refactor(opencode-plugin): migrate auto recall from chat.messages.transform to chat.message hook

[A0nameless0man]

Description

将记忆召回（auto recall）钩子从 experimental.chat.messages.transform 迁移到 chat.message，使用 synthetic: true 持久化注入，解决上下文漂移与重复搜索问题。

Related Issue

N/A

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Refactoring (no functional changes)
• Performance improvement
• Test update

Changes Made

• 移除 experimental.chat.messages.transform 钩子
• 移除死代码：RecallMessagePart、RecallWithParts 接口、extractLatestUserText、appendToLastTextPart 函数
• 新增 chat.message 钩子，使用 synthetic: true 持久化注入 <relevant-memories>
• 新增 runAutoRecall() 共享函数
• 新增 extractMessageText() 替代 extractLatestUserText()，适配 chat.message 的 input.parts 签名
• 净减少 28 行（-82/+54）

Testing

• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• I have tested this on the following platforms:
  • Linux
  • macOS
  • Windows

Checklist

• My code follows the project's coding style
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• Any dependent changes have been merged and published

Additional Notes

问题根因

experimental.chat.messages.transform 是无状态钩子——修改仅作用于内存中的消息拷贝，不写回会话存储。每次 LLM API 调用（含工具调用续接）都会重新触发，导致：

1. 上下文不一致 — 每次工具调用回来，模型看到的记忆集合不同
2. 前序记忆丢失 — 上次注入的记忆在下一次调用时消失
3. KV-cache 命中差 — 消息内容持续变化，服务端缓存无效
4. 冗余 API 调用 — 同一轮对话反复调用 OpenViking search

方案

chat.message 钩子在用户消息创建时触发一次，output.parts 的修改持久化到会话历史。synthetic: true 的 part 在 TUI 中不可见，但对 LLM 作为 synthetic context 发送。

效果对比

# 旧方案
用户输入 → API#1: 搜索→A,B,C → 工具调用 → API#2: 重新搜索→A,D,E → ...

# 新方案
用户输入 → chat.message: 搜索→插入synthetic part(持久化)
         → API#1: 含A,B,C → 工具调用 → API#2: 仍含A,B,C ✓

关键约束

• Part ID 前缀：prt-ov-recall-{ts}-{random}，必须 prt- 开头（OpenCode schema 校验）
• 幂等性：保留 <relevant-memories> 标记检测，防止重加载会话后重复注入
• 兼容性：旧会话不含 synthetic part，chat.message 仅对新消息触发

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

🏅 Score: 60

🧪 No relevant tests

⏱️ Estimated effort to review: 3 🔵🔵🔵⚪⚪

🔒 No security concerns identified

✅ No TODO sections

🔀 No multiple PR themes

⚡ Recommended focus areas for review NameError in load_latest_conversation due to missing time_str capture The load_latest_conversation function removed the line that captures time_str from the regex match when processing a new message section, but still uses time_str when parsing timestamps for previous messages, leading to a NameError. match = re.match(r"^## \[([0-9:]+)\] (.+)$", line) if match: # Save previous message if current_speaker and current_content: # Try to parse timestamp, or use current time try: # We don't have the date, just use current time ts = time.time() except ValueError: ts = time.time() messages.append( ChatMessage( channel_id=current_speaker, content="\n".join(current_content).strip(), is_user=(current_speaker == "admin"), timestamp=ts, ) ) # Start new message current_speaker = match.group(2) current_content = [] elif line and not line.startswith("# "): # Content line if current_speaker is not None: current_content.append(line) # Save the last message if current_speaker and current_content: try: ts = time.time() except ValueError: ts = time.time() messages.append( ChatMessage( channel_id=current_speaker, content="\n".join(current_content).strip(), is_user=(current_speaker == "admin"), timestamp=ts, ) ) logger.info( f"Loaded {len(messages)} messages from latest conversation, session_id={session_id}" ) Mismatched command line argument for werewolf server The start script changed from passing --bot-url to --bot-port, but the werewolf server still expects --vikingbot-url as a command line argument, causing a startup failure. args.server_host, "--port", str(args.server_port), "--with-bot", "--bot-port", str(urlparse(args.vikingbot_url).port or 18790),

[github-actions]

PR Code Suggestions ✨

No code suggestions found for the PR.

[yangxinxin-7]

LGTM！这是一个很棒的重构：

1. 解决了无状态钩子导致的上下文漂移和重复搜索问题
2. 代码更简洁，删除了不必要的死代码
3. 幂等性处理得当
4. CI 全部通过