@yeaft/webchat-agent 0.1.992 → 0.1.994

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@yeaft/webchat-agent",
-  "version": "0.1.992",
+  "version": "0.1.994",
   "description": "Remote agent for Yeaft WebChat — connects worker machines to the central server",
   "main": "index.js",
   "type": "module",

package/yeaft/dream/prompts/create.md

@@ -9,7 +9,9 @@ Source session conversations:
 {{siblingsBlock}}
 Task:
 1. Write memory.md from scratch with a reasonable section structure.
-2. Write summary.md (1–3 sentences).
+2. Prioritize reusable experience: workflows, preferences, pitfalls, corrections, project conventions, review/merge/tag lessons, and rules that should change future execution.
+3. Keep current PR/review/blocker detail only when it is still actionable.
+4. Write summary.md (1–3 sentences) with reusable lessons before current state.
 
 Reply with strict JSON of the shape:
 { "memory_md": "...", "summary_md": "..." }
@@ -25,7 +27,9 @@ Scope path: {{target}}   (must be ≤2 levels)
 {{siblingsBlock}}
 任务：
 1. 从零编写 memory.md，并使用合理的章节结构。
-2. 编写 summary.md（1–3 句）。
+2. 优先保留可复用经验：workflow、preference、pitfall、correction、project convention、review/merge/tag 教训，以及会改变后续执行方式的规则。
+3. 当前 PR、review、阻塞细节只有仍可执行时才保留。
+4. 编写 summary.md（1–3 句），先写可复用经验，再写当前状态。
 
 只回复严格 JSON，结构如下：
 { "memory_md": "...", "summary_md": "..." }

package/yeaft/dream/prompts/extract-session.md

@@ -8,12 +8,14 @@ The target session id is provided as `{{sessionId}}`.
 ## What to extract for `sessions/<id>` scope
 
 - **purpose** — what this session exists to do, its charter or mission
-- **members** — users, VPs, roles, and what each is responsible for
-- **conventions** — how the session works: rituals, naming, languages, review flow
+- **members** — users, session members, roles, and what each is responsible for
+- **workflow** — reusable ways this session should work: review flow, merge/tag ritual, handoff order, testing expectations
+- **preferences** — user or project preferences that affect future execution in this session
+- **pitfalls and corrections** — mistakes, rejected approaches, and user corrections that should prevent repeated errors
 - **shared decisions** — durable agreements the session should remember
 - **shared context** — domain or project knowledge the whole session relies on
-- **current state** — recent PRs, reviews, blockers, next steps, or release status that matter now
-- **lessons** — corrections or decisions that should guide future turns in this session
+- **current state** — recent PRs, reviews, blockers, next steps, or release status that matter now, but only when still actionable
+- **lessons** — generalized experience that should guide future turns in this session
 
 ## What NOT to extract here
 
@@ -26,7 +28,9 @@ The target session id is provided as `{{sessionId}}`.
 
 Return a JSON array only. Each item must have `kind`, `body`, `tags`, `sourceMessages`, and `confidence`.
 
-Make the segment useful to a future session member who needs the big picture plus the latest actionable detail.
+Allowed `kind` values include `workflow`, `preference`, `lesson`, `pitfall`, `correction`, `project-convention`, `decision`, `goal`, `fact`, and `context`.
+
+Make the segment useful to a future session member who needs the big picture plus the latest actionable detail. Prefer stable, reusable memory over a chronology of completed actions. A completion record is useful only if it changes what future work should do next.
 
 <!-- lang:zh -->
 # Dream Extract — Session Scope
@@ -38,12 +42,14 @@ Make the segment useful to a future session member who needs the big picture plu
 ## `sessions/<id>` scope 应抽取什么
 
 - **purpose** — 这个 session 存在的目的、charter 或 mission
-- **members** — 用户、VP、角色，以及各自负责什么
-- **conventions** — session 的协作方式：流程、命名、语言、review 规则
+- **members** — 用户、session member、角色，以及各自负责什么
+- **workflow** — 之后可复用的工作方式：review 流程、merge/tag 节奏、交接顺序、测试要求
+- **preferences** — 会影响后续执行的用户偏好或项目偏好
+- **pitfalls and corrections** — 已犯错误、被拒绝的做法、用户纠偏，避免之后重复
 - **shared decisions** — session 之后应该记住的稳定决策
 - **shared context** — 整个 session 依赖的领域或项目信息
-- **current state** — 当前重要的 PR、review、阻塞、下一步或 release 状态
-- **lessons** — 未来 turn 应继续遵守的纠正和决策
+- **current state** — 当前重要的 PR、review、阻塞、下一步或 release 状态，但只保留仍会影响下一步的内容
+- **lessons** — 未来 turn 应继续遵守的规则化经验
 
 ## 不要在这里抽取什么
 
@@ -56,4 +62,6 @@ Make the segment useful to a future session member who needs the big picture plu
 
 只返回 JSON array。每一项必须包含 `kind`、`body`、`tags`、`sourceMessages` 和 `confidence`。
 
-segment 应帮助未来的 session member 快速理解大背景和最新可执行细节。
+允许的 `kind` 包括 `workflow`、`preference`、`lesson`、`pitfall`、`correction`、`project-convention`、`decision`、`goal`、`fact` 和 `context`。
+
+segment 应帮助未来的 session member 快速理解大背景和最新可执行细节。优先保留可复用经验，不要只写完成事项流水账；只有仍会改变下一步执行方式的完成记录才值得保留。

package/yeaft/dream/prompts/summarize-scope.md

@@ -8,7 +8,9 @@ The target scope is `{{scope}}` and contains `{{segmentCount}}` segments listed
 ## Goals
 
 - A future session member should grasp the gist of this scope from your summary alone, without reading all segments.
+- Prioritize reusable experience: workflows, project/user preferences, pitfalls, corrections, review/merge/tag lessons, and conventions that change future behavior.
 - Keep specifics that matter: names, numbers, decisions, durable views, current blockers, and next steps.
+- Do not turn the summary into a chronology of completed work. Mention completed items only when they affect future execution.
 - Stay faithful: do not invent facts not present in the segments. Do not soften decisions or opinions.
 - Write in the same language the segments are mostly in, unless the surrounding language directive says otherwise.
 
@@ -33,7 +35,9 @@ Return only the summary text.
 ## 目标
 
 - 未来的 session member 只读 summary，也应能理解这个 scope 的核心意思。
+- 优先保留可复用经验：工作流、项目/用户偏好、坑点、纠偏、review/merge/tag 教训，以及会改变后续行为的约定。
 - 保留重要具体信息：名字、数字、决策、稳定观点、当前阻塞和下一步。
+- 不要把 summary 写成完成事项流水账；已完成事项只有会影响后续执行时才写。
 - 忠于事实：不要编造 segments 中没有的事实。不要弱化已经明确的决策或观点。
 - 使用 segments 主要使用的语言；如果外层语言指令另有要求，以外层语言指令为准。
 

package/yeaft/dream/prompts/update.md

@@ -18,9 +18,11 @@ Recent session conversations:
 
 Task:
 - Extract from these conversations what is relevant to THIS scope.
+- Prefer reusable experience over chronology: workflows, preferences, pitfalls, corrections, project conventions, review/merge/tag lessons, and rules that should change future execution.
+- Keep current PR/review/blocker detail only when it is still actionable.
 - Integrate it into memory.md (reorganize sections if needed).
 - Drop stale or contradicted entries.
-- Rewrite summary.md (1–3 sentences).
+- Rewrite summary.md (1–3 sentences) so the summary highlights reusable lessons first, then current state.
 - The same conversations are being processed for OTHER scopes too.
   Only handle what is relevant here. Ignore the rest.
 
@@ -53,9 +55,11 @@ Scope: {{target}}
 
 任务：
 - 从这些对话中提取与 THIS scope 相关的内容。
+- 优先保留可复用经验，而不是流水账：workflow、preference、pitfall、correction、project convention、review/merge/tag 教训，以及会改变后续执行方式的规则。
+- 当前 PR、review、阻塞细节只有仍可执行时才保留。
 - 将内容整合进 memory.md（必要时重组章节）。
 - 删除过期或被推翻的条目。
-- 重写 summary.md（1–3 句）。
+- 重写 summary.md（1–3 句），summary 先突出可复用经验，再写当前状态。
 - 同一批对话也会被处理到 OTHER scopes。
   这里只处理与当前 scope 相关的内容，忽略其他内容。
 

package/yeaft/dream/segment-extract.js

@@ -17,7 +17,19 @@ const MAX_MESSAGES = 80;
 const MAX_BODY_CHARS = 1200;
 const MAX_SEGMENTS_PER_SCOPE = 64;
 const RECENT_MESSAGE_COUNT = 8;
-const VALID_KINDS = new Set(['fact', 'preference', 'decision', 'lesson', 'relation', 'goal', 'context']);
+const VALID_KINDS = new Set([
+  'fact',
+  'preference',
+  'decision',
+  'lesson',
+  'relation',
+  'goal',
+  'context',
+  'workflow',
+  'pitfall',
+  'correction',
+  'project-convention',
+]);
 
 /**
  * @param {{
@@ -61,7 +73,10 @@ export async function extractAndWriteMemorySegments(opts) {
     }
 
     const recent = scope === `sessions/${opts.sessionId}`
-      ? [buildRecentSegment({ scope, messages, now })]
+      ? [
+        buildRecentSegment({ scope, messages, now }),
+        buildRecentExperienceSegment({ scope, extracted, now }),
+      ].filter(Boolean)
       : [];
     if (extracted.length === 0 && recent.length === 0) continue;
 
@@ -184,8 +199,47 @@ function buildRecentSegment({ scope, messages, now }) {
   });
 }
 
+function buildRecentExperienceSegment({ scope, extracted, now }) {
+  const experienceSegments = extracted
+    .filter(isExperienceSegment)
+    .slice(-RECENT_MESSAGE_COUNT);
+  if (experienceSegments.length === 0) return null;
+
+  const lines = experienceSegments
+    .map(seg => oneLine(seg.body))
+    .filter(Boolean);
+  if (lines.length === 0) return null;
+
+  const sourceMessages = [...new Set(experienceSegments.flatMap(seg => (
+    Array.isArray(seg.sourceMessages) ? seg.sourceMessages.map(String).filter(Boolean) : []
+  )))];
+  const body = [
+    'Reusable session experience from the latest Dream pass:',
+    ...lines.map(line => `- ${line}`),
+  ].join('\n');
+
+  return makeSegment({
+    scope,
+    kind: 'lesson',
+    tags: ['recent', 'experience', 'workflow'],
+    sourceMessages,
+    createdAt: now,
+    updatedAt: now,
+    body,
+  });
+}
+
+function isExperienceSegment(seg) {
+  if (!seg) return false;
+  if (['workflow', 'preference', 'pitfall', 'correction', 'project-convention', 'lesson'].includes(seg.kind)) {
+    return true;
+  }
+  const tags = Array.isArray(seg.tags) ? seg.tags.map(tag => String(tag).toLowerCase()) : [];
+  return tags.some(tag => ['workflow', 'preference', 'pitfall', 'correction', 'project-convention', 'lesson', 'experience'].includes(tag));
+}
+
 function mergeSegments(existing, incoming) {
-  const incomingRecent = incoming.filter(isRecentSegment).slice(-1);
+  const incomingRecent = latestRecentByFamily(incoming.filter(isRecentSegment));
   const incomingPermanent = incoming.filter(seg => !isRecentSegment(seg));
   const byKey = new Map();
 
@@ -216,6 +270,21 @@ function segmentMergeKey(seg) {
   return `id:${seg.id}`;
 }
 
+function latestRecentByFamily(segments) {
+  const byFamily = new Map();
+  for (const seg of segments) {
+    byFamily.set(recentFamily(seg), seg);
+  }
+  return [...byFamily.values()];
+}
+
+function recentFamily(seg) {
+  const tags = Array.isArray(seg.tags) ? seg.tags.map(String) : [];
+  if (tags.includes('experience')) return 'experience';
+  if (tags.includes('current')) return 'current';
+  return 'recent';
+}
+
 function isRecentSegment(seg) {
   return Array.isArray(seg.tags) && seg.tags.includes('recent');
 }

package/yeaft/memory/segment.js

@@ -11,7 +11,7 @@
  *   ---
  *   id: seg_<8hex>
  *   scope: feature/auth
- *   kind: decision        # fact|preference|decision|lesson|relation|goal|context
+ *   kind: decision        # fact|preference|decision|lesson|relation|goal|context|workflow|pitfall|correction|project-convention
  *   tags: [auth, jwt]
  *   sourceMessages: [m_142, m_143]
  *   createdAt: 2026-04-29T10:11:12Z
@@ -49,6 +49,7 @@ import { createHash } from 'node:crypto';
 
 export const KIND_VALUES = new Set([
   'fact', 'preference', 'decision', 'lesson', 'relation', 'goal', 'context',
+  'workflow', 'pitfall', 'correction', 'project-convention',
 ]);
 
 const SCOPE_RE = /^(user|group\/[\w-]+(?:\/(?:user|vp\/[\w-]+|feature\/[\w-]+|topic\/[\w-]+(?:\/[\w-]+)?))?|sessions\/[\w-]+(?:\/(?:user|vp\/[\w-]+|feature\/[\w-]+|topic\/[\w-]+(?:\/[\w-]+)?))?|chat\/[\w-]+(?:\/vp\/[\w-]+)?|session\/[\w-]+(?:\/vp\/[\w-]+)?)$/;

package/yeaft/prompts.js

@@ -470,9 +470,6 @@ function renderActiveScope(activeScope, lang) {
     : '';
   if (session) lines.push(`session_id: ${session}`);
 
-  const sessionMember = firstNonEmptyString(activeScope.sessionMember, activeScope.vpId);
-  if (sessionMember) lines.push(`session_member: ${sessionMember}`);
-
   const membersLine = renderSessionMembersLine(activeScope.sessionMembers || activeScope.members);
   if (membersLine) lines.push(`session_members: ${membersLine}`);