+                continue
+        task_id, info = generate_and_wait("空图源降级测试")
+        check("generation done (fallback no-image)", task_id is not None, info)
+        if task_id:
+            s, b = call("GET", f"/api/generate/{task_id}/result")
+            rec = b["data"]["record"]
+            check("record still created", bool(rec and rec["id"]), str(rec)[:60])
+    finally:
+        proc.terminate()
+
+    import shutil
+    shutil.rmtree(TMP, ignore_errors=True)
+    print("\n" + ("IMAGE E2E ALL PASSED" if not failures
+                  else "FAILURES: " + "; ".join(failures)))
+    sys.exit(1 if failures else 0)
+
+
+if __name__ == "__main__":
+    main()

tests/e2e_llm.py

@@ -0,0 +1,153 @@
+# -*- coding: utf-8 -*-
+"""LLM 接入集成测试：规划/标题/意图/禁止降级/长文档两级规划。
+
+用法：python tests/e2e_llm.py
+"""
+import os
+import sys
+import time
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT))
+sys.stdout.reconfigure(encoding="utf-8", errors="replace")
+os.environ["PLAN_CHUNK_CHARS"] = "2500"  # 压低阈值，确保长文档用例真正走分段通读
+
+from src.agent import llm  # noqa: E402
+from src.agent.perception import GenerateConfig  # noqa: E402
+from src.agent import planning as P  # noqa: E402
+from src.agent.planning import build_plan, PlanningError  # noqa: E402
+from src.agent.chat import parse_intent_llm, parse_intent  # noqa: E402
+
+SAMPLE = """# 项目背景
+为提升知识周转效率启动平台建设，前期调研两个月
+# 本期成果
+部署完成基础服务，导入文档120份
+混合检索准确率较基线提升明显
+完成三个部门试点接入
+# 下一步
+扩大试点范围，建立运营机制
+"""
+
+failures = []
+
+
+def check(name, cond, extra=""):
+    print(f"  [{'OK' if cond else 'NG'}] {name} {extra}")
+    if not cond:
+        failures.append(name)
+
+
+def make_long_doc(sections=40):
+    """构造约 1.5-2 万字符的多节长文档（含数字，验证分段通读与页数硬约束）。"""
+    parts = []
+    for i in range(1, sections + 1):
+        parts.append(f"# 第{i}章节 阶段性工作汇报")
+        parts.append(f"本阶段完成任务{i}的设计与开发，投入人力3人，周期2周")
+        parts.append(f"完成度达到{i*2}%，质量抽检合格率98%")
+        parts.append("- 关键交付物已通过评审")
+        parts.append("- 遗留问题已登记跟踪")
+        parts.append("下一步将推进与业务系统的对接联调")
+    return "\n".join(parts)
+
+
+def main():
+    print("== 0. 禁止降级：LLM 失败必须报错而非照搬 ==")
+    cfg_nf = GenerateConfig(title="降级测试", user_id=1, content=SAMPLE,
+                            page_min=5, page_max=8)
+    orig_chat = llm.chat_json
+    try:
+        llm.chat_json = lambda *a, **k: None
+        try:
+            build_plan(cfg_nf)
+            check("no silent fallback", False, "未抛出 PlanningError")
+        except PlanningError as e:
+            check("no silent fallback", True, str(e)[:50])
+    finally:
+        llm.chat_json = orig_chat
+
+    ok = llm.ensure_backend()
+    check("llm backend", ok)
+
+    print("== 1. LLM 内容规划（短文档直通） ==")
+    t = time.time()
+    cfg = GenerateConfig(title="知识平台周报", user_id=1, content=SAMPLE,
+                         scene="report", language="zh", page_min=6, page_max=9)
+    r = build_plan(cfg)
+    plan = r.plan
+    slides = plan["slides"]
+    print(f"  mode={r.mode} pages={len(slides)} 耗时={time.time()-t:.0f}s")
+    check("llm mode used", r.mode == "llm")
+    check("page count in range", 6 <= len(slides) <= 9, f"n={len(slides)}")
+    check("cover first / end last", slides[0]["type"] == "cover" and slides[-1]["type"] == "end")
+    notes_n = sum(1 for s in slides if s.get("notes"))
+    check("per-slide notes", notes_n >= len(slides) - 2, f"notes={notes_n}")
+    generic = sum(1 for s in slides if s["title"] in ("",) or s["title"].startswith("要点"))
+    check("no generic titles", generic == 0)
+
+    def avg_len(sl):
+        bl = [len(b) for x in sl if x["type"] == "content" for b in x.get("content", [])]
+        return sum(bl) / max(len(bl), 1)
+    check("full-sentence bullets (short doc)", avg_len(slides) >= 12,
+          f"avg={avg_len(slides):.0f}")
+    for s in slides:
+        print(f"    - [{s['type']:8s}] {s['title'][:24]} notes={'Y' if s.get('notes') else '-'}")
+
+    print("== 2. 长文档两级规划（分段通读→汇总大纲） ==")
+    long_doc = make_long_doc()
+    t = time.time()
+    cfg2 = GenerateConfig(title="季度工作总结", user_id=1, content=long_doc,
+                          scene="report", language="zh", page_min=10, page_max=15)
+    r2 = build_plan(cfg2)
+    slides2 = r2.plan["slides"]
+    n2 = len(slides2)
+    print(f"  mode={r2.mode} pages={n2} 源={len(long_doc)}字 耗时={time.time()-t:.0f}s")
+    check("long doc page hard cap", n2 <= 15 and n2 >= 4, f"n={n2}")
+    notes2 = sum(1 for s in slides2 if s.get("notes"))
+    check("long doc has notes", notes2 >= n2 - 2, f"notes={notes2}")
+    covered = sum(1 for i in range(1, 41) if str(i) in "".join(
+        s["title"] + "".join(s["content"]) for s in slides2))
+    check("late sections represented", covered >= 3, f"covered_sections={covered}")
+    check("full-sentence bullets (long doc)", avg_len(slides2) >= 12,
+          f"avg={avg_len(slides2):.0f}")
+
+    print("== 3. 智能标题 ==")
+    t = time.time()
+    data = llm.chat_json(
+        "根据以下 PPT 内容拟 3 个标题（每个不超过 20 字）。"
+        '输出 JSON {"titles": ["..."]}。\n\n' + SAMPLE)
+    titles = [str(x).strip() for x in (data or {}).get("titles", []) if str(x).strip()]
+    print(f"  titles={titles} 耗时={time.time()-t:.0f}s")
+    check("title suggest", len(titles) >= 1)
+
+    print("== 4. LLM 意图解析（规则无法处理的句式） ==")
+    msg = '第2页末尾添加一条要点：风险与依赖已同步全组'
+    rule = parse_intent(msg, {})
+    llm_it = parse_intent_llm(msg, plan)
+    check("rule cannot parse (no quotes)", rule["type"] == "unknown", str(rule))
+    check("llm parses add", llm_it and llm_it.get("type") == "edit_page"
+          and llm_it.get("op") == "add" and llm_it.get("page_no") == 2,
+          str(llm_it))
+    msg2 = "把标题改成 AI 知识平台建设汇报"
+    it2 = parse_intent_llm(msg2, plan)
+    check("llm parses title", it2 and it2.get("type") == "change_title"
+          and "知识平台" in it2.get("title", ""), str(it2))
+
+    print("== 5. 渲染 LLM plan（含 notes） ==")
+    import json
+    import tempfile
+    from src.engine_bridge import render_plan, verify_output
+    tmp = Path(tempfile.mkdtemp(prefix="llm_plan_"))
+    p = tmp / "plan.json"
+    p.write_text(json.dumps(plan, ensure_ascii=False, indent=2), encoding="utf-8")
+    out = tmp / "out.pptx"
+    render_plan(p, out)
+    vok, detail = verify_output(out, p)
+    check("render+verify llm plan", vok, detail[:60])
+
+    print("\n" + ("LLM E2E ALL PASSED" if not failures else "FAILURES: " + "; ".join(failures)))
+    sys.exit(1 if failures else 0)
+
+
+if __name__ == "__main__":
+    main()

tests/e2e_settings.py

@@ -0,0 +1,200 @@
+"""个人设置 e2e：偏好读写校验、改用户名、改密码、历史语言筛选。
+
+用法：python tests/e2e_settings.py
+"""
+
+import json
+import os
+import sqlite3
+import subprocess
+import sys
+import time
+import urllib.request
+import urllib.error
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT))
+
+BASE = "http://127.0.0.1:5101"
+_TS = int(time.time())
+UA = f"sxa{_TS}"
+UB = f"sxb{_TS}"
+NEW_NAME = f"sxc{_TS}"
+PASS = "set_e2e_123"
+
+COOKIE = {"value": ""}
+
+
+def call(method, path, payload=None):
+    url = BASE + path
+    data = None if payload is None else json.dumps(payload).encode("utf-8")
+    req = urllib.request.Request(url, data=data, method=method)
+    if payload is not None:
+        req.add_header("Content-Type", "application/json")
+    if COOKIE["value"]:
+        req.add_header("Cookie", COOKIE["value"])
+    try:
+        with urllib.request.urlopen(req, timeout=60) as resp:
+            body = resp.read()
+            sc = resp.headers.get("Set-Cookie")
+            if sc and not COOKIE["value"]:
+                COOKIE["value"] = sc.split(";")[0]
+            return resp.status, json.loads(body or b"{}")
+    except urllib.error.HTTPError as e:
+        try:
+            return e.code, json.loads(e.read() or b"{}")
+        except Exception:
+            return e.code, {}
+
+
+def drop_cookie():
+    COOKIE["value"] = ""
+
+
+def insert_record(user_id, title, language, created_at):
+    from config import DB_PATH
+    conn = sqlite3.connect(str(DB_PATH))
+    conn.execute(
+        "INSERT INTO ppt_records (user_id,title,content,scene,language,color_scheme,"
+        "page_min,page_max,status,created_at) VALUES (?,?,?,?,?,?,?,?,?,?)",
+        (user_id, title, "内容示例文本，用于语言筛选测试", "report", language,
+         "blue", 10, 15, "done", created_at),
+    )
+    conn.commit()
+    uid = conn.execute("SELECT id FROM ppt_records WHERE title=?", (title,)).fetchone()[0]
+    conn.close()
+    return uid
+
+
+def main():
+    failures = []
+
+    def check(name, cond, extra=""):
+        print(f"  [{'OK' if cond else 'NG'}] {name} {extra}")
+        if not cond:
+            failures.append(name)
+
+    env = {**os.environ, "PYTHONIOENCODING": "utf-8"}
+    subprocess.run([sys.executable, "-m", "src.db.init"], cwd=str(ROOT),
+                   capture_output=True, env=env)
+    proc = subprocess.Popen(
+        [sys.executable, "-c",
+         f"import sys; sys.path.insert(0, r'{ROOT}');"
+         "from src.web.app import app; app.run(host='127.0.0.1', port=5101)"],
+        cwd=str(ROOT), env=env, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
+    )
+    try:
+        up = False
+        for _ in range(30):
+            time.sleep(0.5)
+            try:
+                call("GET", "/api/me")
+                up = True
+                break
+            except Exception:
+                continue
+        check("server up", up)
+
+        s, b = call("POST", "/api/register",
+                    {"username": UA, "password": PASS, "confirm_password": PASS})
+        check("register A", s == 200 and b.get("code") == 0)
+
+        # ---- GET /api/settings 默认值 ----
+        s, b = call("GET", "/api/settings")
+        p = (b.get("data") or {}).get("preferences") or {}
+        check("settings defaults", s == 200 and b["data"]["username"] == UA
+              and p.get("default_scene") == "training" and p.get("default_color") == "blue"
+              and p.get("default_language") == "zh" and p.get("default_canvas") == "ppt169"
+              and p.get("default_image_strategy") == "web"
+              and p.get("default_page_min") == 10 and p.get("default_page_max") == 15,
+              str(p)[:80])
+
+        # ---- 偏好保存（部分字段）----
+        s, b = call("PUT", "/api/settings/preferences",
+                    {"default_color": "purple", "default_language": "ja",
+                     "default_page_min": 5, "default_page_max": 8})
+        p = ((b.get("data") or {}).get("preferences")) or {}
+        check("save prefs partial", s == 200 and p.get("default_color") == "purple"
+              and p.get("default_language") == "ja" and p.get("default_page_min") == 5
+              and p.get("default_page_max") == 8 and p.get("default_scene") == "training",
+              str(p)[:80])
+        s, b = call("PUT", "/api/settings/preferences", {"default_color": "pink"})
+        check("prefs invalid enum -> 5002", s == 400 and b.get("code") == 5002, str(b)[:60])
+        s, b = call("PUT", "/api/settings/preferences", {"default_page_min": 20, "default_page_max": 8})
+        check("prefs min>max -> 2002", s == 400 and b.get("code") == 2002, str(b)[:60])
+
+        # ---- 改用户名 ----
+        s, b = call("PUT", "/api/settings/username", {"username": "ab"})
+        check("username too short -> 5002", s == 400 and b.get("code") == 5002, str(b)[:50])
+        s, b = call("POST", "/api/register",
+                    {"username": UB, "password": PASS, "confirm_password": PASS})
+        drop_cookie()
+        call("POST", "/api/login", {"username": UA, "password": PASS})
+        s, b = call("PUT", "/api/settings/username", {"username": UB})
+        check("username duplicate -> 1001", s == 409 and b.get("code") == 1001, str(b)[:50])
+        new_name = NEW_NAME
+        s, b = call("PUT", "/api/settings/username", {"username": new_name})
+        s2, b2 = call("GET", "/api/me")
+        check("username changed", s == 200 and b2["data"]["username"] == new_name,
+              str(b2)[:60])
+
+        # ---- 改密码 ----
+        s, b = call("PUT", "/api/settings/password",
+                    {"old_password": "wrong_old", "new_password": "newpwd456",
+                     "confirm_password": "newpwd456"})
+        check("wrong old pwd -> 5001", s == 400 and b.get("code") == 5001, str(b)[:50])
+        s, b = call("PUT", "/api/settings/password",
+                    {"old_password": PASS, "new_password": "newpwd456",
+                     "confirm_password": "different"})
+        check("confirm mismatch -> 1003", s == 400 and b.get("code") == 1003, str(b)[:50])
+        s, b = call("PUT", "/api/settings/password",
+                    {"old_password": PASS, "new_password": "abc",
+                     "confirm_password": "abc"})
+        check("short new pwd -> 5002", s == 400 and b.get("code") == 5002, str(b)[:50])
+        s, b = call("PUT", "/api/settings/password",
+                    {"old_password": PASS, "new_password": "newpwd456",
+                     "confirm_password": "newpwd456"})
+        check("password changed", s == 200 and b.get("code") == 0)
+        drop_cookie()
+        s, _ = call("POST", "/api/login", {"username": new_name, "password": PASS})
+        check("old password rejected -> 401", s == 401)
+        s, b = call("POST", "/api/login", {"username": new_name, "password": "newpwd456"})
+        check("login with new password", s == 200 and b.get("code") == 0)
+
+        # ---- 历史语言筛选 ----
+        uid = None
+        from src.agent import memory as mem
+        u = mem.get_user_by_name(new_name)
+        uid = u["id"]
+        insert_record(uid, "筛选测试中文稿", "zh", "2026-08-25 10:00:00")
+        insert_record(uid, "フィルタ日本語資料", "ja", "2026-08-25 10:01:00")
+        s, b = call("GET", "/api/records?language=ja")
+        titles_ja = [r["title"] for r in b["data"]["records"]]
+        s, b = call("GET", "/api/records?language=zh")
+        titles_zh = [r["title"] for r in b["data"]["records"]]
+        s, b = call("GET", "/api/records")
+        titles_all = [r["title"] for r in b["data"]["records"]]
+        check("filter ja only", "フィルタ日本語資料" in titles_ja
+              and "筛选测试中文稿" not in titles_ja, str(titles_ja)[:70])
+        check("filter zh only", "筛选测试中文稿" in titles_zh
+              and "フィルタ日本語資料" not in titles_zh, str(titles_zh)[:70])
+        check("no filter shows both", "筛选测试中文稿" in titles_all
+              and "フィルタ日本語資料" in titles_all, f"n={len(titles_all)}")
+
+        # 清理测试插入的记录
+        conn = sqlite3.connect(str(ROOT / "src" / "web" / "data" / "ppt.db"))
+        conn.execute("DELETE FROM ppt_records WHERE title IN (?,?)",
+                     ("筛选测试中文稿", "フィルタ日本語資料"))
+        conn.commit()
+        conn.close()
+    finally:
+        proc.terminate()
+
+    print("\n" + ("SETTINGS E2E ALL PASSED" if not failures
+                  else "FAILURES: " + "; ".join(failures)))
+    sys.exit(1 if failures else 0)
+
+
+if __name__ == "__main__":
+    main()

tests/e2e_upload.py

@@ -0,0 +1,225 @@
+"""上传解析 e2e：上传 → 解析回填全链路（覆盖 md/txt/pdf/html/xlsx/pptx/URL + 异常）。
+
+用法：python tests/e2e_upload.py
+"""
+
+import io
+import json
+import os
+import subprocess
+import sys
+import threading
+import time
+import urllib.request
+import urllib.error
+import uuid
+from http.server import BaseHTTPRequestHandler, HTTPServer
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT))
+
+BASE = "http://127.0.0.1:5099"
+USER = f"up_e2e_{int(time.time())}"
+PASS = "up_e2e_123"
+TMP = ROOT / "tests" / "_up"
+
+COOKIE = {"value": ""}
+
+
+def call(method, path, payload=None, raw=False):
+    url = BASE + path
+    data = None if payload is None else json.dumps(payload).encode("utf-8")
+    req = urllib.request.Request(url, data=data, method=method)
+    if payload is not None:
+        req.add_header("Content-Type", "application/json")
+    if COOKIE["value"]:
+        req.add_header("Cookie", COOKIE["value"])
+    try:
+        with urllib.request.urlopen(req, timeout=120) as resp:
+            body = resp.read()
+            sc = resp.headers.get("Set-Cookie")
+            if sc and not COOKIE["value"]:
+                COOKIE["value"] = sc.split(";")[0]
+            return resp.status, (body if raw else json.loads(body or b"{}"))
+    except urllib.error.HTTPError as e:
+        body = e.read()
+        try:
+            parsed = json.loads(body or b"{}")
+        except Exception:
+            parsed = {"raw": body.decode("utf-8", "ignore")}
+        return e.code, parsed
+
+
+def upload(filename, data, with_cookie=True):
+    """multipart 上传，返回 (status, body)。"""
+    boundary = "----pptagente2e" + uuid.uuid4().hex[:8]
+    head = (f'--{boundary}\r\nContent-Disposition: form-data; name="file"; '
+            f'filename="{filename}"\r\nContent-Type: application/octet-stream\r\n\r\n')
+    body = head.encode("utf-8") + data + f"\r\n--{boundary}--\r\n".encode("utf-8")
+    req = urllib.request.Request(BASE + "/api/upload", data=body, method="POST")
+    req.add_header("Content-Type", f"multipart/form-data; boundary={boundary}")
+    if with_cookie and COOKIE["value"]:
+        req.add_header("Cookie", COOKIE["value"])
+    try:
+        with urllib.request.urlopen(req, timeout=60) as resp:
+            return resp.status, json.loads(resp.read() or b"{}")
+    except urllib.error.HTTPError as e:
+        try:
+            return e.code, json.loads(e.read() or b"{}")
+        except Exception:
+            return e.code, {}
+
+
+def make_samples():
+    """生成本地样例文件，返回 {格式名: (文件名, 字节)}。"""
+    TMP.mkdir(exist_ok=True)
+    samples = {}
+
+    md = "# UPLOAD_E2E_MARKER_MD\n知识库平台建设背景说明\n本期完成检索服务上线\n"
+    samples["md"] = ("sample.md", md.encode("utf-8"))
+
+    txt = "UPLOAD_E2E_MARKER_TXT\n知识库平台建设背景说明\n本期完成检索服务上线"
+    samples["txt"] = ("sample.txt", txt.encode("utf-8"))
+
+    html = ("<html><head><style>.x{color:red}</style></head><body>"
+            "<h1>UPLOAD_E2E_MARKER_HTML</h1><p>知识库平台季度汇报</p></body></html>")
+    samples["html"] = ("sample.html", html.encode("utf-8"))
+
+    import fitz
+    doc = fitz.open()
+    page = doc.new_page()
+    page.insert_text((72, 72), "UPLOAD_E2E_MARKER_PDF quarterly report of knowledge platform")
+    buf = io.BytesIO()
+    doc.save(buf)
+    samples["pdf"] = ("sample.pdf", buf.getvalue())
+
+    from openpyxl import Workbook
+    wb = Workbook()
+    ws = wb.active
+    ws["A1"] = "UPLOAD_E2E_MARKER_XLSX"
+    ws["A2"] = "knowledge platform kpi"
+    buf = io.BytesIO()
+    wb.save(buf)
+    samples["xlsx"] = ("sample.xlsx", buf.getvalue())
+
+    from pptx import Presentation
+    from pptx.util import Inches
+    prs = Presentation()
+    slide = prs.slides.add_slide(prs.slide_layouts[5])
+    tb = slide.shapes.add_textbox(Inches(1), Inches(1), Inches(6), Inches(1))
+    tb.text_frame.text = "UPLOAD_E2E_MARKER_PPTX knowledge platform review"
+    buf = io.BytesIO()
+    prs.save(buf)
+    samples["pptx"] = ("sample.pptx", buf.getvalue())
+
+    return samples
+
+
+def start_url_server():
+    """本地起一个极简网页服务器，供 URL 抓取分支测试。"""
+    class H(BaseHTTPRequestHandler):
+        def do_GET(self):
+            body = (b"<html><head><title>t</title></head><body>"
+                    b"<h1>UPLOAD_E2E_MARKER_URL</h1><p>web capture ok</p></body></html>")
+            self.send_response(200)
+            self.send_header("Content-Type", "text/html; charset=utf-8")
+            self.end_headers()
+            self.wfile.write(body)
+
+        def log_message(self, *a):
+            pass
+
+    srv = HTTPServer(("127.0.0.1", 5199), H)
+    threading.Thread(target=srv.serve_forever, daemon=True).start()
+    return srv
+
+
+def main():
+    failures = []
+
+    def check(name, cond, extra=""):
+        print(f"  [{'OK' if cond else 'NG'}] {name} {extra}")
+        if not cond:
+            failures.append(name)
+
+    env = {**os.environ, "PYTHONIOENCODING": "utf-8"}
+    subprocess.run([sys.executable, "-m", "src.db.init"], cwd=str(ROOT),
+                   capture_output=True, env=env)
+    proc = subprocess.Popen(
+        [sys.executable, "-c",
+         f"import sys; sys.path.insert(0, r'{ROOT}');"
+         "from src.web.app import app; app.run(host='127.0.0.1', port=5099)"],
+        cwd=str(ROOT), env=env, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
+    )
+    srv = start_url_server()
+    samples = make_samples()
+    try:
+        up = False
+        for _ in range(30):
+            time.sleep(0.5)
+            try:
+                call("GET", "/api/me")
+                up = True
+                break
+            except Exception:
+                continue
+        check("server up", up)
+
+        s, b = call("POST", "/api/register",
+                    {"username": USER, "password": PASS, "confirm_password": PASS})
+        check("register", s == 200 and b.get("code") == 0)
+
+        # ---- 鉴权与非法输入 ----
+        s, b = upload("x.md", b"# hi", with_cookie=False)
+        check("upload without login -> 401", s == 401, str(s))
+        s, b = upload("evil.exe", b"MZ fake")
+        check("upload exe -> 3001", s == 400 and b.get("code") == 3001, str(b)[:60])
+        s, b = upload("huge.txt", b"A" * (21 * 1024 * 1024))
+        check("upload 21MB -> 3002", b.get("code") == 3002, str(b)[:60])
+        s, b = call("POST", "/api/parse", {"file_id": "nothex"})
+        check("parse bad file_id -> 400", s == 400, str(b)[:60])
+        s, b = call("POST", "/api/parse", {"file_id": "f" * 32})
+        check("parse unknown file_id -> 404", s == 404 and b.get("code") == 3003, str(b)[:60])
+
+        # ---- 各格式解析 ----
+        file_ids = {}
+        for fmt, (fname, data) in samples.items():
+            s, b = upload(fname, data)
+            ok_up = s == 200 and b.get("code") == 0 and b.get("data", {}).get("file_id")
+            check(f"upload {fmt}", bool(ok_up), str(b)[:60])
+            if not ok_up:
+                continue
+            fid = b["data"]["file_id"]
+            file_ids[fmt] = fid
+            marker = {
+                "md": "UPLOAD_E2E_MARKER_MD", "txt": "UPLOAD_E2E_MARKER_TXT",
+                "html": "UPLOAD_E2E_MARKER_HTML", "pdf": "UPLOAD_E2E_MARKER_PDF",
+                "xlsx": "UPLOAD_E2E_MARKER_XLSX", "pptx": "UPLOAD_E2E_MARKER_PPTX",
+            }[fmt]
+            s, b = call("POST", "/api/parse", {"file_id": fid})
+            d = b.get("data") or {}
+            check(f"parse {fmt}", s == 200 and b.get("code") == 0
+                  and marker in d.get("text", "") and d.get("chars", 0) >= 20,
+                  str(b)[:80])
+
+        # ---- URL 抓取分支 ----
+        s, b = call("POST", "/api/parse", {"url": "notaurl"})
+        check("parse bad url -> 400", s == 400, str(b)[:60])
+        s, b = call("POST", "/api/parse", {"url": "http://127.0.0.1:5199/page"})
+        d = b.get("data") or {}
+        check("parse local url", s == 200 and "UPLOAD_E2E_MARKER_URL" in d.get("text", ""),
+              str(b)[:80])
+    finally:
+        proc.terminate()
+        srv.shutdown()
+
+    import shutil
+    shutil.rmtree(TMP, ignore_errors=True)
+    print("\n" + ("UPLOAD E2E ALL PASSED" if not failures
+                  else "FAILURES: " + "; ".join(failures)))
+    sys.exit(1 if failures else 0)
+
+
+if __name__ == "__main__":
+    main()

tests/e2e_web.py

@@ -0,0 +1,197 @@
+"""Web 全链路端到端测试：初始化DB → 启动Flask → 注册 → 生成 → 轮询 → 下载。
+
+用法：
+    python tests/e2e_web.py
+"""
+
+import json
+import sys
+import time
+import urllib.request
+import urllib.error
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT))
+
+BASE = "http://127.0.0.1:5099"
+USER = f"e2e_user_{int(time.time())}"
+PASS = "e2e_pass_123"
+SAMPLE = """# 项目背景
+为提升内部知识周转效率，启动知识库平台建设
+已完成技术选型与原型验证
+# 本期成果
+部署完成基础服务，导入文档120份
+混合检索准确率显著优于基线方案
+# 下一步
+扩大试点范围至三个部门
+建立内容运营与质量巡检机制
+"""
+
+
+def call(method: str, path: str, payload=None, raw=False):
+    url = BASE + path
+    data = None if payload is None else json.dumps(payload).encode("utf-8")
+    req = urllib.request.Request(url, data=data, method=method)
+    if payload is not None:
+        req.add_header("Content-Type", "application/json")
+    # cookie 会话
+    try:
+        with urllib.request.urlopen(req, timeout=30) as resp:
+            body = resp.read()
+            set_cookie = resp.headers.get("Set-Cookie")
+            return resp.status, (body if raw else json.loads(body or b"{}")), set_cookie
+    except urllib.error.HTTPError as e:
+        body = e.read()
+        try:
+            return e.code, json.loads(body or b"{}"), e.headers.get("Set-Cookie")
+        except Exception:
+            return e.code, {"raw": body.decode("utf-8", "ignore")}, e.headers.get("Set-Cookie")
+
+
+COOKIE = {"value": ""}
+
+
+def callx(method, path, payload=None, raw=False):
+    status, body, sc = call(method, path, payload, raw)
+    if sc and COOKIE["value"] == "":
+        COOKIE["value"] = sc.split(";")[0]
+    return status, body
+
+
+# 注入 cookie
+_orig_call = call
+
+
+def call(method, path, payload=None, raw=False):  # noqa: F811
+    url = BASE + path
+    data = None if payload is None else json.dumps(payload).encode("utf-8")
+    req = urllib.request.Request(url, data=data, method=method)
+    if payload is not None:
+        req.add_header("Content-Type", "application/json")
+    if COOKIE["value"]:
+        req.add_header("Cookie", COOKIE["value"])
+    try:
+        with urllib.request.urlopen(req, timeout=60) as resp:
+            body = resp.read()
+            sc = resp.headers.get("Set-Cookie")
+            if sc and not COOKIE["value"]:
+                COOKIE["value"] = sc.split(";")[0]
+            return resp.status, (body if raw else json.loads(body or b"{}"))
+    except urllib.error.HTTPError as e:
+        body = e.read()
+        try:
+            parsed = json.loads(body or b"{}")
+        except Exception:
+            parsed = {"raw": body.decode("utf-8", "ignore")}
+        sc = e.headers.get("Set-Cookie")
+        if sc and not COOKIE["value"]:
+            COOKIE["value"] = sc.split(";")[0]
+        return e.code, parsed
+
+
+def main():
+    failures = []
+
+    def check(name, cond, extra=""):
+        print(f"  [{'OK' if cond else 'NG'}] {name} {extra}")
+        if not cond:
+            failures.append(name)
+
+    print("== 初始化数据库 ==")
+    import subprocess
+    r = subprocess.run([sys.executable, "-m", "src.db.init"], cwd=str(ROOT),
+                       capture_output=True, text=True, encoding="utf-8", errors="replace",
+                       env={**__import__("os").environ, "PYTHONIOENCODING": "utf-8"})
+    check("db init", r.returncode == 0, r.stdout.strip()[-40:] + r.stderr.strip()[-80:])
+
+    print("== 启动 Flask ==")
+    proc = subprocess.Popen(
+        [sys.executable, "-c",
+         f"import sys; sys.path.insert(0, r'{ROOT}');"
+         "from src.web.app import app; app.run(host='127.0.0.1', port=5099)"],
+        cwd=str(ROOT), env={**__import__("os").environ, "PYTHONIOENCODING": "utf-8"},
+        stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
+    )
+    try:
+        up = False
+        for _ in range(30):
+            time.sleep(0.5)
+            try:
+                s, _b = call("GET", "/api/me")
+                up = True
+                break
+            except Exception:
+                continue
+        check("server up", up)
+
+        print("== 认证链路 ==")
+        s, b = call("POST", "/api/register",
+                    {"username": USER, "password": PASS, "confirm_password": PASS})
+        check("register", s == 200 and b.get("code") == 0, str(b)[:60])
+
+        s, b = call("GET", "/api/me")
+        check("me with session", s == 200 and b.get("data", {}).get("username") == USER)
+        prefs = b.get("data", {}).get("preferences") or {}
+
+        print("== 校验拦截 ==")
+        s, b = call("POST", "/api/generate",
+                    {**{"title": "t", "content": "太短", "scene": "report"},
+                     **{k: prefs.get("default_" + k[0:-4], v) for k, v in []}})
+        check("short content rejected (2001)", b.get("code") == 2001, str(b)[:60])
+
+        print("== 真实生成 ==")
+        gen_payload = {
+            "title": "E2E冒烟报告",
+            "content": SAMPLE,
+            "scene": "report", "language": "zh", "canvas": "ppt169",
+            "image_strategy": "off", "color_scheme": "blue",
+            "page_min": 5, "page_max": 8,
+        }
+        s, b = call("POST", "/api/generate", gen_payload)
+        check("generate accepted", s == 200 and b.get("code") == 0, str(b)[:80])
+        task_id = (b.get("data") or {}).get("task_id")
+
+        final = None
+        for _ in range(120):
+            time.sleep(1)
+            s, b = call("GET", f"/api/generate/{task_id}/status")
+            d = b.get("data") or {}
+            if d.get("status") in ("done", "failed"):
+                final = d
+                break
+        check("generation done", final and final.get("status") == "done", str(final)[:120])
+
+        s, b = call("GET", f"/api/generate/{task_id}/result")
+        record = (b.get("data") or {}).get("record") or {}
+        rid = record.get("id")
+        check("result has record", bool(rid), f"id={rid} pages={record.get('page_count')} size={record.get('file_size')}")
+        check("page_count sane", 5 <= (record.get("page_count") or 0) <= 8,
+              str(record.get("page_count")))
+
+        print("== 历史与下载 ==")
+        s, b = call("GET", "/api/records?keyword=E2E")
+        recs = ((b.get("data") or {}).get("records")) or []
+        check("records list contains new", any(x["id"] == rid for x in recs))
+
+        s, body_bytes = call("GET", f"/api/files/{rid}/download", raw=True)
+        ok_pptx = body_bytes[:2] == b"P" + b"K"  # PK zip header
+        check("download pptx bytes", s == 200 and ok_pptx, f"{len(body_bytes)} bytes head={body_bytes[:2]}")
+
+        pptx_path = ROOT / "tests" / "_e2e_download.pptx"
+        pptx_path.write_bytes(body_bytes)
+        from pptx import Presentation
+        prs = Presentation(str(pptx_path))
+        texts = "\n".join(sh.text_frame.text for sl in prs.slides
+                          for sh in sl.shapes if sh.has_text_frame)
+        check("pptx contains title", "E2E冒烟报告" in texts)
+        check("pptx contains content keyword", "混合检索" in texts)
+    finally:
+        proc.terminate()
+
+    print("\n" + ("E2E ALL PASSED" if not failures else "E2E FAILURES: " + "; ".join(failures)))
+    sys.exit(1 if failures else 0)
+
+
+if __name__ == "__main__":
+    main()

tests/smoke_build_plan.py

@@ -0,0 +1,98 @@
+"""冒烟测试：build_plan 结构 + aura 引擎直接渲染。
+
+用法：
+    python tests/smoke_build_plan.py          # 仅构建 plan 并打印结构
+    python tests/smoke_build_plan.py --render # 追加真实渲染 + 保真验证
+"""
+
+import json
+import sys
+import tempfile
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
+
+from src.agent.perception import GenerateConfig  # noqa: E402
+from src.agent.planning import build_plan  # noqa: E402
+
+SAMPLE_ZH = """# 本月完成情况
+完成模块A开发，进度符合计划
+修复线上3个缺陷，系统可用性提升至99.9%
+新增用户反馈收集通道
+# 下月计划
+启动模块B概要设计
+组织跨部门代码评审
+开展性能压测与容量评估
+补充核心链路自动化测试用例
+# 风险与求助
+人力存在缺口，希望增援1名后端
+第三方接口联调依赖对方排期
+"""
+
+SAMPLE_JA = """# 今月の進捗
+モジュールAの開発完了、予定どおり
+本番障害3件を修正し、可用性99.9%を達成
+# 来月の計画
+モジュールBの概要設計を開始
+性能テストを実施する
+"""
+
+
+def make_cfg(lang="zh"):
+    return GenerateConfig(
+        title="七月项目周报" if lang == "zh" else "7月度プロジェクト報告",
+        user_id=1, content=SAMPLE_ZH if lang == "zh" else SAMPLE_JA,
+        scene="report", language=lang, canvas="ppt169",
+        image_strategy="off",
+        color_scheme="blue", page_min=6, page_max=10,
+    )
+
+
+def show(plan):
+    print(f"design={plan['design']} slides={len(plan['slides'])}")
+    for i, s in enumerate(plan["slides"], 1):
+        n = len(s.get("content") or [])
+        print(f"  {i:02d}. [{s['type']:9s}] {s.get('title','')[:30]} bullets={n}")
+
+
+def main():
+    failures = []
+    for lang in ("zh", "ja"):
+        r = build_plan(make_cfg(lang))
+        plan = r.plan
+        print(f"\n===== {lang} ===== warnings={r.warnings}")
+        show(plan)
+        types = [s["type"] for s in plan["slides"]]
+        if types[0] != "cover" or types[-1] != "end":
+            failures.append(f"{lang}: 首尾页面类型错误")
+        if not (6 <= len(plan["slides"]) <= 10) and lang == "zh":
+            failures.append(f"{lang}: 页数 {len(plan['slides'])} 超出目标区间")
+        if lang == "ja" and plan["design"] != "report-jp":
+            failures.append("ja 未映射到 report-jp")
+        if lang == "zh" and plan["design"] != "corp-blue":
+            failures.append("zh/blue 未映射到 corp-blue")
+
+    if "--render" in sys.argv:
+        from config import COMPANY_TEMPLATE
+        from src.engine_bridge import render_plan, verify_output
+        tmp = Path(tempfile.mkdtemp(prefix="aura_smoke_"))
+        for lang in ("zh", "ja"):
+            r = build_plan(make_cfg(lang))
+            p = tmp / f"plan_{lang}.json"
+            p.write_text(json.dumps(r.plan, ensure_ascii=False, indent=2),
+                         encoding="utf-8")
+            out = tmp / f"smoke_{lang}.pptx"
+            tpl = COMPANY_TEMPLATE if (lang == "ja" and COMPANY_TEMPLATE.exists()) else None
+            render_plan(p, out, template=tpl)
+            ok, detail = verify_output(out, p)
+            print(f"\n[render {lang}] -> {out}")
+            print(f"[verify {lang}] passed={ok}\n{detail}")
+            if not ok:
+                failures.append(f"{lang}: 渲染保真验证未通过")
+
+    print("\n" + ("ALL SMOKE PASSED" if not failures else "FAILURES: " + "; ".join(failures)))
+    sys.exit(1 if failures else 0)
+
+
+if __name__ == "__main__":
+    main()