← 返回 Skills 市场
123
总下载
0
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install conversations
功能描述
从 OpenClaw sessions/*.jsonl 导入对话历史到本地 SQLite，配合 FTS5 全文搜索。成为真正的记忆库，支持语义化查询历史对话，让整理和回顾更高效。装好后跑一次 import，之后随时可以 query。
使用说明 (SKILL.md)
conversations-universal
对话历史的本地记忆库。
使用方式
第一次使用
- 运行导入脚本，把历史 session 读入本地库：
python {baseDir}/scripts/import_sessions.py
- 之后每次想回顾历史，直接用 query 脚本：
python {baseDir}/scripts/query_conversations.py \x3C搜索内容>
查询示例
# 查询某天的讨论
python scripts/query_conversations.py "我们讨论过内存泄漏"
# 查询某个话题
python scripts/query_conversations.py "OpenClaw 配置"
自动导入
建议配合 cron 定期导入新 session：
openclaw cron add \
--name "对话历史导入" \
--cron "0 */6 * * *" \
--session isolated \
--message "python {baseDir}/scripts/import_sessions.py" \
--announce
工作原理
- 存储：SQLite + FTS5 全文搜索，零外部依赖
- 导入：读取
agents/main/sessions/*.jsonl，解析 role/content，写入 chunks 表 - 搜索：FTS5 MATCH，支持自然语言查询
环境变量
| 变量 | 默认值 | 说明 |
|---|---|---|
| OPENCLAW_WORKSPACE | 自动推断 | session 文件所在目录 |
| CONVERSATIONS_DB | {workspace}/conversations.db |
数据库路径 |
| DRY_RUN | false | 试运行，不写入 |
依赖
- Python 3.8+（内置 sqlite3，FTS5 需要 3.9+）
- OpenClaw sessions 目录
与 conversations 1.0 的区别
| conversations 1.0 | conversations-universal | |
|---|---|---|
| 搜索方式 | Ollama 向量搜索 | SQLite FTS5 |
| 外部依赖 | 需要 Ollama | 纯 Python |
| 搜索质量 | 高（语义） | 中（关键词） |
| 适用场景 | 已有 Ollama 的用户 | 所有人 |
安全使用建议
This skill is internally coherent and implements what it claims: it reads your OpenClaw session JSONL files and builds a local SQLite database for keyword/FTS search. Before installing: (1) be aware it will aggregate your chat contents into a local DB (no network exfiltration in the code), so don’t install if you don’t want local copies of conversations; (2) run import with DRY_RUN first to verify paths and parsing; (3) ensure your Python+SQLite build supports FTS5 if you want full-text search (otherwise it falls back to LIKE); (4) back up or inspect any existing conversations.db before running to avoid accidental overwrites; and (5) note the SKILL.md documents OPENCLAW_WORKSPACE / CONVERSATIONS_DB / DRY_RUN as config vars even though the registry didn't mark required env vars.
功能分析
Type: OpenClaw Skill
Name: conversations
Version: 2.0.0
The skill bundle is a utility for indexing and searching local OpenClaw conversation history using SQLite and FTS5. The scripts (import_sessions.py and query_conversations.py) operate entirely on local files within the OpenClaw workspace, contain no network activity, and use only standard Python libraries without any suspicious execution patterns or data exfiltration logic.
能力评估
Purpose & Capability
Name/description claim: import OpenClaw sessions into local SQLite with FTS5 and provide query tooling. The included scripts implement exactly that: they locate sessions under an OpenClaw workspace, parse JSONL message records, deduplicate by MD5, create a local conversations.db, and provide LIKE/FTS5 search. No unexpected external services, binaries, or credentials are required.
Instruction Scope
SKILL.md and the scripts instruct running the bundled Python scripts and optionally scheduling via openclaw cron. They access files under agents/main/sessions (or the workspace path) and write a local conversations.db. This is in-scope for the stated purpose. Minor discrepancy: SKILL.md documents environment variables (OPENCLAW_WORKSPACE, CONVERSATIONS_DB, DRY_RUN) that are used by the scripts but the registry did not list any required env vars — these are configuration vars (not secrets) and optional, but the documentation/registry mismatch is worth noting.
Install Mechanism
Instruction-only with bundled Python scripts; there is no network download, no package install, and no install spec. This minimizes install-time risk. Scripts rely on Python 3.8+/3.9+ and built-in sqlite3; no third-party packages or external URLs are fetched.
Credentials
The skill requests no credentials or privileged config paths. It does read session files from the OpenClaw workspace (which contain chat history) and writes a local DB. The referenced environment variables are for configuration only (workspace path, DB path, dry-run) and do not contain secrets; still, the skill will aggregate potentially sensitive chat content locally—this is expected but worth informing the user.
Persistence & Privilege
always is false and the skill does not request persistent platform-level privileges or modify other skills. It creates/updates a conversations.db in the workspace but does not alter agent configurations beyond scheduling guidance (cron example).
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install conversations - 安装完成后，直接呼叫该 Skill 的名称或使用
/conversations触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v2.0.0
这是一个更普适的版本，任何装了 OpenClaw 的人都能用。
搜索方式：SQLite FTS5 全文搜索
外部依赖：纯 Python，零依赖
搜索质量：中（关键词匹配）
安装难度：低（装好 Python 就能跑）
v1.0.1
- 更新描述和文档，强调支持 OpenClaw sessions 文件及本地 SQLite 数据库导入与查询
- 增加环境变量说明（OPENCLAW_WORKSPACE、CONVERSATIONS_DB），支持自定义数据库路径
- 优化数据库结构和脚本说明，明确 chunks/embeddings 表作用
- 完善导入和查询脚本用法说明，增加示例与参数细节
- 建议工作流程更具体，包括定期导入和多样化查询方式
v1.0.0
Initial release of the conversations skill for managing conversation history.
- Import past conversations from sessions/*.jsonl into a centralized conversations.db, with deduplication and assistant/user filtering.
- Query historical conversations by stats, recent entries, random samples, keyword search, session, or timestamp.
- Provides command-line scripts for both import (import_sessions.py) and flexible querying (query_conversations.py).
- Stores all conversation and embedding data in workspace/conversations.db for further analysis or retrieval.
元数据
常见问题
conversations 是什么？
从 OpenClaw sessions/*.jsonl 导入对话历史到本地 SQLite，配合 FTS5 全文搜索。成为真正的记忆库，支持语义化查询历史对话，让整理和回顾更高效。装好后跑一次 import，之后随时可以 query。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 123 次。
如何安装 conversations？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install conversations」即可一键安装，无需额外配置。
conversations 是免费的吗？
是的，conversations 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
conversations 支持哪些平台？
conversations 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 conversations？
由 shu0yu（@shu0yu）开发并维护，当前版本 v2.0.0。
推荐 Skills