agent-shared-workspace 0.1.0__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- agent_shared_workspace-0.1.0/LICENSE +21 -0
- agent_shared_workspace-0.1.0/PKG-INFO +218 -0
- agent_shared_workspace-0.1.0/README.en.md +194 -0
- agent_shared_workspace-0.1.0/README.md +232 -0
- agent_shared_workspace-0.1.0/pyproject.toml +49 -0
- agent_shared_workspace-0.1.0/setup.cfg +4 -0
- agent_shared_workspace-0.1.0/src/agent_shared_workspace/__init__.py +3 -0
- agent_shared_workspace-0.1.0/src/agent_shared_workspace/cli.py +1094 -0
- agent_shared_workspace-0.1.0/src/agent_shared_workspace/mcp_server.py +385 -0
- agent_shared_workspace-0.1.0/src/agent_shared_workspace.egg-info/PKG-INFO +218 -0
- agent_shared_workspace-0.1.0/src/agent_shared_workspace.egg-info/SOURCES.txt +14 -0
- agent_shared_workspace-0.1.0/src/agent_shared_workspace.egg-info/dependency_links.txt +1 -0
- agent_shared_workspace-0.1.0/src/agent_shared_workspace.egg-info/entry_points.txt +3 -0
- agent_shared_workspace-0.1.0/src/agent_shared_workspace.egg-info/top_level.txt +1 -0
- agent_shared_workspace-0.1.0/tests/test_cli.py +337 -0
- agent_shared_workspace-0.1.0/tests/test_mcp_and_packaging.py +151 -0
|
@@ -0,0 +1,21 @@
|
|
|
1
|
+
MIT License
|
|
2
|
+
|
|
3
|
+
Copyright (c) 2026 MAHONGHAO1
|
|
4
|
+
|
|
5
|
+
Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
6
|
+
of this software and associated documentation files (the "Software"), to deal
|
|
7
|
+
in the Software without restriction, including without limitation the rights
|
|
8
|
+
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
|
9
|
+
copies of the Software, and to permit persons to whom the Software is
|
|
10
|
+
furnished to do so, subject to the following conditions:
|
|
11
|
+
|
|
12
|
+
The above copyright notice and this permission notice shall be included in all
|
|
13
|
+
copies or substantial portions of the Software.
|
|
14
|
+
|
|
15
|
+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
16
|
+
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
17
|
+
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
|
18
|
+
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
19
|
+
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
20
|
+
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
|
21
|
+
SOFTWARE.
|
|
@@ -0,0 +1,218 @@
|
|
|
1
|
+
Metadata-Version: 2.4
|
|
2
|
+
Name: agent-shared-workspace
|
|
3
|
+
Version: 0.1.0
|
|
4
|
+
Summary: Plain-text shared memory hub for multiple coding agents: Markdown SSOT + concurrency-safe CLI + zero-dependency MCP server.
|
|
5
|
+
Author: MAHONGHAO1
|
|
6
|
+
License-Expression: MIT
|
|
7
|
+
Project-URL: Homepage, https://github.com/MAHONGHAO1/agent-shared-workspace
|
|
8
|
+
Project-URL: Repository, https://github.com/MAHONGHAO1/agent-shared-workspace
|
|
9
|
+
Project-URL: Issues, https://github.com/MAHONGHAO1/agent-shared-workspace/issues
|
|
10
|
+
Keywords: ai-agents,multi-agent,memory,mcp,cli,markdown,claude-code,codex
|
|
11
|
+
Classifier: Development Status :: 4 - Beta
|
|
12
|
+
Classifier: Intended Audience :: Developers
|
|
13
|
+
Classifier: Programming Language :: Python :: 3
|
|
14
|
+
Classifier: Programming Language :: Python :: 3.9
|
|
15
|
+
Classifier: Programming Language :: Python :: 3.10
|
|
16
|
+
Classifier: Programming Language :: Python :: 3.11
|
|
17
|
+
Classifier: Programming Language :: Python :: 3.12
|
|
18
|
+
Classifier: Topic :: Software Development :: Libraries :: Python Modules
|
|
19
|
+
Classifier: Topic :: Utilities
|
|
20
|
+
Requires-Python: >=3.9
|
|
21
|
+
Description-Content-Type: text/markdown
|
|
22
|
+
License-File: LICENSE
|
|
23
|
+
Dynamic: license-file
|
|
24
|
+
|
|
25
|
+
# Agent Shared Workspace
|
|
26
|
+
|
|
27
|
+
[](https://github.com/MAHONGHAO1/agent-shared-workspace/actions/workflows/tests.yml)
|
|
28
|
+
[](https://www.python.org/)
|
|
29
|
+
[](#)
|
|
30
|
+
[](LICENSE)
|
|
31
|
+
|
|
32
|
+
[中文](README.md) | **English**
|
|
33
|
+
|
|
34
|
+
**A dependency-free shared memory hub for multiple coding agents: plain-text Markdown SSOT + concurrency-safe CLI + Git versioning.**
|
|
35
|
+
|
|
36
|
+
---
|
|
37
|
+
|
|
38
|
+
## Why This Exists
|
|
39
|
+
|
|
40
|
+
If you run Codex, Claude Code, DeepSeek Harness, Antigravity, or other coding agents on the same codebase, each agent keeps its own isolated memory:
|
|
41
|
+
|
|
42
|
+
- Three agents hit the exact same bug three times.
|
|
43
|
+
- Yesterday's root-cause analysis disappears in today's session.
|
|
44
|
+
- Passing current project status between agents turns into repetitive copy-pasting.
|
|
45
|
+
- Vector-DB memory feels like a black box: hard to inspect, hard to edit, and easy to pollute silently.
|
|
46
|
+
|
|
47
|
+
This tool takes a simpler route: **shared memory lives in plain Markdown files inside a local workspace, every agent reads/writes through one stdlib-only CLI guarded by a cross-process file lock, and Git handles history and rollback.**
|
|
48
|
+
|
|
49
|
+
## Architecture
|
|
50
|
+
|
|
51
|
+
```mermaid
|
|
52
|
+
flowchart LR
|
|
53
|
+
subgraph Agents["Multiple Coding Agents (Thin Entry Pointers / Skill)"]
|
|
54
|
+
A1["Codex\n~/.codex/AGENTS.md"]
|
|
55
|
+
A2["Claude Code\n~/.claude/CLAUDE.md"]
|
|
56
|
+
A3["DeepSeek Harness\n~/.dsh/AGENTS.md"]
|
|
57
|
+
A4["Antigravity\n.agents/skills/"]
|
|
58
|
+
end
|
|
59
|
+
|
|
60
|
+
subgraph CLI["Unified Memory CLI (Python Stdlib Only)"]
|
|
61
|
+
TOOL["workspace_tool.py\ninit · exp · project · handoff · guard · commit"]
|
|
62
|
+
LOCK[".workspace.lock\nmsvcrt / fcntl exclusive lock"]
|
|
63
|
+
TOOL --- LOCK
|
|
64
|
+
end
|
|
65
|
+
|
|
66
|
+
subgraph SSOT["Plain-Text Single Source of Truth + Git"]
|
|
67
|
+
M1["AGENTS.md / NEW_PROJECT_SOP.md"]
|
|
68
|
+
M2["WORKSPACE.md"]
|
|
69
|
+
M3["EXPERIENCES.md"]
|
|
70
|
+
M4["Project_XXX/docs/当前状态与交接.md"]
|
|
71
|
+
GIT[("Local Git\nAuditable · Reversible")]
|
|
72
|
+
end
|
|
73
|
+
|
|
74
|
+
A1 & A2 & A3 & A4 --> TOOL
|
|
75
|
+
TOOL --> M1 & M2 & M3 & M4
|
|
76
|
+
M1 & M2 & M3 & M4 --> GIT
|
|
77
|
+
```
|
|
78
|
+
|
|
79
|
+
## Real CLI Output
|
|
80
|
+
|
|
81
|
+

|
|
82
|
+
|
|
83
|
+
## One-Line Install (4 Distribution Channels)
|
|
84
|
+
|
|
85
|
+
### 1. Agent Skill (`skills.sh` / Codex / Claude Code / Antigravity / Cursor)
|
|
86
|
+
```bash
|
|
87
|
+
npx skills add MAHONGHAO1/agent-shared-workspace
|
|
88
|
+
```
|
|
89
|
+
Ships with a self-contained stdlib `scripts/workspace_tool.py` inside the skill folder.
|
|
90
|
+
|
|
91
|
+
### 2. Claude Code Plugin (Skill + MCP Server)
|
|
92
|
+
```text
|
|
93
|
+
/plugin marketplace add MAHONGHAO1/agent-shared-workspace
|
|
94
|
+
/plugin install agent-shared-workspace
|
|
95
|
+
```
|
|
96
|
+
|
|
97
|
+
### 3. MCP Server (Zero Dependencies — Cursor / Claude Desktop / Antigravity)
|
|
98
|
+
```json
|
|
99
|
+
{
|
|
100
|
+
"mcpServers": {
|
|
101
|
+
"agent-shared-workspace": {
|
|
102
|
+
"command": "python",
|
|
103
|
+
"args": ["-m", "agent_shared_workspace.mcp_server"],
|
|
104
|
+
"env": {
|
|
105
|
+
"PYTHONPATH": "/path/to/agent-shared-workspace/src",
|
|
106
|
+
"AGENT_WORKSPACE_ROOT": "/path/to/your-workspace"
|
|
107
|
+
}
|
|
108
|
+
}
|
|
109
|
+
}
|
|
110
|
+
}
|
|
111
|
+
```
|
|
112
|
+
> Or install locally via `pip install .` to expose `agent-workspace` (CLI) and `agent-workspace-mcp` (MCP stdio server).
|
|
113
|
+
|
|
114
|
+
### 4. DeepSeek Harness Plugin Bundle
|
|
115
|
+
Ready-to-publish DSH bundle lives in [`integrations/deepseek-harness/`](integrations/deepseek-harness/) (`dsh-agent-shared-workspace`, mounts the skill via `@deepseek-ai/dsh-skill-filesystem`).
|
|
116
|
+
|
|
117
|
+
## Quick Start
|
|
118
|
+
|
|
119
|
+
Zero third-party packages required (Python 3.9+ stdlib only):
|
|
120
|
+
|
|
121
|
+
```bash
|
|
122
|
+
# 1. Point to your workspace directory (defaults to cwd if unset)
|
|
123
|
+
export AGENT_WORKSPACE_ROOT=~/agent-workspace
|
|
124
|
+
# Windows PowerShell: $env:AGENT_WORKSPACE_ROOT="D:\agent-workspace"
|
|
125
|
+
|
|
126
|
+
# 2. Initialize skeleton files, workspace.config.json, .gitignore, and Git
|
|
127
|
+
python src/workspace_tool.py init --lang en
|
|
128
|
+
|
|
129
|
+
# 3. Create and register a project by SOP
|
|
130
|
+
python src/workspace_tool.py project new "ApiGateway" --owner codex
|
|
131
|
+
|
|
132
|
+
# 4. Record a tagged pitfall/experience
|
|
133
|
+
python src/workspace_tool.py exp add \
|
|
134
|
+
--title "Non-root container fails with EACCES on bind mount" \
|
|
135
|
+
--tag "docker,linux" \
|
|
136
|
+
--phenomenon "Container process gets Permission denied writing to mounted volume" \
|
|
137
|
+
--cause "Host directory is owned by root while container runs as uid 10001" \
|
|
138
|
+
--solution "Set user: in compose or chown in entrypoint" \
|
|
139
|
+
--criteria "Container successfully writes a 1 KB test file" \
|
|
140
|
+
--agent "claude-code"
|
|
141
|
+
|
|
142
|
+
# 5. Search by keyword and/or tag
|
|
143
|
+
python src/workspace_tool.py exp search --tag docker
|
|
144
|
+
python src/workspace_tool.py exp search "EACCES" --tag linux
|
|
145
|
+
python src/workspace_tool.py exp get EXP-001
|
|
146
|
+
|
|
147
|
+
# 6. Record a structured handoff (syncs handoff doc, project README, and board)
|
|
148
|
+
python src/workspace_tool.py handoff 001 --agent "codex" \
|
|
149
|
+
--status "Rate limiter verified" \
|
|
150
|
+
--done "Implemented token bucket and 12 unit tests" \
|
|
151
|
+
--evidence "tests/test_ratelimit.py 12/12" \
|
|
152
|
+
--todo "Canary rollout"
|
|
153
|
+
|
|
154
|
+
# 7. Check for uncommitted changes and commit
|
|
155
|
+
python src/workspace_tool.py guard
|
|
156
|
+
python src/workspace_tool.py commit -m "docs: initialize workspace and log EXP-001"
|
|
157
|
+
```
|
|
158
|
+
|
|
159
|
+
## Commands
|
|
160
|
+
|
|
161
|
+
| Command | Purpose |
|
|
162
|
+
| --- | --- |
|
|
163
|
+
| `init [--lang zh\|en] [--no-git]` | Idempotently scaffold docs, `workspace.config.json`, `.gitignore`, and Git |
|
|
164
|
+
| `status [--guard]` | Show active projects and Git status; `--guard` exits with code 1 if dirty |
|
|
165
|
+
| `guard` | Lightweight dirty-tree check (exit 0 when clean, exit 1 when dirty) |
|
|
166
|
+
| `exp search [query] [--tag t] [--list-tags]` | Search experiences by keyword/tag or list all tags with counts |
|
|
167
|
+
| `exp get EXP-001` | Inspect index metadata and full entry body (or sink path) |
|
|
168
|
+
| `exp add --title ... [--tag a,b] [--sink 001]` | Append a new experience with auto-incremented ID and index update |
|
|
169
|
+
| `exp update EXP-001 --note ...` | Append new evidence to an existing experience instead of duplicating |
|
|
170
|
+
| `project list` | List all `Project_XXX_*` directories and highest ID |
|
|
171
|
+
| `project new <name> --owner <agent>` | Create a project directory, README, handoff file, and register it |
|
|
172
|
+
| `project update <id> --status ...` | Sync status to both project README and workspace board |
|
|
173
|
+
| `handoff <id> --agent ... --status ... --done ...` | Append a handoff record and atomically update README and board |
|
|
174
|
+
| `commit -m <message>` | Stage and commit workspace changes (automatically excludes `.workspace.lock`) |
|
|
175
|
+
|
|
176
|
+
## Built-in Safeguards
|
|
177
|
+
|
|
178
|
+
- **Cross-process Reentrant File Lock**: All write commands acquire `.workspace.lock` using `msvcrt.locking` on Windows and `fcntl.flock` on Linux/macOS. Concurrent agents serialize cleanly with zero ID collisions or torn Markdown. Timeout defaults to 10s (`AGENT_WORKSPACE_LOCK_TIMEOUT` to override).
|
|
179
|
+
- **Dirty-Tree Guard**: `guard` and `status --guard` return exit code `1` whenever uncommitted edits remain, suitable for shell hooks or end-of-session checks.
|
|
180
|
+
- **Configurable Filenames**: `init --lang en` writes `WORKSPACE.md` / `EXPERIENCES.md` / `NEW_PROJECT_SOP.md` and saves the mapping in `workspace.config.json`.
|
|
181
|
+
|
|
182
|
+
## Wiring Into Your Agents
|
|
183
|
+
|
|
184
|
+
Copy the short snippets from [`examples/AGENTS-snippet.md`](examples/AGENTS-snippet.md) into each agent's native instruction file:
|
|
185
|
+
|
|
186
|
+
| Agent | Entry File |
|
|
187
|
+
| --- | --- |
|
|
188
|
+
| Codex | `~/.codex/AGENTS.md` |
|
|
189
|
+
| Claude Code | `~/.claude/CLAUDE.md` |
|
|
190
|
+
| DeepSeek Harness | `~/.dsh/AGENTS.md` |
|
|
191
|
+
| Antigravity | `<workspace>/.agents/skills/shared-workspace/SKILL.md` (see [`examples/SKILL.md`](examples/SKILL.md)) |
|
|
192
|
+
|
|
193
|
+
Keep agent entry files thin (commands & triggers only). Do not paste full workspace docs into every system prompt.
|
|
194
|
+
|
|
195
|
+
## Why Not a Vector Database?
|
|
196
|
+
|
|
197
|
+
| Aspect | Vector RAG | This Approach |
|
|
198
|
+
| --- | --- | --- |
|
|
199
|
+
| Readability | Retrieved chunks, opaque ranking | Plain Markdown humans read and edit directly |
|
|
200
|
+
| Correctability | Requires re-embedding / re-indexing | Edit a paragraph or run `exp update` |
|
|
201
|
+
| Versioning | Separate snapshot tooling | Native Git diff, blame, and revert |
|
|
202
|
+
| Dependencies | Vector store + embedding service | Python stdlib only |
|
|
203
|
+
| Failure mode | Silently retrieves irrelevant memory | Explicit miss — visible and debuggable |
|
|
204
|
+
|
|
205
|
+
Trade-off: keyword, tag, and ID lookup only (no semantic similarity). Ideal for solo developers or small teams managing dozens to hundreds of verified lessons across multiple local agents.
|
|
206
|
+
|
|
207
|
+
## Tests & CI
|
|
208
|
+
|
|
209
|
+
```bash
|
|
210
|
+
python -m unittest discover -s tests -v
|
|
211
|
+
cd integrations/deepseek-harness && npm run check
|
|
212
|
+
```
|
|
213
|
+
|
|
214
|
+
**15 Python end-to-end tests + 3 Node.js DSH bundle checks** run in isolated temp directories on both **Ubuntu** and **Windows** in GitHub Actions, including an **8-process concurrent write stress test**, **lock-timeout verification**, and a full **MCP stdio JSON-RPC lifecycle test**.
|
|
215
|
+
|
|
216
|
+
## License
|
|
217
|
+
|
|
218
|
+
[MIT](LICENSE)
|
|
@@ -0,0 +1,194 @@
|
|
|
1
|
+
# Agent Shared Workspace
|
|
2
|
+
|
|
3
|
+
[](https://github.com/MAHONGHAO1/agent-shared-workspace/actions/workflows/tests.yml)
|
|
4
|
+
[](https://www.python.org/)
|
|
5
|
+
[](#)
|
|
6
|
+
[](LICENSE)
|
|
7
|
+
|
|
8
|
+
[中文](README.md) | **English**
|
|
9
|
+
|
|
10
|
+
**A dependency-free shared memory hub for multiple coding agents: plain-text Markdown SSOT + concurrency-safe CLI + Git versioning.**
|
|
11
|
+
|
|
12
|
+
---
|
|
13
|
+
|
|
14
|
+
## Why This Exists
|
|
15
|
+
|
|
16
|
+
If you run Codex, Claude Code, DeepSeek Harness, Antigravity, or other coding agents on the same codebase, each agent keeps its own isolated memory:
|
|
17
|
+
|
|
18
|
+
- Three agents hit the exact same bug three times.
|
|
19
|
+
- Yesterday's root-cause analysis disappears in today's session.
|
|
20
|
+
- Passing current project status between agents turns into repetitive copy-pasting.
|
|
21
|
+
- Vector-DB memory feels like a black box: hard to inspect, hard to edit, and easy to pollute silently.
|
|
22
|
+
|
|
23
|
+
This tool takes a simpler route: **shared memory lives in plain Markdown files inside a local workspace, every agent reads/writes through one stdlib-only CLI guarded by a cross-process file lock, and Git handles history and rollback.**
|
|
24
|
+
|
|
25
|
+
## Architecture
|
|
26
|
+
|
|
27
|
+
```mermaid
|
|
28
|
+
flowchart LR
|
|
29
|
+
subgraph Agents["Multiple Coding Agents (Thin Entry Pointers / Skill)"]
|
|
30
|
+
A1["Codex\n~/.codex/AGENTS.md"]
|
|
31
|
+
A2["Claude Code\n~/.claude/CLAUDE.md"]
|
|
32
|
+
A3["DeepSeek Harness\n~/.dsh/AGENTS.md"]
|
|
33
|
+
A4["Antigravity\n.agents/skills/"]
|
|
34
|
+
end
|
|
35
|
+
|
|
36
|
+
subgraph CLI["Unified Memory CLI (Python Stdlib Only)"]
|
|
37
|
+
TOOL["workspace_tool.py\ninit · exp · project · handoff · guard · commit"]
|
|
38
|
+
LOCK[".workspace.lock\nmsvcrt / fcntl exclusive lock"]
|
|
39
|
+
TOOL --- LOCK
|
|
40
|
+
end
|
|
41
|
+
|
|
42
|
+
subgraph SSOT["Plain-Text Single Source of Truth + Git"]
|
|
43
|
+
M1["AGENTS.md / NEW_PROJECT_SOP.md"]
|
|
44
|
+
M2["WORKSPACE.md"]
|
|
45
|
+
M3["EXPERIENCES.md"]
|
|
46
|
+
M4["Project_XXX/docs/当前状态与交接.md"]
|
|
47
|
+
GIT[("Local Git\nAuditable · Reversible")]
|
|
48
|
+
end
|
|
49
|
+
|
|
50
|
+
A1 & A2 & A3 & A4 --> TOOL
|
|
51
|
+
TOOL --> M1 & M2 & M3 & M4
|
|
52
|
+
M1 & M2 & M3 & M4 --> GIT
|
|
53
|
+
```
|
|
54
|
+
|
|
55
|
+
## Real CLI Output
|
|
56
|
+
|
|
57
|
+

|
|
58
|
+
|
|
59
|
+
## One-Line Install (4 Distribution Channels)
|
|
60
|
+
|
|
61
|
+
### 1. Agent Skill (`skills.sh` / Codex / Claude Code / Antigravity / Cursor)
|
|
62
|
+
```bash
|
|
63
|
+
npx skills add MAHONGHAO1/agent-shared-workspace
|
|
64
|
+
```
|
|
65
|
+
Ships with a self-contained stdlib `scripts/workspace_tool.py` inside the skill folder.
|
|
66
|
+
|
|
67
|
+
### 2. Claude Code Plugin (Skill + MCP Server)
|
|
68
|
+
```text
|
|
69
|
+
/plugin marketplace add MAHONGHAO1/agent-shared-workspace
|
|
70
|
+
/plugin install agent-shared-workspace
|
|
71
|
+
```
|
|
72
|
+
|
|
73
|
+
### 3. MCP Server (Zero Dependencies — Cursor / Claude Desktop / Antigravity)
|
|
74
|
+
```json
|
|
75
|
+
{
|
|
76
|
+
"mcpServers": {
|
|
77
|
+
"agent-shared-workspace": {
|
|
78
|
+
"command": "python",
|
|
79
|
+
"args": ["-m", "agent_shared_workspace.mcp_server"],
|
|
80
|
+
"env": {
|
|
81
|
+
"PYTHONPATH": "/path/to/agent-shared-workspace/src",
|
|
82
|
+
"AGENT_WORKSPACE_ROOT": "/path/to/your-workspace"
|
|
83
|
+
}
|
|
84
|
+
}
|
|
85
|
+
}
|
|
86
|
+
}
|
|
87
|
+
```
|
|
88
|
+
> Or install locally via `pip install .` to expose `agent-workspace` (CLI) and `agent-workspace-mcp` (MCP stdio server).
|
|
89
|
+
|
|
90
|
+
### 4. DeepSeek Harness Plugin Bundle
|
|
91
|
+
Ready-to-publish DSH bundle lives in [`integrations/deepseek-harness/`](integrations/deepseek-harness/) (`dsh-agent-shared-workspace`, mounts the skill via `@deepseek-ai/dsh-skill-filesystem`).
|
|
92
|
+
|
|
93
|
+
## Quick Start
|
|
94
|
+
|
|
95
|
+
Zero third-party packages required (Python 3.9+ stdlib only):
|
|
96
|
+
|
|
97
|
+
```bash
|
|
98
|
+
# 1. Point to your workspace directory (defaults to cwd if unset)
|
|
99
|
+
export AGENT_WORKSPACE_ROOT=~/agent-workspace
|
|
100
|
+
# Windows PowerShell: $env:AGENT_WORKSPACE_ROOT="D:\agent-workspace"
|
|
101
|
+
|
|
102
|
+
# 2. Initialize skeleton files, workspace.config.json, .gitignore, and Git
|
|
103
|
+
python src/workspace_tool.py init --lang en
|
|
104
|
+
|
|
105
|
+
# 3. Create and register a project by SOP
|
|
106
|
+
python src/workspace_tool.py project new "ApiGateway" --owner codex
|
|
107
|
+
|
|
108
|
+
# 4. Record a tagged pitfall/experience
|
|
109
|
+
python src/workspace_tool.py exp add \
|
|
110
|
+
--title "Non-root container fails with EACCES on bind mount" \
|
|
111
|
+
--tag "docker,linux" \
|
|
112
|
+
--phenomenon "Container process gets Permission denied writing to mounted volume" \
|
|
113
|
+
--cause "Host directory is owned by root while container runs as uid 10001" \
|
|
114
|
+
--solution "Set user: in compose or chown in entrypoint" \
|
|
115
|
+
--criteria "Container successfully writes a 1 KB test file" \
|
|
116
|
+
--agent "claude-code"
|
|
117
|
+
|
|
118
|
+
# 5. Search by keyword and/or tag
|
|
119
|
+
python src/workspace_tool.py exp search --tag docker
|
|
120
|
+
python src/workspace_tool.py exp search "EACCES" --tag linux
|
|
121
|
+
python src/workspace_tool.py exp get EXP-001
|
|
122
|
+
|
|
123
|
+
# 6. Record a structured handoff (syncs handoff doc, project README, and board)
|
|
124
|
+
python src/workspace_tool.py handoff 001 --agent "codex" \
|
|
125
|
+
--status "Rate limiter verified" \
|
|
126
|
+
--done "Implemented token bucket and 12 unit tests" \
|
|
127
|
+
--evidence "tests/test_ratelimit.py 12/12" \
|
|
128
|
+
--todo "Canary rollout"
|
|
129
|
+
|
|
130
|
+
# 7. Check for uncommitted changes and commit
|
|
131
|
+
python src/workspace_tool.py guard
|
|
132
|
+
python src/workspace_tool.py commit -m "docs: initialize workspace and log EXP-001"
|
|
133
|
+
```
|
|
134
|
+
|
|
135
|
+
## Commands
|
|
136
|
+
|
|
137
|
+
| Command | Purpose |
|
|
138
|
+
| --- | --- |
|
|
139
|
+
| `init [--lang zh\|en] [--no-git]` | Idempotently scaffold docs, `workspace.config.json`, `.gitignore`, and Git |
|
|
140
|
+
| `status [--guard]` | Show active projects and Git status; `--guard` exits with code 1 if dirty |
|
|
141
|
+
| `guard` | Lightweight dirty-tree check (exit 0 when clean, exit 1 when dirty) |
|
|
142
|
+
| `exp search [query] [--tag t] [--list-tags]` | Search experiences by keyword/tag or list all tags with counts |
|
|
143
|
+
| `exp get EXP-001` | Inspect index metadata and full entry body (or sink path) |
|
|
144
|
+
| `exp add --title ... [--tag a,b] [--sink 001]` | Append a new experience with auto-incremented ID and index update |
|
|
145
|
+
| `exp update EXP-001 --note ...` | Append new evidence to an existing experience instead of duplicating |
|
|
146
|
+
| `project list` | List all `Project_XXX_*` directories and highest ID |
|
|
147
|
+
| `project new <name> --owner <agent>` | Create a project directory, README, handoff file, and register it |
|
|
148
|
+
| `project update <id> --status ...` | Sync status to both project README and workspace board |
|
|
149
|
+
| `handoff <id> --agent ... --status ... --done ...` | Append a handoff record and atomically update README and board |
|
|
150
|
+
| `commit -m <message>` | Stage and commit workspace changes (automatically excludes `.workspace.lock`) |
|
|
151
|
+
|
|
152
|
+
## Built-in Safeguards
|
|
153
|
+
|
|
154
|
+
- **Cross-process Reentrant File Lock**: All write commands acquire `.workspace.lock` using `msvcrt.locking` on Windows and `fcntl.flock` on Linux/macOS. Concurrent agents serialize cleanly with zero ID collisions or torn Markdown. Timeout defaults to 10s (`AGENT_WORKSPACE_LOCK_TIMEOUT` to override).
|
|
155
|
+
- **Dirty-Tree Guard**: `guard` and `status --guard` return exit code `1` whenever uncommitted edits remain, suitable for shell hooks or end-of-session checks.
|
|
156
|
+
- **Configurable Filenames**: `init --lang en` writes `WORKSPACE.md` / `EXPERIENCES.md` / `NEW_PROJECT_SOP.md` and saves the mapping in `workspace.config.json`.
|
|
157
|
+
|
|
158
|
+
## Wiring Into Your Agents
|
|
159
|
+
|
|
160
|
+
Copy the short snippets from [`examples/AGENTS-snippet.md`](examples/AGENTS-snippet.md) into each agent's native instruction file:
|
|
161
|
+
|
|
162
|
+
| Agent | Entry File |
|
|
163
|
+
| --- | --- |
|
|
164
|
+
| Codex | `~/.codex/AGENTS.md` |
|
|
165
|
+
| Claude Code | `~/.claude/CLAUDE.md` |
|
|
166
|
+
| DeepSeek Harness | `~/.dsh/AGENTS.md` |
|
|
167
|
+
| Antigravity | `<workspace>/.agents/skills/shared-workspace/SKILL.md` (see [`examples/SKILL.md`](examples/SKILL.md)) |
|
|
168
|
+
|
|
169
|
+
Keep agent entry files thin (commands & triggers only). Do not paste full workspace docs into every system prompt.
|
|
170
|
+
|
|
171
|
+
## Why Not a Vector Database?
|
|
172
|
+
|
|
173
|
+
| Aspect | Vector RAG | This Approach |
|
|
174
|
+
| --- | --- | --- |
|
|
175
|
+
| Readability | Retrieved chunks, opaque ranking | Plain Markdown humans read and edit directly |
|
|
176
|
+
| Correctability | Requires re-embedding / re-indexing | Edit a paragraph or run `exp update` |
|
|
177
|
+
| Versioning | Separate snapshot tooling | Native Git diff, blame, and revert |
|
|
178
|
+
| Dependencies | Vector store + embedding service | Python stdlib only |
|
|
179
|
+
| Failure mode | Silently retrieves irrelevant memory | Explicit miss — visible and debuggable |
|
|
180
|
+
|
|
181
|
+
Trade-off: keyword, tag, and ID lookup only (no semantic similarity). Ideal for solo developers or small teams managing dozens to hundreds of verified lessons across multiple local agents.
|
|
182
|
+
|
|
183
|
+
## Tests & CI
|
|
184
|
+
|
|
185
|
+
```bash
|
|
186
|
+
python -m unittest discover -s tests -v
|
|
187
|
+
cd integrations/deepseek-harness && npm run check
|
|
188
|
+
```
|
|
189
|
+
|
|
190
|
+
**15 Python end-to-end tests + 3 Node.js DSH bundle checks** run in isolated temp directories on both **Ubuntu** and **Windows** in GitHub Actions, including an **8-process concurrent write stress test**, **lock-timeout verification**, and a full **MCP stdio JSON-RPC lifecycle test**.
|
|
191
|
+
|
|
192
|
+
## License
|
|
193
|
+
|
|
194
|
+
[MIT](LICENSE)
|
|
@@ -0,0 +1,232 @@
|
|
|
1
|
+
# Agent Shared Workspace
|
|
2
|
+
|
|
3
|
+
[](https://github.com/MAHONGHAO1/agent-shared-workspace/actions/workflows/tests.yml)
|
|
4
|
+
[](https://www.python.org/)
|
|
5
|
+
[](#)
|
|
6
|
+
[](LICENSE)
|
|
7
|
+
|
|
8
|
+
**中文** | [English](README.en.md)
|
|
9
|
+
|
|
10
|
+
**给多 Agent 用的一套共享记忆:纯文本 SSOT + 零依赖并发安全 CLI + Git 版本化。**
|
|
11
|
+
|
|
12
|
+
---
|
|
13
|
+
|
|
14
|
+
## 它解决什么问题
|
|
15
|
+
|
|
16
|
+
当你同时用 Codex、Claude Code、DeepSeek Harness、Antigravity 等多个 Coding Agent 做同一批项目时,它们各记各的,很快就会出现:
|
|
17
|
+
|
|
18
|
+
- 同一个坑,三个 Agent 踩三遍;
|
|
19
|
+
- 昨天的定位结论,今天换个 Agent 全部重来;
|
|
20
|
+
- 想把当前进度同步给下一个 Agent,只能反复复制粘贴;
|
|
21
|
+
- 上向量数据库做记忆,召回黑盒不可控,人既看不清它为什么想起这条,也很难直接改。
|
|
22
|
+
|
|
23
|
+
这套方案用最朴素的做法解决它:**记忆就是本地仓库里的几份 Markdown,所有 Agent 通过同一个零依赖 CLI(带跨进程文件锁)读写,Git 负责版本与回滚。**
|
|
24
|
+
|
|
25
|
+
## 架构一览
|
|
26
|
+
|
|
27
|
+
```mermaid
|
|
28
|
+
flowchart LR
|
|
29
|
+
subgraph Agents["多个 Coding Agent(薄入口指针 / Skill)"]
|
|
30
|
+
A1["Codex\n~/.codex/AGENTS.md"]
|
|
31
|
+
A2["Claude Code\n~/.claude/CLAUDE.md"]
|
|
32
|
+
A3["DeepSeek Harness\n~/.dsh/AGENTS.md"]
|
|
33
|
+
A4["Antigravity\n.agents/skills/"]
|
|
34
|
+
end
|
|
35
|
+
|
|
36
|
+
subgraph CLI["统一记忆 CLI(Python 标准库)"]
|
|
37
|
+
TOOL["workspace_tool.py\ninit · exp · project · handoff · guard · commit"]
|
|
38
|
+
LOCK[".workspace.lock\nmsvcrt / fcntl 跨进程排他锁"]
|
|
39
|
+
TOOL --- LOCK
|
|
40
|
+
end
|
|
41
|
+
|
|
42
|
+
subgraph SSOT["纯文本单一真相源 + Git"]
|
|
43
|
+
M1["AGENTS.md / SOP"]
|
|
44
|
+
M2["全局工作台.md / WORKSPACE.md"]
|
|
45
|
+
M3["全局复用与踩坑日志.md / EXPERIENCES.md"]
|
|
46
|
+
M4["Project_XXX/docs/当前状态与交接.md"]
|
|
47
|
+
GIT[("Local Git\n可审计 · 可回滚")]
|
|
48
|
+
end
|
|
49
|
+
|
|
50
|
+
A1 & A2 & A3 & A4 --> TOOL
|
|
51
|
+
TOOL --> M1 & M2 & M3 & M4
|
|
52
|
+
M1 & M2 & M3 & M4 --> GIT
|
|
53
|
+
```
|
|
54
|
+
|
|
55
|
+
## 实际运行效果
|
|
56
|
+
|
|
57
|
+

|
|
58
|
+
|
|
59
|
+
## 一键安装(4 种方式任选)
|
|
60
|
+
|
|
61
|
+
### 1. Agent Skill(`skills.sh` / Codex / Claude Code / Antigravity / Cursor)
|
|
62
|
+
```bash
|
|
63
|
+
npx skills add MAHONGHAO1/agent-shared-workspace
|
|
64
|
+
```
|
|
65
|
+
技能目录自带零依赖 `scripts/workspace_tool.py`,装完即可直接使用。
|
|
66
|
+
|
|
67
|
+
### 2. Claude Code 插件(含 Skill + MCP Server)
|
|
68
|
+
```text
|
|
69
|
+
/plugin marketplace add MAHONGHAO1/agent-shared-workspace
|
|
70
|
+
/plugin install agent-shared-workspace
|
|
71
|
+
```
|
|
72
|
+
|
|
73
|
+
### 3. MCP Server(零外部依赖,支持 Cursor / Claude Desktop / Antigravity)
|
|
74
|
+
```json
|
|
75
|
+
{
|
|
76
|
+
"mcpServers": {
|
|
77
|
+
"agent-shared-workspace": {
|
|
78
|
+
"command": "python",
|
|
79
|
+
"args": ["-m", "agent_shared_workspace.mcp_server"],
|
|
80
|
+
"env": {
|
|
81
|
+
"PYTHONPATH": "/path/to/agent-shared-workspace/src",
|
|
82
|
+
"AGENT_WORKSPACE_ROOT": "/path/to/your-workspace"
|
|
83
|
+
}
|
|
84
|
+
}
|
|
85
|
+
}
|
|
86
|
+
}
|
|
87
|
+
```
|
|
88
|
+
> 亦支持标准 Python 包安装:`pip install .` 后可直接运行 `agent-workspace`(CLI)与 `agent-workspace-mcp`(MCP stdio Server)。
|
|
89
|
+
|
|
90
|
+
### 4. DeepSeek Harness 插件包
|
|
91
|
+
仓库内置即用型 DSh 插件包位于 [`integrations/deepseek-harness/`](integrations/deepseek-harness/)(包名 `dsh-agent-shared-workspace`,通过 `@deepseek-ai/dsh-skill-filesystem` 自动挂载共享记忆技能与 CLI)。
|
|
92
|
+
|
|
93
|
+
## 快速开始(30 秒跑通)
|
|
94
|
+
|
|
95
|
+
无需安装任何第三方依赖(Python 3.9+ 标准库即可运行):
|
|
96
|
+
|
|
97
|
+
```bash
|
|
98
|
+
# 1. 指定你的共享工作空间目录(不设则默认当前目录)
|
|
99
|
+
export AGENT_WORKSPACE_ROOT=~/agent-workspace
|
|
100
|
+
# Windows PowerShell: $env:AGENT_WORKSPACE_ROOT="D:\agent-workspace"
|
|
101
|
+
|
|
102
|
+
# 2. 一键初始化骨架文档、workspace.config.json、.gitignore 与本地 Git
|
|
103
|
+
# 默认中文骨架(全局工作台.md 等);英文骨架传 --lang en(WORKSPACE.md 等)
|
|
104
|
+
python src/workspace_tool.py init --lang zh
|
|
105
|
+
|
|
106
|
+
# 3. 按 SOP 新建项目并自动登记
|
|
107
|
+
python src/workspace_tool.py project new "API网关改造" --owner codex
|
|
108
|
+
|
|
109
|
+
# 4. 沉淀一条带标签的踩坑经验
|
|
110
|
+
python src/workspace_tool.py exp add \
|
|
111
|
+
--title "容器挂载卷非root写入报EACCES" \
|
|
112
|
+
--tag "docker,linux" \
|
|
113
|
+
--phenomenon "bind mount 后进程以 non-root 运行,创建目录报 Permission denied" \
|
|
114
|
+
--cause "宿主机目录属主是 root,容器内 uid 10001 无写权限" \
|
|
115
|
+
--solution "compose 里指定 user: 或 entrypoint 内 chown" \
|
|
116
|
+
--criteria "容器内成功写入 1 KB 校验文件" \
|
|
117
|
+
--agent "claude-code"
|
|
118
|
+
|
|
119
|
+
# 5. 按关键词或标签检索经验
|
|
120
|
+
python src/workspace_tool.py exp search --tag docker
|
|
121
|
+
python src/workspace_tool.py exp search "权限" --tag linux
|
|
122
|
+
python src/workspace_tool.py exp get EXP-001
|
|
123
|
+
|
|
124
|
+
# 6. 记录阶段交接(自动同步交接文档、项目 README 与全局登记表)
|
|
125
|
+
python src/workspace_tool.py handoff 001 --agent "codex" \
|
|
126
|
+
--status "网关限流联调完成" \
|
|
127
|
+
--done "完成令牌桶与 12 项单测" \
|
|
128
|
+
--evidence "tests/test_ratelimit.py 12/12" \
|
|
129
|
+
--todo "灰度发布"
|
|
130
|
+
|
|
131
|
+
# 7. 退出前巡检并固化到本地 Git
|
|
132
|
+
python src/workspace_tool.py guard
|
|
133
|
+
python src/workspace_tool.py commit -m "docs: 初始化工作区并沉淀 EXP-001"
|
|
134
|
+
```
|
|
135
|
+
|
|
136
|
+
## 目录结构(中英文自动对齐)
|
|
137
|
+
|
|
138
|
+
`init --lang zh` 与 `init --lang en` 会自动生成对应的 `workspace.config.json`,工具会按配置自动识别文件名,不再出现文档与代码不一致:
|
|
139
|
+
|
|
140
|
+
| 角色 | 中文骨架(`--lang zh`) | 英文骨架(`--lang en`) |
|
|
141
|
+
| --- | --- | --- |
|
|
142
|
+
| 规则入口 | `AGENTS.md` | `AGENTS.md` |
|
|
143
|
+
| 项目登记表 | `全局工作台.md` | `WORKSPACE.md` |
|
|
144
|
+
| 经验索引与正文 | `全局复用与踩坑日志.md` | `EXPERIENCES.md` |
|
|
145
|
+
| 建项规范 | `新项目创建SOP.md` | `NEW_PROJECT_SOP.md` |
|
|
146
|
+
| 文件名映射配置 | `workspace.config.json` | `workspace.config.json` |
|
|
147
|
+
|
|
148
|
+
```text
|
|
149
|
+
your-workspace/
|
|
150
|
+
├── AGENTS.md
|
|
151
|
+
├── workspace.config.json
|
|
152
|
+
├── 全局工作台.md # 或 WORKSPACE.md
|
|
153
|
+
├── 全局复用与踩坑日志.md # 或 EXPERIENCES.md
|
|
154
|
+
├── 新项目创建SOP.md # 或 NEW_PROJECT_SOP.md
|
|
155
|
+
└── Project_001_API网关改造/
|
|
156
|
+
├── README.md
|
|
157
|
+
└── docs/
|
|
158
|
+
└── 当前状态与交接.md
|
|
159
|
+
```
|
|
160
|
+
|
|
161
|
+
## 命令一览
|
|
162
|
+
|
|
163
|
+
| 命令 | 作用 |
|
|
164
|
+
| --- | --- |
|
|
165
|
+
| `init [--lang zh\|en] [--no-git]` | 幂等初始化骨架文档、配置文件、`.gitignore` 与 Git(不覆盖已有文件) |
|
|
166
|
+
| `status [--guard]` | 查看进行中项目与 Git 状态;加 `--guard` 在有未提交改动时返回退出码 1 |
|
|
167
|
+
| `guard` | 轻量未提交巡检:干净返回 0,有未提交改动打印提醒并返回 1 |
|
|
168
|
+
| `exp search [关键词] [--tag 标签] [--list-tags]` | 按关键词、标签或组合检索经验,或列出全部标签统计 |
|
|
169
|
+
| `exp get EXP-001` | 查看指定经验的索引与正文详情(已下沉条目会提示目标路径) |
|
|
170
|
+
| `exp add --title ... [--tag a,b] [--sink P001]` | 沉淀新经验,自动分配连续编号、更新计数与索引(支持下沉到项目 docs) |
|
|
171
|
+
| `exp update EXP-001 --note ...` | 为已有经验追加新证据(同一问题不重复开条) |
|
|
172
|
+
| `project list` | 列出工作空间下所有项目及最大编号 |
|
|
173
|
+
| `project new <名称> --owner <agent>` | 按 SOP 建目录、生成 README 与交接文件,并自动登记 |
|
|
174
|
+
| `project update <编号> --status ...` | 同步更新项目 README 与全局登记表状态 |
|
|
175
|
+
| `handoff <编号> --agent ... --status ... --done ...` | 写入结构化交接,并原子同步 README 与登记表 |
|
|
176
|
+
| `commit -m <说明>` | 暂存并提交工作空间改动(自动排除临时锁文件 `.workspace.lock`) |
|
|
177
|
+
|
|
178
|
+
## 核心可靠性机制
|
|
179
|
+
|
|
180
|
+
1. **跨进程可重入文件锁(零外部依赖)**:所有写操作(`init` / `exp add` / `exp update` / `project new` / `project update` / `handoff` / `commit`)统一获取 `.workspace.lock` 排他锁(Windows 用 `msvcrt.locking`,Linux/macOS 用 `fcntl.flock`)。多 Agent 并发写自动排队串行,杜绝编号碰撞和 Markdown 撕裂;超时默认 10s(可通过 `AGENT_WORKSPACE_LOCK_TIMEOUT` 调整),且同进程嵌套调用可重入不死锁。
|
|
181
|
+
2. **脏工作树守卫(`guard` / `status --guard`)**:配合 shell hook 或 Agent 收尾约定,防止改完忘记 `commit`。
|
|
182
|
+
3. **标签不破坏原表格**:`--tag` 以 `#tag` 形式写入标题与正文元数据,保持原有 5 列 Markdown 表格完全向后兼容,人工 `grep` 同样直接命中。
|
|
183
|
+
|
|
184
|
+
## 接入各 Agent
|
|
185
|
+
|
|
186
|
+
把 [`examples/AGENTS-snippet.md`](examples/AGENTS-snippet.md) 里的短约定贴进各 Agent 的原生入口:
|
|
187
|
+
|
|
188
|
+
| Agent | 启动入口 |
|
|
189
|
+
| --- | --- |
|
|
190
|
+
| Codex | `~/.codex/AGENTS.md` |
|
|
191
|
+
| Claude Code | `~/.claude/CLAUDE.md` |
|
|
192
|
+
| DeepSeek Harness | `~/.dsh/AGENTS.md` |
|
|
193
|
+
| Antigravity | `<workspace>/.agents/skills/shared-workspace/SKILL.md`(见 [`examples/SKILL.md`](examples/SKILL.md)) |
|
|
194
|
+
|
|
195
|
+
> 原则:各 Agent 入口只放**什么时候调用哪个命令**的薄指针,不要把全局长文档整份塞进系统提示词,避免每轮开局白耗几千 token。
|
|
196
|
+
|
|
197
|
+
## 为什么不用向量数据库
|
|
198
|
+
|
|
199
|
+
| 维度 | 向量 RAG | 本方案 |
|
|
200
|
+
| --- | --- | --- |
|
|
201
|
+
| 可读性 | 召回碎片,人难核对来源 | 纯文本 Markdown,人直接读改 |
|
|
202
|
+
| 可修正性 | 改一条往往要重算 embedding / 重建索引 | 直接改一个段落或 `exp update` |
|
|
203
|
+
| 版本与回滚 | 需额外设计快照 | Git 原生 diff / blame / revert |
|
|
204
|
+
| 运行依赖 | 向量库 + embedding 模型/服务 | Python 标准库,零外部依赖 |
|
|
205
|
+
| 失败方式 | 静默召回不相关片段 | 查不到就是查不到,明确可见 |
|
|
206
|
+
|
|
207
|
+
代价也很透明:它只做关键词、标签与编号检索,不做语义相似度匹配。对**单人或小团队、几十到几百条经验**的本地多 Agent 协作非常顺手;超大规模知识库应考虑专业检索引擎。
|
|
208
|
+
|
|
209
|
+
## 测试与 CI
|
|
210
|
+
|
|
211
|
+
```bash
|
|
212
|
+
python -m unittest discover -s tests -v
|
|
213
|
+
cd integrations/deepseek-harness && npm run check
|
|
214
|
+
```
|
|
215
|
+
|
|
216
|
+
仓库内置 **15 项 Python 端到端测试 + 3 项 Node DSh 插件校验**(全部在临时目录运行,不碰真实工作区),在 GitHub Actions 上覆盖 **Ubuntu + Windows** 双平台矩阵:
|
|
217
|
+
- `init` 中英文骨架初始化与幂等性
|
|
218
|
+
- `exp add / update / search / --tag / --list-tags`
|
|
219
|
+
- `project new / handoff / status / guard / commit`
|
|
220
|
+
- **8 进程真实并发写压测**(验证文件锁串行化与编号连续无损)
|
|
221
|
+
- **锁持有超时退出验证**
|
|
222
|
+
- **MCP stdio JSON-RPC 全生命周期测试**(`initialize` / `tools/list` / `tools/call`)与多渠道清单同步校验
|
|
223
|
+
|
|
224
|
+
## 边界
|
|
225
|
+
|
|
226
|
+
- 面向单机、本地多 Agent 协作,不提供跨机器实时同步(跨机通过 Git push/pull 协作)。
|
|
227
|
+
- 不替代项目设计文档:交接只写当前状态、证据与下一步,详细方案留在项目 `docs/`。
|
|
228
|
+
- 不做黑盒自动提取:什么值得记由 Agent 或人判断,CLI 负责保证结构、编号、并发与版本一致。
|
|
229
|
+
|
|
230
|
+
## License
|
|
231
|
+
|
|
232
|
+
[MIT](LICENSE)
|