fix(lark): deliver App messages from original private conversations

docs/architecture/rfcs/app-conversation-and-async-inbox-v0.md

@@ -154,6 +154,46 @@ Packaged browser evidence qualifies the local UI path separately. Live Lark
 return delivery, installed service adoption, split-root composition and
 Goal-less project peer handoff remain open; this checkpoint does not close them.
 
+The existing Lark reply workers now also project committed App-origin user and
+answer messages from that exact owner-private Session. Each provider outbox
+retains the canonical message identity, frozen text and delivery attempt. The
+original Lark input supplies address provenance only: projection never readmits
+it or starts model work. Before every write and receipt recovery, Core rechecks
+the Session, current binding and audience, and the provider verifies the original
+source. Posts identify the App speaker, stay in canonical append order within
+that outbox and retain complete long text through the existing rich-post parts.
+Distinct canonical messages and parts get distinct provider idempotency keys,
+including when their text is identical; callers without that identity retain
+their existing receipt digest. Known attempts are read back before continuation;
+a started write without a provider locator stays unconfirmed after restart.
+Revocation or changed identity prevents further delivery. Disable the existing
+connection to revoke it; rollback must preserve its private delivery journals.
+
+Use **Continue original conversation** above to reach this path. Ordinary App
+steward history is still a separate entry; it is not silently merged or retargeted.
+This checkpoint covers text projection, not reply/thread relation mirroring,
+attachment transfer, retroactive insertion into Lark chronology or all-channel
+broadcast. An App attachment produces an explicit unsynchronized notice and a
+partial outbox, never a full-delivery claim. Synthetic real-HTTP/Core journeys
+cover repeated inputs, long text, restart, missing/unknown receipts, revocation
+and provider replacement. Installed provider readback remains a separate gate.
+
+The implementation decision follows primary-source comparisons, not a claim
+that shared context already guarantees two-way delivery:
+
+- [OpenClaw history delta](https://github.com/openclaw/openclaw/blob/b3ba4d622638ac23f64c0cec415bf5bfe81f5cf5/src/gateway/server-methods/chat-history-delta.ts)
+  resets incomplete/scoped projections; its
+  [send retry owner](https://github.com/openclaw/openclaw/blob/b3ba4d622638ac23f64c0cec415bf5bfe81f5cf5/src/gateway/server-methods/chat-send-retry.ts)
+  reconciles already-started execution. These inform recovery, not a Lark bridge.
+- [nanobot unified session keys](https://github.com/HKUDS/nanobot/blob/9e9e22d3319678178dba05a76f0ea61064b78ee8/nanobot/session/keys.py)
+  and [last-destination routing](https://github.com/HKUDS/nanobot/blob/9e9e22d3319678178dba05a76f0ea61064b78ee8/nanobot/agent/loop.py)
+  share context; a last destination does not replicate a transcript to every surface.
+- [mautrix message mappings](https://github.com/mautrix/go/blob/a4237e5b9f8f0e9b35c81d5310615eb70c202cea/bridgev2/database/message.go)
+  and [portal echo handling](https://github.com/mautrix/go/blob/a4237e5b9f8f0e9b35c81d5310615eb70c202cea/bridgev2/portal.go)
+  motivate stable transaction/part identities and echo suppression. A Matrix
+  homeserver/bridge would add separate infrastructure and identity ownership.
+  LoopX borrows the design, not MPL-2.0 code; the other two snapshots are MIT.
+
 Synthetic product previews: [desktop](../../assets/personal-workspace/private-project-conversations.png),
 [narrow](../../assets/personal-workspace/private-project-conversations-narrow.png),
 [revoked workspace](../../assets/personal-workspace/private-project-workspace-revoked.png).

docs/architecture/rfcs/app-conversation-and-async-inbox-v0.zh-CN.md

@@ -109,6 +109,36 @@ Web 来源，通过已有 Core 队列及新鲜来源准入。连接列表独立
 真实 Lark 结果回传、安装态服务采用、分离运行根组合及无 Goal 项目的 peer 交办
 仍未关闭，不由此检查点代替。
 
+现有 Lark 回复 worker 现在也投递该本人私聊确切 Session 中已提交的 App 用户消息和
+回答。每份 provider outbox 保留 canonical 消息身份、冻结正文和发送尝试。原 Lark
+输入只提供地址来源证据；投影不会重新受理它或启动模型工作。每次发送和恢复回执前，
+Core 重查 Session、当前绑定与受众，provider 核验原来源。消息标明 App 发言者，在
+该 outbox 内按 canonical 追加顺序发送，长正文沿用既有富文本分段，保留完整内容。
+即使正文相同，不同 canonical 消息及分段也使用不同 provider 幂等键；未提供该身份
+的既有调用者保持原回执 digest。恢复先读回已知发送，再继续未发送部分；已经开始但
+没有 provider locator 的写入，重启后仍保留未确认。撤权或身份变化阻止后续投递。
+关闭既有连接即可撤销授权；回滚须保留私有投递 journal。
+
+通过上述 **继续原会话** 使用该路径。普通 App 管家历史仍是独立入口，不会被静默
+合并或重定向。本检查点覆盖文本投影，不覆盖回复/话题关系镜像、附件传输、向 Lark
+历史回插旧时间或全渠道广播。App 附件会产生明确的未同步提示，并保留 partial
+outbox，不声明完整送达。真实 HTTP/Core 配合合成 provider 的旅程覆盖重复输入、
+长正文、重启、回执缺失/未知、撤权及 App 身份替换；安装态 provider 读回另行验收。
+
+实现判断来自一手源码比较，不把共享上下文视为双向送达已经成立：
+
+- [OpenClaw 历史增量](https://github.com/openclaw/openclaw/blob/b3ba4d622638ac23f64c0cec415bf5bfe81f5cf5/src/gateway/server-methods/chat-history-delta.ts)
+  对不完整/受限投影要求重置；其
+  [发送重试 owner](https://github.com/openclaw/openclaw/blob/b3ba4d622638ac23f64c0cec415bf5bfe81f5cf5/src/gateway/server-methods/chat-send-retry.ts)
+  对已经开始的执行先对账。这些机制用于恢复设计，不代表已有 Lark 桥。
+- [nanobot 统一会话键](https://github.com/HKUDS/nanobot/blob/9e9e22d3319678178dba05a76f0ea61064b78ee8/nanobot/session/keys.py)
+  和[最后目的地路由](https://github.com/HKUDS/nanobot/blob/9e9e22d3319678178dba05a76f0ea61064b78ee8/nanobot/agent/loop.py)
+  可以共享上下文，但最后目的地不等于向每个界面复制完整记录。
+- [mautrix 消息映射](https://github.com/mautrix/go/blob/a4237e5b9f8f0e9b35c81d5310615eb70c202cea/bridgev2/database/message.go)
+  及 [portal 回声处理](https://github.com/mautrix/go/blob/a4237e5b9f8f0e9b35c81d5310615eb70c202cea/bridgev2/portal.go)
+  支持稳定事务/分段身份和回声抑制的设计判断。采用 Matrix homeserver/桥还会增加
+  独立基础设施与身份 owner；LoopX 借鉴设计，不复制 MPL-2.0 代码，另两个快照为 MIT。
+
 ![合成私聊工作区设置](../../assets/personal-workspace/private-project-conversations.png)
 ![窄屏私聊设置](../../assets/personal-workspace/private-project-conversations-narrow.png)
 ![工作区撤权读回](../../assets/personal-workspace/private-project-workspace-revoked.png)

loopx/extensions/lark/inbox_reply.py

@@ -318,6 +318,7 @@ def _deliver_lark_inbox_outbound(
     text: str,
     content_format: str = "text",
     reply_placement: str | None = None,
+    delivery_id: str | None = None,
     execute: bool = False,
     provider_preflight: bool = False,
     runner: CommandRunner = _default_runner,
@@ -412,12 +413,16 @@ def _deliver_lark_inbox_outbound(
         )
         else "source_thread"
     )
+    if delivery_id is not None and (not isinstance(delivery_id, str)
+            or not re.fullmatch(r"[A-Za-z0-9][A-Za-z0-9._:/-]{0,159}", delivery_id)):
+        raise ValueError("invalid native delivery identity")
     digest = hashlib.sha256(
         json.dumps(
             {
                 "message_id": source_message_id,
                 "placement": placement,
                 "text": reply_text,
+                **({"delivery_id": delivery_id} if delivery_id is not None else {}),
                 **({"content_format": "markdown"} if markdown else {}),
                 **({"attachment_keys": attachment_keys} if attachment_keys else {}),
             },
@@ -587,6 +592,7 @@ def verify_source() -> tuple[bool, bool]:
                         "source_membership_verifier": source_membership_verifier,
                         "source_authority_verifier": source_authority_verifier,
                         **({"reply_placement": reply_placement} if reply_placement is not None else {}),
+                        **({"delivery_id": delivery_id} if delivery_id is not None else {}),
                         "finalize_reactions": finalize_reactions,
                         "attachment_keys": attachment_keys, "attachment_names": attachment_names})
     guidance = None
@@ -810,6 +816,7 @@ def reply_lark_event_inbox(
     text: str,
     content_format: str = "text",
     reply_placement: str | None = None,
+    delivery_id: str | None = None,
     execute: bool = False,
     provider_preflight: bool = False,
     runner: CommandRunner = _default_runner,
@@ -853,6 +860,7 @@ def reply_lark_event_inbox(
         text=text,
         content_format=content_format,
         reply_placement=reply_placement,
+        delivery_id=delivery_id,
         execute=execute,
         provider_preflight=provider_preflight,
         runner=runner,
@@ -878,6 +886,7 @@ def verify_lark_inbox_reply(
     config_path: str | Path,
     message_id: str,
     reply_placement: str | None = None,
+    delivery_id: str | None = None,
     text: str,
     attempt: Mapping[str, Any],
     runner: CommandRunner = _default_runner,
@@ -922,14 +931,15 @@ def verify_lark_inbox_reply(
     if batch:
         original = reply_lark_event_inbox(project=project, config_path=config_path,
             message_id=message_id, text=text, content_format="markdown", attachment_keys=attachment_keys,
-            reply_placement=reply_placement)
+            reply_placement=reply_placement, delivery_id=delivery_id)
         receipt = original["idempotency_key"]
         return verify_post_parts(receipt=receipt, intent_digest=_intent_digest(profile, chat_id, receipt),
             attempt=attempt, resume=resume_parts, arguments={"project": project,
                 "config_path": config_path, "message_id": message_id, "runner": runner,
                 "source_membership_verifier": source_membership_verifier,
                 "source_authority_verifier": source_authority_verifier,
                 **({"reply_placement": reply_placement} if reply_placement is not None else {}),
+                **({"delivery_id": delivery_id} if delivery_id is not None else {}),
                 "before_send": before_send, "finalize_reactions": finalize_reactions,
                 "text": normalize_lark_outbound_text(text, limit=None, preserve_format=True),
                 "attachment_keys": attachment_keys, "attachment_names": attachment_names})
@@ -947,6 +957,7 @@ def verify_lark_inbox_reply(
             text=candidate_text,
             content_format=content_format,
             reply_placement=reply_placement,
+            delivery_id=delivery_id,
             execute=False,
             runner=runner,
             attachment_keys=attachment_keys,

loopx/extensions/lark/post_delivery.py

@@ -7,6 +7,7 @@
 from __future__ import annotations
 
 import json
+import hashlib
 import re
 from pathlib import Path
 from typing import Any
@@ -19,6 +20,12 @@
 POST_PARTS_LOCATOR = re.compile(r"^lark-parts:([a-f0-9]{64})$")
 
 
+def _part_identity(arguments, index):
+    identity = arguments.get("delivery_id")
+    return ({"delivery_id": hashlib.sha256(f"{identity}:part:{index}".encode()).hexdigest()}
+            if identity is not None else {})
+
+
 def _path(project, config_path, intent_digest):
     config = load_lark_event_inbox_config(project=project, config_path=config_path)
     return (
@@ -124,6 +131,7 @@ def options(index):
                 "source_authority_verifier": arguments.get("source_authority_verifier"),
                 **({"reply_placement": arguments["reply_placement"]}
                    if arguments.get("reply_placement") is not None else {}),
+                **_part_identity(arguments, index),
                 "before_send": (lambda _part_intent: outer_hook(intent_digest))
                 if outer_hook
                 else None,
@@ -249,6 +257,7 @@ def verify_post_parts(*, receipt, intent_digest, attempt, arguments, resume=Fals
             source_authority_verifier=arguments.get("source_authority_verifier"),
             **({"reply_placement": arguments["reply_placement"]}
                if arguments.get("reply_placement") is not None else {}),
+            **_part_identity(arguments, index),
             finalize_reactions=False,
             attachment_keys=arguments.get("attachment_keys", ()) if final else (),
             attachment_names=arguments.get("attachment_names", ()) if final else (),