dev-memory-cli 0.22.1 → 0.22.2

package/lib/dev_memory_setup.py

@@ -6,7 +6,7 @@ lazy-inits files on first write. Setup's job is now:
 
   1. Ensure skeleton exists (idempotent).
   2. Scan unsorted.md and present each entry to the user for classification.
-  3. Route user's choices into decisions/progress/risks/glossary/shared-*.
+  3. Route user's choices into decisions/risks/glossary/shared-*.
   4. Mark manifest.setup_completed = true.
 """
 
@@ -92,7 +92,7 @@ def command_init(args):
 # plan.json format:
 # {
 #   "classifications": [
-#     {"entry": "original bullet text", "kind": "decision|progress|next|risk|glossary|source|shared-*|skip"},
+#     {"entry": "original bullet text", "kind": "decision|risk|glossary|source|shared-*|skip"},
 #     ...
 #   ],
 #   "clear_unsorted_on_done": true
@@ -138,7 +138,12 @@ def _apply_classifications(paths, classifications):
         from dev_memory_common import append_to_section
         body = "\n".join(f"- {e}" for e in entries)
         path = paths[file_key]
-        append_to_section(path, section, body)
+        append_to_section(
+            path,
+            section,
+            body,
+            max_entries=20 if file_key.startswith("repo_") else None,
+        )
         tally[f"{file_key}:{section}"] = len(entries)
     return {"applied": tally, "skipped": skipped}
 

package/lib/dev_memory_summary.py

@@ -68,16 +68,26 @@ def _extract_codex(obj):
     }
 
 
-def _iter_core_messages(transcript_path):
+def _iter_core_messages(transcript_path, since_size=0):
     if not transcript_path:
         return []
     path = Path(transcript_path).expanduser()
     if not path.exists():
         return []
     out = []
-    with path.open(encoding="utf-8") as f:
-        for line in f:
-            line = line.strip()
+    with path.open("rb") as f:
+        while True:
+            start = f.tell()
+            raw = f.readline()
+            if not raw:
+                break
+            end = f.tell()
+            if end <= since_size:
+                continue
+            try:
+                line = raw.decode("utf-8").strip()
+            except UnicodeDecodeError:
+                continue
             if not line:
                 continue
             try:
@@ -86,12 +96,16 @@ def _iter_core_messages(transcript_path):
                 continue
             msg = _extract_claude(obj) or _extract_codex(obj)
             if msg:
+                msg["start_offset"] = start
+                msg["end_offset"] = end
                 out.append(msg)
     return out
 
 
 def _truncate(text, max_chars):
     text = (text or "").strip()
+    if not max_chars or max_chars < 0:
+        return text
     if len(text) <= max_chars:
         return text
     return text[: max_chars - 3].rstrip() + "..."
@@ -142,8 +156,8 @@ def _summary_job(job):
 def extract_core_payload(
     job,
     *,
-    max_messages=40,
-    max_message_chars=2000,
+    max_messages=0,
+    max_message_chars=0,
     max_memory_chars=6000,
     since_size=0,
     include_message_metadata=False,
@@ -159,14 +173,9 @@ def extract_core_payload(
         ("repo/decisions.md", repo_dir / "repo" / "decisions.md"),
         ("repo/glossary.md", repo_dir / "repo" / "glossary.md"),
     ]
-    messages = _iter_core_messages(job.get("transcript_path"))
-    if since_size:
-        # JSONL is line-oriented but byte offsets can land mid-line. Keep this
-        # option conservative for now: expose cursor metadata and let workers
-        # use recent-tail until a precise offset reader is needed.
-        pass
+    messages = _iter_core_messages(job.get("transcript_path"), since_size=since_size)
     messages = [m for m in messages if not _is_nonsemantic_user_text(m["text"])]
-    recent = messages[-max_messages:]
+    recent = messages[-max_messages:] if max_messages and max_messages > 0 else messages
     core_messages = []
     for m in recent:
         item = {
@@ -213,10 +222,10 @@ def main():
     sub = parser.add_subparsers(dest="command", required=True)
     p = sub.add_parser("extract-core", help="Extract core transcript messages and current memory for a summary job")
     p.add_argument("job", help="Path to a session-summary pending job JSON")
-    p.add_argument("--max-messages", type=int, default=40)
-    p.add_argument("--max-message-chars", type=int, default=2000)
+    p.add_argument("--max-messages", type=int, default=0, help="0 keeps every semantic message")
+    p.add_argument("--max-message-chars", type=int, default=0, help="0 keeps complete message text")
     p.add_argument("--max-memory-chars", type=int, default=6000)
-    p.add_argument("--since-size", type=int, default=0, help="Reserved cursor hint for future byte-offset extraction")
+    p.add_argument("--since-size", type=int, default=0, help="Read JSONL records ending after this byte offset")
     p.add_argument(
         "--include-message-metadata",
         action="store_true",

package/lib/ui-app.html

@@ -75,6 +75,14 @@ button { font-family: inherit; }
 .topbar .brand { display: flex; align-items: baseline; gap: 12px; min-width: 0; }
 .topbar .title { font-weight: 600; font-size: 15px; }
 .topbar .meta { color: var(--fg-muted); font-size: 12px; overflow: hidden; text-overflow: ellipsis; white-space: nowrap; }
+.view-tabs { display: flex; gap: 4px; }
+.view-tab { padding: 6px 10px; border: 1px solid var(--border); background: var(--bg-card); color: var(--fg); border-radius: var(--radius-sm); cursor: pointer; }
+.view-tab.active { color: var(--accent-fg); background: var(--accent); border-color: var(--accent); }
+.scan-table { width: 100%; border-collapse: collapse; background: var(--bg-card); border: 1px solid var(--border); font-size: 12px; }
+.scan-table th, .scan-table td { padding: 9px 10px; text-align: left; border-bottom: 1px solid var(--border-subtle); vertical-align: top; }
+.scan-table th { color: var(--fg-muted); background: var(--bg); font-weight: 600; }
+.scan-table tr:last-child td { border-bottom: 0; }
+.scan-table .num { text-align: right; font-variant-numeric: tabular-nums; }
 
 .main { display: grid; grid-template-columns: 320px 1fr; overflow: hidden; min-height: 0; }
 
@@ -453,6 +461,10 @@ button.icon-btn:hover {
       <span class="meta" id="meta">加载中…</span>
     </div>
     <div class="actions">
+      <span class="view-tabs">
+        <button class="view-tab active" id="memoryView">记忆</button>
+        <button class="view-tab" id="scanView">会话扫描</button>
+      </span>
       <button class="icon-btn" id="reload" title="重新加载">↻ 刷新</button>
     </div>
   </div>
@@ -490,6 +502,8 @@ button.icon-btn:hover {
 
 var state = {
   tree: null,
+  scan: null,
+  view: "memory",
   selectedRepoKey: null,
   selectedBranch: null,
   fileCache: new Map(),
@@ -672,6 +686,117 @@ function stat(label, value, small) {
   );
 }
 
+function formatBytes(value) {
+  var n = Number(value || 0);
+  if (n < 1024) return n + " B";
+  if (n < 1024 * 1024) return (n / 1024).toFixed(1) + " KB";
+  if (n < 1024 * 1024 * 1024) return (n / 1024 / 1024).toFixed(1) + " MB";
+  return (n / 1024 / 1024 / 1024).toFixed(2) + " GB";
+}
+
+function renderScan() {
+  var data = state.scan;
+  var content = q("#content");
+  var sidebar = q(".sidebar");
+  sidebar.style.display = "none";
+  content.style.gridColumn = "1 / -1";
+  content.innerHTML = "";
+  if (!data || !data.exists) {
+    content.appendChild(el("div", { class: "placeholder" }, "暂无会话扫描记录"));
+    return;
+  }
+  content.appendChild(el("div", { class: "repo-header" }, el("h2", null, "会话扫描"), el("div", { class: "path mono" }, data.scanRoot)));
+  var totals = el("div", { class: "stat-row" });
+  totals.appendChild(stat("扫描运行", data.run_count));
+  totals.appendChild(stat("涉及仓库", data.repos.length));
+  totals.appendChild(stat("总结 Token", data.usage ? (data.usage.total_tokens || 0).toLocaleString() : "未上报"));
+  totals.appendChild(stat("Token 未上报调用", data.unavailable_invocations));
+  content.appendChild(totals);
+
+  content.appendChild(el("h3", { class: "section-title" }, "仓库统计"));
+  var repoTable = el("table", { class: "scan-table" });
+  repoTable.innerHTML = "<thead><tr><th>仓库 / 分支</th><th class='num'>扫描次数</th><th class='num'>会话</th><th class='num'>原始大小</th><th class='num'>新增数据</th><th class='num'>总结 Token</th><th>最近扫描</th></tr></thead>";
+  var repoBody = el("tbody");
+  data.repos.forEach(function (repo) {
+    repoBody.appendChild(el("tr", null,
+      el("td", null, el("div", { class: "mono" }, repo.repo_key), el("div", { class: "sub" }, repo.branches.join(", "))),
+      el("td", { class: "num" }, repo.scan_count),
+      el("td", { class: "num" }, repo.session_count),
+      el("td", { class: "num" }, formatBytes(repo.raw_bytes)),
+      el("td", { class: "num" }, formatBytes(repo.new_bytes)),
+      el("td", { class: "num" }, Number(repo.summary_tokens || 0).toLocaleString()),
+      el("td", null, formatDateTime(repo.last_scanned_at))
+    ));
+  });
+  repoTable.appendChild(repoBody);
+  content.appendChild(repoTable);
+
+  content.appendChild(el("h3", { class: "section-title" }, "最近运行"));
+  var runTable = el("table", { class: "scan-table" });
+  runTable.innerHTML = "<thead><tr><th>时间 / Run ID</th><th>执行器</th><th class='num'>会话</th><th class='num'>新增数据</th><th class='num'>Token</th><th>结果</th></tr></thead>";
+  var runBody = el("tbody");
+  data.runs.forEach(function (run) {
+    var usage = run.summary_usage && run.summary_usage.total_tokens;
+    runBody.appendChild(el("tr", null,
+      el("td", null, el("div", null, formatDateTime(run.started_at)), el("div", { class: "mono" }, run.run_id)),
+      el("td", null, [run.executor || "-", run.model || ""].filter(Boolean).join(" / ")),
+      el("td", { class: "num" }, run.session_count),
+      el("td", { class: "num" }, formatBytes(run.new_bytes)),
+      el("td", { class: "num" }, usage == null ? "未上报" : Number(usage).toLocaleString()),
+      el("td", null, run.done_count + " 完成 · " + run.failed_count + " 失败 · " + run.skipped_count + " 跳过")
+    ));
+  });
+  runTable.appendChild(runBody);
+  content.appendChild(runTable);
+
+  var recentSessions = [];
+  data.runs.forEach(function (run) {
+    (run.sessions || []).forEach(function (session) {
+      recentSessions.push({ run_id: run.run_id, started_at: run.started_at, session: session });
+    });
+  });
+  recentSessions.sort(function (a, b) { return String(b.started_at || "").localeCompare(String(a.started_at || "")); });
+  recentSessions = recentSessions.slice(0, 200);
+  content.appendChild(el("h3", { class: "section-title" }, "最近会话"));
+  var sessionTable = el("table", { class: "scan-table" });
+  sessionTable.innerHTML = "<thead><tr><th>会话 / 仓库</th><th class='num'>原始大小</th><th class='num'>本次新增</th><th class='num'>原会话 Token</th><th class='num'>总结 Token</th><th>状态</th></tr></thead>";
+  var sessionBody = el("tbody");
+  recentSessions.forEach(function (row) {
+    var session = row.session;
+    sessionBody.appendChild(el("tr", null,
+      el("td", null,
+        el("div", { class: "mono" }, session.session_id || "-"),
+        el("div", null, [session.repo_key, session.branch].filter(Boolean).join(" / ")),
+        el("div", { class: "sub mono" }, session.path || "")
+      ),
+      el("td", { class: "num" }, formatBytes(session.raw_size)),
+      el("td", { class: "num" }, formatBytes(session.new_bytes)),
+      el("td", { class: "num" }, session.session_usage && session.session_usage.total_tokens != null ? Number(session.session_usage.total_tokens).toLocaleString() : "未上报"),
+      el("td", { class: "num" }, session.summary_usage && session.summary_usage.total_tokens != null ? Number(session.summary_usage.total_tokens).toLocaleString() : "未上报"),
+      el("td", null, session.status + (session.reason ? " · " + session.reason : ""))
+    ));
+  });
+  sessionTable.appendChild(sessionBody);
+  content.appendChild(sessionTable);
+}
+
+function setView(view) {
+  state.view = view;
+  q("#memoryView").classList.toggle("active", view === "memory");
+  q("#scanView").classList.toggle("active", view === "scan");
+  if (view === "scan") {
+    fetch("/api/session-scan").then(function (r) { return r.json(); }).then(function (data) {
+      state.scan = data;
+      q("#meta").textContent = data.run_count + " 次扫描 · " + data.repos.length + " 个仓库 · " + data.scanRoot;
+      renderScan();
+    }).catch(function (err) { q("#meta").textContent = "错误：" + err.message; });
+  } else {
+    q(".sidebar").style.display = "block";
+    q("#content").style.gridColumn = "";
+    loadTree();
+  }
+}
+
 function renderRepoList(repos, filterText) {
   var list = q("#repoList");
   list.innerHTML = "";
@@ -1068,8 +1193,10 @@ q("#search").addEventListener("input", function (e) {
 });
 q("#reload").addEventListener("click", function () {
   state.fileCache.clear();
-  loadTree();
+  setView(state.view);
 });
+q("#memoryView").addEventListener("click", function () { setView("memory"); });
+q("#scanView").addEventListener("click", function () { setView("scan"); });
 q("#modalClose").addEventListener("click", closeModal);
 q("#modalEdit").addEventListener("click", enterEditMode);
 q("#modalSave").addEventListener("click", saveModalEdit);

package/lib/ui-server.js

@@ -13,6 +13,11 @@ function getStorageRoot() {
   return process.env.DEV_ASSETS_ROOT || DEFAULT_STORAGE_ROOT;
 }
 
+function getScanRoot() {
+  return process.env.DEV_MEMORY_SCAN_ROOT
+    || path.join(path.dirname(getStorageRoot()), "jobs", "session-scan");
+}
+
 function safeReadDir(dir) {
   try {
     return fs.readdirSync(dir, { withFileTypes: true });
@@ -135,6 +140,62 @@ function buildTree() {
   return { storageRoot: root, exists: true, repos };
 }
 
+function buildSessionScanData() {
+  const root = getScanRoot();
+  const runsDir = path.join(root, "runs");
+  const runs = [];
+  for (const ent of safeReadDir(runsDir)) {
+    if (!ent.isFile() || !ent.name.endsWith(".json")) continue;
+    const value = safeReadJson(path.join(runsDir, ent.name));
+    if (value && typeof value === "object") runs.push(value);
+  }
+  runs.sort((a, b) => String(b.started_at || "").localeCompare(String(a.started_at || "")));
+  const repos = new Map();
+  const usage = {};
+  let unavailableInvocations = 0;
+  for (const run of runs) {
+    unavailableInvocations += Number(run.usage_unavailable_invocations || 0);
+    for (const [key, value] of Object.entries(run.summary_usage || {})) {
+      if (Number.isFinite(value)) usage[key] = (usage[key] || 0) + value;
+    }
+    for (const session of run.sessions || []) {
+      if (!session.repo_key) continue;
+      const row = repos.get(session.repo_key) || {
+        repo_key: session.repo_key,
+        branches: new Set(), sessions: new Set(), scan_count: 0,
+        raw_bytes: 0, new_bytes: 0, summary_tokens: 0, done: 0, failed: 0, last_scanned_at: null,
+      };
+      row.scan_count += 1;
+      row.sessions.add(session.session_id);
+      if (session.branch) row.branches.add(session.branch);
+      row.raw_bytes += Number(session.raw_size || 0);
+      row.new_bytes += Number(session.new_bytes || 0);
+      row.summary_tokens += Number((session.summary_usage || {}).total_tokens || 0);
+      if (session.status === "done") row.done += 1;
+      if (session.status === "failed") row.failed += 1;
+      if (!row.last_scanned_at || session.last_scanned_at > row.last_scanned_at) {
+        row.last_scanned_at = session.last_scanned_at;
+      }
+      repos.set(session.repo_key, row);
+    }
+  }
+  const repoRows = Array.from(repos.values()).map((row) => ({
+    ...row,
+    branches: Array.from(row.branches).sort(),
+    session_count: row.sessions.size,
+    sessions: undefined,
+  })).sort((a, b) => String(b.last_scanned_at || "").localeCompare(String(a.last_scanned_at || "")));
+  return {
+    scanRoot: root,
+    exists: !!safeStat(root),
+    run_count: runs.length,
+    usage: Object.keys(usage).length ? usage : null,
+    unavailable_invocations: unavailableInvocations,
+    repos: repoRows,
+    runs: runs.slice(0, 100),
+  };
+}
+
 function resolveSafePath(relPath) {
   const root = getStorageRoot();
   if (!relPath || typeof relPath !== "string") return null;
@@ -299,6 +360,11 @@ function start({ host = "127.0.0.1", port = 0, openBrowserFlag = true, readOnly
       res.end(JSON.stringify(tree));
       return;
     }
+    if (url.pathname === "/api/session-scan" && isRead) {
+      res.writeHead(200, { "content-type": "application/json; charset=utf-8" });
+      res.end(JSON.stringify(buildSessionScanData()));
+      return;
+    }
     if (url.pathname === "/api/injection-preview" && isRead) {
       const repoKey = url.searchParams.get("repo") || "";
       const branch = url.searchParams.get("branch") || "";
@@ -380,4 +446,4 @@ function start({ host = "127.0.0.1", port = 0, openBrowserFlag = true, readOnly
   return server;
 }
 
-module.exports = { start, buildTree, getStorageRoot };
+module.exports = { start, buildTree, buildSessionScanData, getStorageRoot, getScanRoot };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dev-memory-cli",
-  "version": "0.22.1",
+  "version": "0.22.2",
   "description": "CLI for dev-memory hooks and repo-local setup (formerly @xluos/dev-assets-cli)",
   "license": "MIT",
   "repository": {

package/scripts/hooks/_common.py

@@ -41,6 +41,7 @@ CONTEXT_SCRIPT = PACKAGE_ROOT / "lib" / "dev_memory_context.py"
 CAPTURE_SCRIPT = PACKAGE_ROOT / "lib" / "dev_memory_capture.py"
 SUMMARY_WORKER_SCRIPT = PACKAGE_ROOT / "scripts" / "hooks" / "session_summary_worker.py"
 DEFAULT_CONFIG_PATH = Path(os.environ.get("DEV_MEMORY_CONFIG_PATH", "~/.dev-memory/config.json")).expanduser()
+DEV_MEMORY_HOME = Path(os.environ.get("DEV_MEMORY_HOME", "~/.dev-memory")).expanduser()
 WORKSPACE_CONFIG_NAMES = (".dev-memory-workspace.json", ".dev-assets-workspace.json")
 
 
@@ -121,6 +122,28 @@ def read_hook_input():
         return {"raw": raw[:4000]}
 
 
+def register_session_scan_candidate(hook_input):
+    """Persist a lightweight Codex Stop hint without starting summarization."""
+    if not isinstance(hook_input, dict):
+        return None
+    session_id = hook_session_id(hook_input)
+    transcript_path = _hook_transcript_path(hook_input)
+    if not session_id and not transcript_path:
+        return None
+    source = session_id or transcript_path
+    digest = hashlib.sha1(str(source).encode("utf-8")).hexdigest()[:20]
+    path = DEV_MEMORY_HOME / "jobs" / "session-scan" / "candidates" / f"{digest}.json"
+    _atomic_write_json(path, {
+        "schema_version": 1,
+        "event": "Stop",
+        "registered_at": now_iso(),
+        "session_id": session_id,
+        "transcript_path": transcript_path,
+        "cwd": _first_string(hook_input.get("cwd"), str(REPO_ROOT)),
+    })
+    return path
+
+
 def resolve_assets_for(repo_root):
     """Resolve asset paths for an explicit repo root (workspace-mode friendly)."""
     repo_root_str = str(repo_root)
@@ -219,7 +242,7 @@ def extract_section(path, title):
     return None if is_placeholder(body) else body
 
 
-def extract_repo_file_body(path):
+def extract_repo_file_body(path, *, newest_first=False):
     """Extract full body of a repo-level file, skipping the H1 title and
     ``## 仓库`` metadata section. Placeholder-only sections within the file
     are dropped individually so that real content in sibling sections is
@@ -235,6 +258,11 @@ def extract_repo_file_body(path):
     for sec in sections:
         sec_body = strip_managed_markers(sec).strip()
         if sec_body and not is_placeholder(sec_body):
+            if newest_first and sec_body.startswith("## "):
+                heading, separator, body = sec_body.partition("\n")
+                if separator and body.strip():
+                    blocks = _split_recent_blocks(body)
+                    sec_body = heading + "\n\n" + "\n\n".join(blocks)
             kept.append(sec_body)
     body = "\n\n".join(kept).strip()
     return body or None
@@ -443,11 +471,14 @@ def _brief_section_keys(profile):
     return tuple(_BRIEF_PROFILES.get(profile, _BRIEF_PROFILES["standard"]).keys())
 
 
-def _extract_sections(paths, keys):
+def _extract_sections(paths, keys, *, repo_newest_first=False):
     out = []
     for file_key, title in keys:
         if title is None:
-            body = extract_repo_file_body(paths[file_key])
+            body = extract_repo_file_body(
+                paths[file_key],
+                newest_first=repo_newest_first and file_key in {"repo_decisions", "repo_glossary"},
+            )
             display_title = _REPO_DISPLAY_TITLES.get(file_key, paths[file_key].stem)
         else:
             body = extract_section(paths[file_key], title)
@@ -471,7 +502,7 @@ def _build_context_from_assets(assets, *, full=True, heading=None, brief_profile
 
     paths = assets["paths"]
     keys = _FULL_SECTION_KEYS if full else _brief_section_keys(brief_profile)
-    sections = _extract_sections(paths, keys)
+    sections = _extract_sections(paths, keys, repo_newest_first=full)
     max_lines, max_chars = (8, 700) if full else (3, 200)
     brief_limits = _BRIEF_PROFILES.get(brief_profile, _BRIEF_PROFILES["standard"])
 
@@ -898,9 +929,14 @@ def build_summary_input(job_path):
     job = json.loads(Path(job_path).read_text(encoding="utf-8"))
     return extract_core_payload(
         job,
-        max_messages=_int_env("DEV_MEMORY_SESSION_SUMMARY_MAX_MESSAGES", 30),
-        max_message_chars=_int_env("DEV_MEMORY_SESSION_SUMMARY_MAX_MESSAGE_CHARS", 1600),
+        max_messages=_int_env("DEV_MEMORY_SESSION_SUMMARY_MAX_MESSAGES", 0),
+        max_message_chars=_int_env("DEV_MEMORY_SESSION_SUMMARY_MAX_MESSAGE_CHARS", 0),
         max_memory_chars=_int_env("DEV_MEMORY_SESSION_SUMMARY_MAX_MEMORY_CHARS", 4000),
+        since_size=(
+            ((job.get("previous_job") or {}).get("processed") or {}).get("transcript_size", 0)
+            if isinstance(job.get("previous_job"), dict)
+            else 0
+        ),
     )
 
 
@@ -918,7 +954,8 @@ def build_summary_prompt(job_path, summary_input=None, summary_input_path=None):
         summary_input = build_summary_input(job_path)
     summary_input_json = json.dumps(summary_input, ensure_ascii=False, indent=2)
     input_path_line = f"- summary input JSON: {summary_input_path}\n" if summary_input_path else ""
-    return f"""你是 dev-memory 的后台会话总结 worker。
+    return f"""DEV_MEMORY_INTERNAL_SESSION_SUMMARY_V1
+你是 dev-memory 的后台会话总结 worker。
 
 输入：
 - job JSON: {job_path}

package/scripts/hooks/stop.py

@@ -5,13 +5,19 @@ from _common import (
     is_workspace_mode,
     log,
     maybe_record_head,
+    read_hook_input,
     record_head_all_repos,
+    register_session_scan_candidate,
     resolve_assets,
 )
 
 
 def main():
+    hook_input = read_hook_input()
     try:
+        candidate = register_session_scan_candidate(hook_input)
+        if candidate:
+            log(f"[dev-memory][Stop] registered session scan candidate {candidate}")
         if is_no_git_mode():
             log("[dev-memory][Stop] no-git mode: nothing to record (no HEAD)")
             return 0