Fix/openclaw addmsg

[chenjw]

Description

1. 提交到session的内容，user和assistant分开，去掉了记忆召回部分，工具part修复input的问题。
2. 修复agfs session文件中写入文件，会变成追加写的问题。
3. 去掉了get_content返回的历史abstract 列表。

Related Issue

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

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

Screenshots (if applicable)

Additional Notes

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 4 🔵🔵🔵🔵⚪

🏅 Score: 55

🧪 No relevant tests

🔒 No security concerns identified

✅ No TODO sections

🔀 Multiple PR themes Sub-PR theme: Fix localfs write truncation logic Relevant files: crates/ragfs/src/plugins/localfs/mod.rs Sub-PR theme: OpenClaw plugin addmsg improvements Relevant files: examples/openclaw-plugin/client.ts examples/openclaw-plugin/context-engine.ts examples/openclaw-plugin/index.ts examples/openclaw-plugin/text-utils.ts bot/scripts/install_local_openclaw_plugin.sh Sub-PR theme: Session message filtering and extraction fixes Relevant files: openviking/session/session.py

⚡ Recommended focus areas for review API Breaking Change Removed 'pre_archive_abstracts' from get_session_context response, breaking existing clients relying on this field. # 不再返回 pre_archive_abstracts（只保留 latest_archive_overview） included_pre_archive_abstracts: List[Dict[str, str]] = [] pre_archive_tokens = 0 archive_tokens = latest_archive_tokens + pre_archive_tokens included_archives = len(included_pre_archive_abstracts) dropped_archives = max( 0, context["total_archives"] - context["failed_archives"] - included_archives ) return { "latest_archive_overview": ( latest_archive["overview"] if include_latest_overview else "" ), # 不再返回 pre_archive_abstracts Incorrect Truncation Logic Truncation flag check uses exact match instead of bitwise check, failing to truncate when only Truncate flag is set or other flags are combined. // Determine if we should truncate based on flags let should_truncate = matches!(flags, WriteFlag::Create | WriteFlag::Truncate); // Open or create file with truncate support let mut file = fs::OpenOptions::new() .write(true) .create(true) .truncate(should_truncate) .open(&local_path) .map_err(|e| Error::plugin(format!("failed to open file: {}", e)))?; Private Attribute Access Accessing private attributes _uri_to_path and agfs of viking_fs, breaking encapsulation and risking future compatibility issues. # Workaround: RAGFSBindingClient.write is not truncating, so delete file first # Use AGFS directly to avoid lock issues with viking_fs.rm try: path = viking_fs._uri_to_path(f"{self._session_uri}/messages.jsonl", ctx=self.ctx) viking_fs.agfs.rm(path) except Exception as e: logger.warning(f"[_write_to_agfs_async] Failed to delete messages.jsonl: {e}")

[github-actions]

PR Code Suggestions ✨

Explore these optional code suggestions:

Category	Suggestion	Impact
Possible issue	Add error handling for archive timestamp retrieval Add error handling around archive reading and timestamp parsing to avoid breaking commit or context retrieval when archives are missing/corrupted. Return None on failure, which will skip filtering and use all messages. openviking/session/session.py [185-201] async def _get_latest_archive_last_msg_time(self) -> Optional[float]: """获取上一次归档的最后一条消息的时间戳（毫秒），用于过滤需要提取记忆的消息。""" # 使用 compression_index - 1 获取上一次归档的索引 if self._compression.compression_index <= 1: return None # 获取上一次归档的目录 archive_uri = ( f"{self._session_uri}/history/archive_{self._compression.compression_index - 1:03d}" ) - messages = await self._read_archive_messages(archive_uri) - if messages and messages[-1].created_at: - # 解析 ISO 时间戳为毫秒 - from dateutil import parser + try: + messages = await self._read_archive_messages(archive_uri) + if messages and messages[-1].created_at: + # 解析 ISO 时间戳为毫秒 + from dateutil import parser - dt = parser.parse(messages[-1].created_at) - return dt.timestamp() * 1000 + dt = parser.parse(messages[-1].created_at) + return dt.timestamp() * 1000 + except Exception as e: + logger.warning(f"[_get_latest_archive_last_msg_time] Failed to read archive timestamp: {e}") return None Suggestion importance[1-10]: 7 __ Why: Adds error handling around archive reading and timestamp parsing to prevent breaking commit/context retrieval when archives are missing/corrupted, which improves robustness.	Medium

[qin-ctx]

Thanks for the PR. I left 4 blocking comments.

The main issues are: tool-call history no longer round-trips through afterTurn -> commit -> assemble; get_session_context() silently removes pre_archive_abstracts even though the public contract and consumers still require it; archive summaries are now generated from the filtered extraction subset instead of the actual archived message set; and compacted system-message extraction currently captures from the first ] instead of the last one.

CI is also failing on this branch: lint reports unused locals in session.py, and API integration tests already fail on the missing pre_archive_abstracts contract.

[qin-ctx]

[Bug] (blocking) This persists tool results as a user message containing a tool part, but the assemble path only reconstructs tool calls/results from tool parts on assistant messages and only when tool_id is present. extractNewTurnMessages() also does not persist the original tool call id, so a tool turn cannot round-trip through afterTurn -> commit -> assemble: the next assemble loses the toolUse/toolResult structure entirely. Please persist a stable tool identifier and store tool-call/tool-result data in a shape that the reader can actually reconstruct.

[qin-ctx]

[Bug] (blocking) messages_for_extraction is passed straight into _run_memory_extraction(), but that method also generates the archive summary and writes .abstract.md / .overview.md. That means the timestamp filter now changes the archive metadata as well, not just the memory-extraction input. If any archived messages are filtered out here, the summary no longer describes the actual messages.jsonl that was written for this archive. The filter should apply only to the extraction input, not to archive rendering.

[qin-ctx]

[Bug] (blocking) Removing pre_archive_abstracts here is a breaking API change, not an internal cleanup. The OpenClaw plugin type still requires it, the CLI renderer still expects it, the docs still publish it, and API/session tests still assert that it exists. CI is already failing in scenarios/sessions/test_session_commit.py for this exact reason. If the intent is to retire archive-index support, that needs a coordinated migration; otherwise this field needs to remain in the response.

[qin-ctx]

[Bug] (blocking) The implementation does not match the comment here. text.match(/\]\s*(.+)$/) matches from the first ], not the last one, so an input like System: [t1] Compacted ... [t2] real content becomes Compacted ... [t2] real content instead of just real content. That writes system noise back into the session and pollutes downstream extraction.