dsh-session-manager 0.5.2 → 0.5.4

package/lib/clipboard-parser.js

@@ -1,208 +1,208 @@
-// Clipboard parser for AI-generated annotation payloads. Accepts a raw
-// string from navigator.clipboard.readText() and extracts a structured
-// {tags, note, priority?} payload. Tolerates fenced code blocks and
-// conversational wrappers, then enforces the same normalization rules
-// as the annotation store so the import button can call
-// annotationStore.save without re-validating.
-
-import { ANNOTATION_LIMITS, normalizeAnnotationPatch } from "./annotation-store.js";
-
-const NUL = String.fromCharCode(0);
-const BACKSLASH = String.fromCharCode(92);
-const QUOTE = String.fromCharCode(34);
-const NEWLINE = String.fromCharCode(10);
-const TAG_REJECT = /[,，\x00-\x1f]/;
-
-const tagRejectReason = (key, reason) => ({ field: "tags", kind: "rejected", key, reason });
-
-const trimLength = (value, max) => value.length > max ? value.slice(0, max) : value;
-
-const rejectNonPositive = value => value !== null && (!Number.isInteger(value) || value < 1 || value > 5);
-
-const normalizeStringArray = (value, warnings) => {
-  if (!Array.isArray(value)) throw new Error("tags must be an array of strings");
-  const seen = new Set();
-  const tags = [];
-  let dropped = 0;
-  for (const item of value) {
-    if (typeof item !== "string") { dropped++; continue; }
-    const trimmed = item.trim();
-    if (!trimmed) { dropped++; continue; }
-    if (trimmed.length > ANNOTATION_LIMITS.tagLength || TAG_REJECT.test(trimmed)) {
-      warnings.push(tagRejectReason(trimmed, "每个标签最多 32 字符，不能包含逗号或控制字符"));
-      dropped++;
-      continue;
-    }
-    const key = trimmed.toLowerCase();
-    if (seen.has(key)) continue;
-    seen.add(key);
-    tags.push(trimmed);
-  }
-  return { tags, dropped };
-};
-
-const normalizeNote = (value, warnings) => {
-  if (typeof value !== "string") throw new Error("note must be a string");
-  if (value.includes(NUL)) throw new Error("note must not contain NUL characters");
-  if (value.length > ANNOTATION_LIMITS.noteLength) {
-    warnings.push({ field: "note", kind: "truncated", from: value.length, to: ANNOTATION_LIMITS.noteLength });
-    return trimLength(value, ANNOTATION_LIMITS.noteLength);
-  }
-  return value;
-};
-
-const normalizePriority = (value, warnings) => {
-  if (value === undefined) return undefined;
-  // Treat AI-returned null as the default priority (3 = Normal) so the UI
-  // never has to special-case unset entries.
-  if (value === null) return 3;
-  if (rejectNonPositive(value)) {
-    warnings.push({ field: "priority", kind: "rejected", value, reason: "优先级必须为 1–5 的整数；省略或 null 会按 3（普通）保存" });
-    return undefined;
-  }
-  return value;
-};
-
-// Find the first balanced top-level JSON object in the text. Handles
-// strings, escapes and nested objects without relying on regular
-// expressions.
-const scanForJsonObject = text => {
-  const len = text.length;
-  for (let i = 0; i < len; i++) {
-    const ch = text[i];
-    if (ch !== "{") continue;
-    let depth = 0;
-    let inString = false;
-    let escape = false;
-    for (let j = i; j < len; j++) {
-      const c = text[j];
-      if (inString) {
-        if (escape) escape = false;
-        else if (c === BACKSLASH) escape = true;
-        else if (c === QUOTE) inString = false;
-        continue;
-      }
-      if (c === QUOTE) inString = true;
-      else if (c === "{") depth++;
-      else if (c === "}") {
-        depth--;
-        if (depth === 0) return text.slice(i, j + 1);
-      }
-    }
-  }
-  return null;
-};
-
-// Strip common conversational wrappers (Markdown fences, leading prose)
-// before scanning for JSON. The fence language tag is honored when
-// present so we still find JSON inside triple-backtick fences.
-const stripWrappers = text => {
-  // Strip UTF-8 BOM (U+FEFF) if present so JSON.parse doesn't reject it,
-  // then trim leading and trailing whitespace (the prior /s+/g was a literal
-  // "s+" regex and stripped only literal "s" characters, not whitespace).
-  let trimmed = text.charCodeAt(0) === 0xFEFF ? text.slice(1) : text;
-  trimmed = trimmed.replace(/^\s+|\s+$/g, "");
-  const fence = trimmed.match(/^```(?:json|JSON)?\s*([\s\S]*?)\s*```$/);
-  if (fence) return fence[1];
-  const braceStart = trimmed.indexOf("{");
-  const braceEnd = trimmed.lastIndexOf("}");
-  if (braceStart >= 0 && braceEnd > braceStart) return trimmed.slice(braceStart, braceEnd + 1);
-  return trimmed;
-};
-
-const ALLOWED_KEYS = new Set(["tags", "note", "priority"]);
-
-// Parse a clipboard string into an annotation patch. The returned patch
-// is compatible with annotationStore.save: tags/note/priority are
-// normalized the same way as the server-side validation. Surfaces a
-// list of warnings for truncated notes and dropped tags instead of
-// failing the import.
-export function parseClipboardAnnotation(raw) {
-  if (typeof raw !== "string" || raw.trim() === "") {
-    return { ok: false, error: "剪贴板为空" };
-  }
-  const stripped = stripWrappers(raw);
-  let parsed;
-  try { parsed = JSON.parse(stripped); }
-  catch {
-    const fallback = scanForJsonObject(stripped);
-    if (fallback === null) return { ok: false, error: "无法在剪贴板内容中找到合法的 JSON 对象" };
-    try { parsed = JSON.parse(fallback); }
-    catch { return { ok: false, error: "剪贴板中的 JSON 无法解析" }; }
-  }
-  if (parsed === null || typeof parsed !== "object" || Array.isArray(parsed)) {
-    return { ok: false, error: "剪贴板内容不是 JSON 对象" };
-  }
-  const warnings = [];
-  const unknown = Object.keys(parsed).filter(key => !ALLOWED_KEYS.has(key));
-  for (const key of unknown) warnings.push({ field: key, kind: "ignored", reason: "未知的标记字段，已忽略" });
-
-  let tags = [];
-  let note = "";
-  let priority;
-  if ("tags" in parsed) {
-    try { const result = normalizeStringArray(parsed.tags, warnings); tags = result.tags; }
-    catch (error) { return { ok: false, error: error.message }; }
-  }
-  if ("note" in parsed) {
-    try { note = normalizeNote(parsed.note, warnings); }
-    catch (error) { return { ok: false, error: error.message }; }
-  }
-  if ("priority" in parsed) {
-    const normalized = normalizePriority(parsed.priority, warnings);
-    if (normalized !== undefined) priority = normalized;
-  }
-  const patch = { tags, note, ...(priority !== undefined ? { priority } : {}) };
-  try { normalizeAnnotationPatch(patch); }
-  catch (error) { return { ok: false, error: error.message }; }
-  return { ok: true, annotation: patch, warnings };
-}
-
-// Static prompt templates copied to the clipboard. The Chinese version
-// is the default; English is exposed for users who prefer it. Both ask
-// the model to return strict JSON so the importer can validate it.
-export const CLIPBOARD_PROMPTS = Object.freeze({
-  zh: [
-    "请根据当前对话内容，为这个会话生成整理标记。",
-    "",
-    "要求：",
-    "1. 生成最多 20 个标签，每个标签最多 32 个字符。",
-    "2. 标签应简洁、具体，便于之后搜索。",
-    "3. 标签不要重复，不要生成无意义的标签。",
-    "4. 生成一段不超过 2000 个字符的备注，总结本次对话的主题、结论、未完成事项或下一步行动。",
-    "5. 不要编造对话中没有出现的信息。",
-    "6. 只返回合法 JSON，不要使用 Markdown 代码块，不要附加解释。",
-    "",
-    "返回格式：",
-    "{",
-    "  ", "tags", ": [", "标签1", ", ", "标签2", "],",
-    "  ", "note", ": ", "会话备注", ",",
-    "}"
-  ].join(NEWLINE),
-  en: [
-    "Summarize this conversation into session annotations.",
-    "",
-    "Requirements:",
-    "1. Up to 20 tags, at most 32 characters each.",
-    "2. Tags should be concise and help future searches.",
-    "3. Do not repeat tags or invent meaningless ones.",
-    "4. Write a note under 2000 characters summarizing the topic, conclusions, outstanding items and next steps.",
-    "5. Do not fabricate details that are not present in the conversation.",
-    "6. Return only valid JSON, no Markdown fences, no extra commentary.",
-    "",
-    "Return shape:",
-    "{",
-    "  ", "tags", ": [", "tag1", ", ", "tag2", "],",
-    "  ", "note", ": ", "session note", ",",
-    "}"
-  ].join(NEWLINE)
-});
-
-// Resolve the locale key from the active dictionary. Falls back to zh.
-export function clipboardPromptFor(dictionaries, language) {
-  const available = dictionaries ? Object.keys(dictionaries) : [];
-  const key = available.includes(language) ? language
-    : available.includes("zh") ? "zh"
-    : available[0] || "zh";
-  return { key, text: CLIPBOARD_PROMPTS[key] ?? CLIPBOARD_PROMPTS.zh };
+// Clipboard parser for AI-generated annotation payloads. Accepts a raw
+// string from navigator.clipboard.readText() and extracts a structured
+// {tags, note, priority?} payload. Tolerates fenced code blocks and
+// conversational wrappers, then enforces the same normalization rules
+// as the annotation store so the import button can call
+// annotationStore.save without re-validating.
+
+import { ANNOTATION_LIMITS, normalizeAnnotationPatch } from "./annotation-store.js";
+
+const NUL = String.fromCharCode(0);
+const BACKSLASH = String.fromCharCode(92);
+const QUOTE = String.fromCharCode(34);
+const NEWLINE = String.fromCharCode(10);
+const TAG_REJECT = /[,，\x00-\x1f]/;
+
+const tagRejectReason = (key, reason) => ({ field: "tags", kind: "rejected", key, reason });
+
+const trimLength = (value, max) => value.length > max ? value.slice(0, max) : value;
+
+const rejectNonPositive = value => value !== null && (!Number.isInteger(value) || value < 1 || value > 5);
+
+const normalizeStringArray = (value, warnings) => {
+  if (!Array.isArray(value)) throw new Error("tags must be an array of strings");
+  const seen = new Set();
+  const tags = [];
+  let dropped = 0;
+  for (const item of value) {
+    if (typeof item !== "string") { dropped++; continue; }
+    const trimmed = item.trim();
+    if (!trimmed) { dropped++; continue; }
+    if (trimmed.length > ANNOTATION_LIMITS.tagLength || TAG_REJECT.test(trimmed)) {
+      warnings.push(tagRejectReason(trimmed, "每个标签最多 32 字符，不能包含逗号或控制字符"));
+      dropped++;
+      continue;
+    }
+    const key = trimmed.toLowerCase();
+    if (seen.has(key)) continue;
+    seen.add(key);
+    tags.push(trimmed);
+  }
+  return { tags, dropped };
+};
+
+const normalizeNote = (value, warnings) => {
+  if (typeof value !== "string") throw new Error("note must be a string");
+  if (value.includes(NUL)) throw new Error("note must not contain NUL characters");
+  if (value.length > ANNOTATION_LIMITS.noteLength) {
+    warnings.push({ field: "note", kind: "truncated", from: value.length, to: ANNOTATION_LIMITS.noteLength });
+    return trimLength(value, ANNOTATION_LIMITS.noteLength);
+  }
+  return value;
+};
+
+const normalizePriority = (value, warnings) => {
+  if (value === undefined) return undefined;
+  // Treat AI-returned null as the default priority (3 = Normal) so the UI
+  // never has to special-case unset entries.
+  if (value === null) return 3;
+  if (rejectNonPositive(value)) {
+    warnings.push({ field: "priority", kind: "rejected", value, reason: "优先级必须为 1–5 的整数；省略或 null 会按 3（普通）保存" });
+    return undefined;
+  }
+  return value;
+};
+
+// Find the first balanced top-level JSON object in the text. Handles
+// strings, escapes and nested objects without relying on regular
+// expressions.
+const scanForJsonObject = text => {
+  const len = text.length;
+  for (let i = 0; i < len; i++) {
+    const ch = text[i];
+    if (ch !== "{") continue;
+    let depth = 0;
+    let inString = false;
+    let escape = false;
+    for (let j = i; j < len; j++) {
+      const c = text[j];
+      if (inString) {
+        if (escape) escape = false;
+        else if (c === BACKSLASH) escape = true;
+        else if (c === QUOTE) inString = false;
+        continue;
+      }
+      if (c === QUOTE) inString = true;
+      else if (c === "{") depth++;
+      else if (c === "}") {
+        depth--;
+        if (depth === 0) return text.slice(i, j + 1);
+      }
+    }
+  }
+  return null;
+};
+
+// Strip common conversational wrappers (Markdown fences, leading prose)
+// before scanning for JSON. The fence language tag is honored when
+// present so we still find JSON inside triple-backtick fences.
+const stripWrappers = text => {
+  // Strip UTF-8 BOM (U+FEFF) if present so JSON.parse doesn't reject it,
+  // then trim leading and trailing whitespace (the prior /s+/g was a literal
+  // "s+" regex and stripped only literal "s" characters, not whitespace).
+  let trimmed = text.charCodeAt(0) === 0xFEFF ? text.slice(1) : text;
+  trimmed = trimmed.replace(/^\s+|\s+$/g, "");
+  const fence = trimmed.match(/^```(?:json|JSON)?\s*([\s\S]*?)\s*```$/);
+  if (fence) return fence[1];
+  const braceStart = trimmed.indexOf("{");
+  const braceEnd = trimmed.lastIndexOf("}");
+  if (braceStart >= 0 && braceEnd > braceStart) return trimmed.slice(braceStart, braceEnd + 1);
+  return trimmed;
+};
+
+const ALLOWED_KEYS = new Set(["tags", "note", "priority"]);
+
+// Parse a clipboard string into an annotation patch. The returned patch
+// is compatible with annotationStore.save: tags/note/priority are
+// normalized the same way as the server-side validation. Surfaces a
+// list of warnings for truncated notes and dropped tags instead of
+// failing the import.
+export function parseClipboardAnnotation(raw) {
+  if (typeof raw !== "string" || raw.trim() === "") {
+    return { ok: false, error: "剪贴板为空" };
+  }
+  const stripped = stripWrappers(raw);
+  let parsed;
+  try { parsed = JSON.parse(stripped); }
+  catch {
+    const fallback = scanForJsonObject(stripped);
+    if (fallback === null) return { ok: false, error: "无法在剪贴板内容中找到合法的 JSON 对象" };
+    try { parsed = JSON.parse(fallback); }
+    catch { return { ok: false, error: "剪贴板中的 JSON 无法解析" }; }
+  }
+  if (parsed === null || typeof parsed !== "object" || Array.isArray(parsed)) {
+    return { ok: false, error: "剪贴板内容不是 JSON 对象" };
+  }
+  const warnings = [];
+  const unknown = Object.keys(parsed).filter(key => !ALLOWED_KEYS.has(key));
+  for (const key of unknown) warnings.push({ field: key, kind: "ignored", reason: "未知的标记字段，已忽略" });
+
+  let tags = [];
+  let note = "";
+  let priority;
+  if ("tags" in parsed) {
+    try { const result = normalizeStringArray(parsed.tags, warnings); tags = result.tags; }
+    catch (error) { return { ok: false, error: error.message }; }
+  }
+  if ("note" in parsed) {
+    try { note = normalizeNote(parsed.note, warnings); }
+    catch (error) { return { ok: false, error: error.message }; }
+  }
+  if ("priority" in parsed) {
+    const normalized = normalizePriority(parsed.priority, warnings);
+    if (normalized !== undefined) priority = normalized;
+  }
+  const patch = { tags, note, ...(priority !== undefined ? { priority } : {}) };
+  try { normalizeAnnotationPatch(patch); }
+  catch (error) { return { ok: false, error: error.message }; }
+  return { ok: true, annotation: patch, warnings };
+}
+
+// Static prompt templates copied to the clipboard. The Chinese version
+// is the default; English is exposed for users who prefer it. Both ask
+// the model to return strict JSON so the importer can validate it.
+export const CLIPBOARD_PROMPTS = Object.freeze({
+  zh: [
+    "请根据当前对话内容，为这个会话生成整理标记。",
+    "",
+    "要求：",
+    "1. 生成最多 20 个标签，每个标签最多 32 个字符。",
+    "2. 标签应简洁、具体，便于之后搜索。",
+    "3. 标签不要重复，不要生成无意义的标签。",
+    "4. 生成一段不超过 2000 个字符的备注，总结本次对话的主题、结论、未完成事项或下一步行动。",
+    "5. 不要编造对话中没有出现的信息。",
+    "6. 只返回合法 JSON，不要使用 Markdown 代码块，不要附加解释。",
+    "",
+    "返回格式：",
+    "{",
+    "  ", "tags", ": [", "标签1", ", ", "标签2", "],",
+    "  ", "note", ": ", "会话备注", ",",
+    "}"
+  ].join(NEWLINE),
+  en: [
+    "Summarize this conversation into session annotations.",
+    "",
+    "Requirements:",
+    "1. Up to 20 tags, at most 32 characters each.",
+    "2. Tags should be concise and help future searches.",
+    "3. Do not repeat tags or invent meaningless ones.",
+    "4. Write a note under 2000 characters summarizing the topic, conclusions, outstanding items and next steps.",
+    "5. Do not fabricate details that are not present in the conversation.",
+    "6. Return only valid JSON, no Markdown fences, no extra commentary.",
+    "",
+    "Return shape:",
+    "{",
+    "  ", "tags", ": [", "tag1", ", ", "tag2", "],",
+    "  ", "note", ": ", "session note", ",",
+    "}"
+  ].join(NEWLINE)
+});
+
+// Resolve the locale key from the active dictionary. Falls back to zh.
+export function clipboardPromptFor(dictionaries, language) {
+  const available = dictionaries ? Object.keys(dictionaries) : [];
+  const key = available.includes(language) ? language
+    : available.includes("zh") ? "zh"
+    : available[0] || "zh";
+  return { key, text: CLIPBOARD_PROMPTS[key] ?? CLIPBOARD_PROMPTS.zh };
 }

package/lib/index.js

@@ -23,11 +23,11 @@
  * Optional live-writer/coordinator internals are guarded before mutation.
  */
 import { lstat, mkdir, readdir, realpath, rename, rm } from "node:fs/promises";
-import { dirname, join } from "node:path";
+import { basename, dirname, join } from "node:path";
 import { randomBytes } from "node:crypto";
 import { createAnnotationStore } from "./annotation-store.js";
 import { createDshAdapter, normalizeSessionHeader } from "./compat/dsh-adapter.js";
-import { ARTIFACT_NAMES, assertSessionId, assertSessionDirectory, encodeSessionSegment, readSessionHeader, readSessionFile, writeTempFile, replaceSessionFile } from "./session-files.js";
+import { ARTIFACT_NAMES, assertSessionId, assertSessionDirectory, encodeSessionSegment, parseSessionGenerationFromFilename, readSessionHeader, readSessionFile, writeTempFile, replaceSessionFile } from "./session-files.js";
 
 export const name = "dsh-session-manager";
 
@@ -411,6 +411,17 @@ ${rest}`, "utf8");
           throw new Error("持久化后端无法定位会话工件路径");
         }
         const newPath = newLocation.path;
+        // Keep the stored generation in sync with the filename selected by
+        // persistence.locate(). persistence.locate() always returns the
+        // SESSION_FORMAT_VERSION path (DSH 0.1.7+ ships v4 by default),
+        // regardless of oldHeader.version. Without the stamp below, the
+        // rewritten file is `session.v4.jsonl.zstd` carrying a v0/v3 header,
+        // which JsonlSessionPersistence.listArtifacts rejects on the next
+        // startup (`session generation filename identifies v4, but its
+        // header identifies v0`), breaking persistence.list() and crashing
+        // the workspace registry while it boots. (issue #18)
+        const expectedVersion = parseSessionGenerationFromFilename(basename(newPath));
+        if (typeof expectedVersion === "number") newHeader.version = expectedVersion;
         const root = typeof dshHomePath === "function" ? dshHomePath("sessions") : undefined;
         await assertSessionDirectory(root, dirname(oldPath), sessionId);
         await assertSessionDirectory(root, dirname(newPath), sessionId, { allowMissing: true });