crewswarm 0.8.2-beta → 0.9.0
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.
package/README.md
CHANGED
|
@@ -1,315 +1,205 @@
|
|
|
1
1
|
# crewswarm
|
|
2
2
|
|
|
3
|
-
**
|
|
4
|
-
|
|
5
|
-
crewswarm is an open-source AI workspace for software development. It combines multi-agent orchestration, project-aware memory, local tool execution, chat surfaces, and editor/MCP integrations into one stack you can run yourself.
|
|
3
|
+
**The only multi-engine AI coding platform.** Switch between Claude Code, Cursor, Gemini, Codex, and OpenCode mid-conversation. Parallel agents. Persistent sessions. No vendor lock-in.
|
|
6
4
|
|
|
5
|
+
[](https://www.npmjs.com/package/crewswarm)
|
|
6
|
+
[]()
|
|
7
7
|
[](LICENSE)
|
|
8
8
|
[](https://nodejs.org)
|
|
9
9
|
[](https://crewswarm.ai)
|
|
10
|
+
[](https://github.com/sponsors/crewswarm)
|
|
10
11
|
|
|
11
12
|
|
|
12
13
|
|
|
13
14
|
---
|
|
14
15
|
|
|
15
|
-
##
|
|
16
|
-
|
|
17
|
-
Most AI dev tools are just a chat box bolted onto an editor.
|
|
18
|
-
|
|
19
|
-
**crewswarm** is different:
|
|
20
|
-
|
|
21
|
-
- **Local-first** — run it on your own machine
|
|
22
|
-
- **Real execution** — agents write files, run commands, and operate on actual projects
|
|
23
|
-
- **Multi-agent** — planner, coder, QA, fixer, security, GitHub, and more
|
|
24
|
-
- **Persistent context** — memory and session history survive beyond one chat
|
|
25
|
-
- **Multiple control surfaces** — dashboard, CLI, Telegram, SwiftBar, MCP/editor integrations
|
|
26
|
-
- **Model-flexible** — use Groq, OpenAI, Anthropic, Gemini, Mistral, DeepSeek, xAI, Ollama, and more
|
|
27
|
-
|
|
28
|
-
It is built for:
|
|
29
|
-
- solo builders
|
|
30
|
-
- AI-native dev teams
|
|
31
|
-
- local-first users who do not want SaaS lock-in
|
|
32
|
-
- people building real software with agent workflows, not toy demos
|
|
33
|
-
|
|
34
|
-
---
|
|
35
|
-
|
|
36
|
-
## Why it matters
|
|
37
|
-
|
|
38
|
-
Most “agent” tools still fake the important part.
|
|
16
|
+
## Why crewswarm?
|
|
39
17
|
|
|
40
|
-
|
|
18
|
+
**Rate limits are real.** Every $20/month AI coding plan has them. Claude, Cursor, Codex — you'll hit the wall mid-feature.
|
|
41
19
|
|
|
42
|
-
|
|
43
|
-
- write real files
|
|
44
|
-
- work across multiple steps
|
|
45
|
-
- keep project memory
|
|
46
|
-
- coordinate multiple specialists
|
|
47
|
-
- run locally without disappearing into someone else’s cloud
|
|
20
|
+
crewswarm is the only tool where you seamlessly switch to another engine and keep your session context. Or pick the best CLI for each job:
|
|
48
21
|
|
|
49
|
-
|
|
22
|
+
| Engine | Best for | Key strength |
|
|
23
|
+
|---|---|---|
|
|
24
|
+
| **Claude Code** | Large refactors, frontend | Full workspace context, session resume |
|
|
25
|
+
| **Cursor CLI** | Architecture, complex reasoning | Parallel waves, isolated contexts |
|
|
26
|
+
| **Gemini CLI** | Research, SEO, free fallback | Free: 60 req/min, Google Search built in |
|
|
27
|
+
| **Codex CLI** | Backend, fast iteration | Full sandbox, no approval prompts |
|
|
28
|
+
| **OpenCode** | Provider flexibility | Any model (Groq/DeepSeek/Ollama) |
|
|
29
|
+
| **crew-cli** | Orchestration, quality workflows | 20+ agents, sandbox, 3x parallel speedup |
|
|
50
30
|
|
|
51
31
|
---
|
|
52
32
|
|
|
53
33
|
## Quickstart
|
|
54
34
|
|
|
55
|
-
|
|
35
|
+
```bash
|
|
36
|
+
npm install -g crewswarm
|
|
37
|
+
crewswarm
|
|
38
|
+
```
|
|
56
39
|
|
|
57
|
-
|
|
58
|
-
- At least one LLM provider key for best results
|
|
59
|
-
- Groq is the fastest free starting point: [console.groq.com](https://console.groq.com)
|
|
40
|
+
That's it. Dashboard opens at `localhost:4319`, Vibe IDE at `localhost:3333`.
|
|
60
41
|
|
|
61
|
-
###
|
|
42
|
+
### Or install from source
|
|
62
43
|
|
|
63
44
|
```bash
|
|
64
45
|
git clone https://github.com/crewswarm/crewswarm
|
|
65
46
|
cd crewswarm
|
|
66
47
|
bash install.sh
|
|
67
|
-
```
|
|
68
|
-
|
|
69
|
-
**Fresh machine shortcut:**
|
|
70
|
-
```bash
|
|
71
|
-
bash <(curl -fsSL https://raw.githubusercontent.com/crewswarm/crewswarm/main/install.sh)
|
|
72
|
-
```
|
|
73
|
-
|
|
74
|
-
### Start
|
|
75
|
-
```bash
|
|
76
48
|
npm run doctor
|
|
77
|
-
npm
|
|
49
|
+
npm start
|
|
78
50
|
```
|
|
79
51
|
|
|
80
|
-
|
|
81
|
-
`http://127.0.0.1:4319`
|
|
52
|
+
### What does it cost?
|
|
82
53
|
|
|
83
|
-
|
|
54
|
+
**$0.** crewswarm is free and open source (MIT). You bring your own API keys — or use CLI OAuth (Claude, Cursor, Gemini login once, no keys needed). Free options: Gemini CLI (1,000 req/day), Groq (free tier), Ollama (fully local).
|
|
84
55
|
|
|
85
56
|
---
|
|
86
57
|
|
|
87
|
-
|
|
58
|
+
## What makes it different
|
|
88
59
|
|
|
89
|
-
|
|
90
|
-
|
|
91
|
-
|
|
92
|
-
|
|
93
|
-
|
|
94
|
-
|
|
95
|
-
|
|
96
|
-
|
|
97
|
-
-
|
|
98
|
-
|
|
99
|
-
This is not simulated. It works against real files and your real local workspace.
|
|
100
|
-
|
|
101
|
-
---
|
|
102
|
-
|
|
103
|
-
### Headless / non-interactive install
|
|
104
|
-
|
|
105
|
-
For Cursor, Codex, CI, or remote shell setups:
|
|
106
|
-
```bash
|
|
107
|
-
CREWSWARM_SETUP_MCP=1 \
|
|
108
|
-
CREWSWARM_START_NOW=1 \
|
|
109
|
-
bash install.sh --non-interactive
|
|
110
|
-
```
|
|
111
|
-
|
|
112
|
-
Optional flags:
|
|
113
|
-
- `CREWSWARM_BUILD_CREWCHAT=1`
|
|
114
|
-
- `CREWSWARM_SETUP_TELEGRAM=1`
|
|
115
|
-
- `CREWSWARM_SETUP_WHATSAPP=1`
|
|
116
|
-
- `CREWSWARM_ENABLE_AUTONOMOUS=1`
|
|
60
|
+
| Capability | crewswarm | Cursor | Windsurf | Devin | Copilot |
|
|
61
|
+
|---|---|---|---|---|---|
|
|
62
|
+
| Multi-engine (6 CLIs) | Yes | No | No | No | No |
|
|
63
|
+
| Native session resume | Yes | No | No | No | No |
|
|
64
|
+
| Parallel agent waves | Yes | No | No | Partial | No |
|
|
65
|
+
| Browser IDE + terminal | Vibe | Desktop | Desktop | Yes | Yes |
|
|
66
|
+
| 20+ specialist agents | Yes | 1 | 1 | 1 | 1 |
|
|
67
|
+
| PM Loop (autonomous roadmap) | Yes | No | No | Partial | No |
|
|
68
|
+
| Local-first / no cloud | Yes | Partial | No | No | No |
|
|
69
|
+
| Open source | Yes | No | No | No | No |
|
|
117
70
|
|
|
118
71
|
---
|
|
119
72
|
|
|
120
|
-
##
|
|
73
|
+
## How it works
|
|
121
74
|
|
|
122
|
-
**
|
|
123
|
-
|
|
124
|
-
|
|
125
|
-
**
|
|
126
|
-
Natural language requests are broken into structured tasks and routed to the right agent.
|
|
127
|
-
|
|
128
|
-
**Shared memory**
|
|
129
|
-
Project context persists through files like `brain.md`, `session-log.md`, `current-state.md`, and `orchestration-protocol.md`.
|
|
130
|
-
|
|
131
|
-
**Fault tolerance**
|
|
132
|
-
Retries, escalation to fixer agents, task leases, and dead-letter replay support are built in.
|
|
133
|
-
|
|
134
|
-
**Command approval gate**
|
|
135
|
-
Potentially risky shell commands require approval before execution.
|
|
136
|
-
|
|
137
|
-
**Multi-engine support**
|
|
138
|
-
Route work through different engines and environments, including Codex, Claude Code, Cursor, Gemini, OpenCode, and `crew-cli`.
|
|
139
|
-
|
|
140
|
-
**Multiple control surfaces**
|
|
141
|
-
Use crewswarm from the web dashboard, CLI, Telegram, macOS SwiftBar, or MCP/editor integrations.
|
|
142
|
-
|
|
143
|
-
---
|
|
144
|
-
|
|
145
|
-
## Why crewswarm vs other frameworks
|
|
146
|
-
|
|
147
|
-
| Feature | crewswarm | LangChain / LangGraph | AutoGen | CrewAI |
|
|
148
|
-
|---|---|---|---|---|
|
|
149
|
-
| Real file writes | ✅ | ⚠️ | ⚠️ | ⚠️ |
|
|
150
|
-
| PM-led planning | ✅ | ❌ | ❌ | ⚠️ |
|
|
151
|
-
| Persistent memory | ✅ | ⚠️ | ❌ | ⚠️ |
|
|
152
|
-
| Local-first | ✅ | ⚠️ | ⚠️ | ⚠️ |
|
|
153
|
-
| Built-in dashboard | ✅ | ❌ | ❌ | ❌ |
|
|
154
|
-
| Telegram / messaging bridges | ✅ | ❌ | ❌ | ❌ |
|
|
155
|
-
| Easy local startup | ✅ | ⚠️ | ⚠️ | ⚠️ |
|
|
156
|
-
|
|
157
|
-
---
|
|
158
|
-
|
|
159
|
-
## Main components
|
|
160
|
-
|
|
161
|
-
| Component | Purpose |
|
|
162
|
-
|---|---|
|
|
163
|
-
| `crew-lead` | conversational command layer and dispatcher |
|
|
164
|
-
| `crew-pm` | planning, task breakdown, roadmap management |
|
|
165
|
-
| `crew-coder` | implementation |
|
|
166
|
-
| `crew-qa` | testing and validation |
|
|
167
|
-
| `crew-fixer` | debugging and repair |
|
|
168
|
-
| `crew-security` | security review |
|
|
169
|
-
| `crew-github` | Git and repo actions |
|
|
170
|
-
| `dashboard` | browser-based control surface |
|
|
171
|
-
| `crew-cli` | command-line interface |
|
|
172
|
-
| `crew-scribe` | memory and task summarization |
|
|
173
|
-
|
|
174
|
-
---
|
|
175
|
-
|
|
176
|
-
## Architecture
|
|
75
|
+
1. **You write a requirement** — one sentence, one paragraph, or a full spec
|
|
76
|
+
2. **crew-pm plans it** — breaks work into phases, assigns specialists
|
|
77
|
+
3. **Agents execute in parallel** — backend, frontend, tests built simultaneously (3x faster)
|
|
78
|
+
4. **Done. Files on disk.** — real files, real tests, real output
|
|
177
79
|
|
|
178
80
|
```
|
|
179
|
-
Dashboard / Vibe / crew-cli / Telegram /
|
|
180
|
-
|
|
181
|
-
|
|
182
|
-
|
|
183
|
-
|
|
184
|
-
|
|
185
|
-
|
|
81
|
+
Dashboard / Vibe IDE / crew-cli / Telegram / MCP
|
|
82
|
+
|
|
|
83
|
+
crew-lead (router)
|
|
84
|
+
|
|
|
85
|
+
RT Bus
|
|
86
|
+
|
|
|
87
|
+
─────────────────────────────────
|
|
186
88
|
| | | | |
|
|
187
89
|
crew-pm coder qa fixer github
|
|
188
|
-
|
|
189
|
-
|
|
190
|
-
|
|
191
|
-
|
|
90
|
+
|
|
|
91
|
+
Code Engines: Claude · Cursor · Gemini · Codex · OpenCode · crew-cli
|
|
92
|
+
|
|
|
93
|
+
real files, commands, memory
|
|
192
94
|
```
|
|
193
95
|
|
|
194
96
|
---
|
|
195
97
|
|
|
196
|
-
##
|
|
98
|
+
## Surfaces
|
|
197
99
|
|
|
198
|
-
- **
|
|
199
|
-
- **
|
|
200
|
-
- **
|
|
201
|
-
- **
|
|
202
|
-
- **
|
|
203
|
-
- **Live smoke** (stack must be running): `npm run smoke`
|
|
204
|
-
- **Health check:** `npm run health`
|
|
205
|
-
- **Release preflight:** `npm run release:check`
|
|
100
|
+
- **Dashboard** — web control plane at `localhost:4319` (agents, engines, models, build, sessions)
|
|
101
|
+
- **Vibe IDE** — browser-based editor + terminal + chat at `localhost:3333`
|
|
102
|
+
- **crew-cli** — terminal-first with 34+ built-in tools
|
|
103
|
+
- **Telegram** — chat with your crew from your phone
|
|
104
|
+
- **MCP server** — plug crewswarm into any MCP-compatible editor
|
|
206
105
|
|
|
207
106
|
---
|
|
208
107
|
|
|
209
|
-
##
|
|
210
|
-
|
|
211
|
-
crewswarm stores config under your local `~/.crewswarm` directory.
|
|
212
|
-
|
|
213
|
-
Typical settings include provider API keys, agent model assignments, RT auth token, output path, and command allowlist.
|
|
108
|
+
## Per-agent model configuration
|
|
214
109
|
|
|
215
|
-
|
|
110
|
+
Every agent gets its own model. Use cheap models for routing, expensive for coding:
|
|
216
111
|
|
|
217
|
-
**Example agent config**
|
|
218
112
|
```json
|
|
219
113
|
{
|
|
220
114
|
"agents": [
|
|
221
|
-
{ "id": "crew-
|
|
115
|
+
{ "id": "crew-lead", "model": "groq/llama-3.3-70b-versatile" },
|
|
116
|
+
{ "id": "crew-pm", "model": "google/gemini-2.5-flash" },
|
|
222
117
|
{ "id": "crew-coder", "model": "anthropic/claude-sonnet-4-20250514" },
|
|
223
|
-
{ "id": "crew-qa", "model": "
|
|
118
|
+
{ "id": "crew-qa", "model": "google/gemini-2.5-flash" },
|
|
119
|
+
{ "id": "crew-fixer", "model": "openai/codex-mini-latest" }
|
|
224
120
|
]
|
|
225
121
|
}
|
|
226
122
|
```
|
|
227
123
|
|
|
228
|
-
|
|
229
|
-
|
|
230
|
-
## Deployment
|
|
124
|
+
Or skip API keys entirely — use Claude Code, Cursor, or Gemini CLI with OAuth login.
|
|
231
125
|
|
|
232
|
-
|
|
126
|
+
---
|
|
233
127
|
|
|
234
|
-
|
|
128
|
+
## The crew
|
|
235
129
|
|
|
236
|
-
|
|
130
|
+
| Agent | Role |
|
|
131
|
+
|---|---|
|
|
132
|
+
| `crew-lead` | Routes tasks, manages conversation |
|
|
133
|
+
| `crew-pm` | Plans, breaks down, prioritizes |
|
|
134
|
+
| `crew-coder` | Writes code (full-stack) |
|
|
135
|
+
| `crew-coder-back` | Backend specialist |
|
|
136
|
+
| `crew-coder-front` | Frontend specialist |
|
|
137
|
+
| `crew-qa` | Tests and validates |
|
|
138
|
+
| `crew-fixer` | Debugs and repairs |
|
|
139
|
+
| `crew-security` | Security review |
|
|
140
|
+
| `crew-github` | Git, PRs, branches |
|
|
141
|
+
| `crew-architect` | System design |
|
|
142
|
+
| `crew-orchestrator` | Wave dispatch |
|
|
143
|
+
| `crew-copywriter` | Docs and content |
|
|
144
|
+
| `crew-frontend` | UI/UX polish |
|
|
145
|
+
| `crew-main` | General coordination |
|
|
237
146
|
|
|
238
|
-
|
|
239
|
-
```bash
|
|
240
|
-
node -e "console.log(require('crypto').randomBytes(32).toString('hex'))"
|
|
241
|
-
```
|
|
147
|
+
---
|
|
242
148
|
|
|
243
|
-
|
|
149
|
+
## Built with crewswarm
|
|
244
150
|
|
|
245
|
-
**
|
|
151
|
+
- **VS Code extension** — full extension from prompt to package in 10 minutes
|
|
152
|
+
- **crewswarm.ai** — the production website you see, 90% built by the swarm in 30 minutes
|
|
153
|
+
- **Session resume** — native resume across 6 CLI engines, built in one session
|
|
246
154
|
|
|
247
|
-
|
|
155
|
+
---
|
|
248
156
|
|
|
249
|
-
|
|
157
|
+
## Commands
|
|
250
158
|
|
|
251
|
-
|
|
252
|
-
|
|
253
|
-
|
|
254
|
-
|
|
255
|
-
|
|
256
|
-
|
|
257
|
-
|
|
159
|
+
```bash
|
|
160
|
+
crewswarm # Start all services
|
|
161
|
+
crewswarm pm-loop # Run autonomous PM loop
|
|
162
|
+
npm run doctor # Preflight check
|
|
163
|
+
npm run restart-all # Restart the stack
|
|
164
|
+
npm test # Run 647 tests
|
|
165
|
+
crew exec "Build X" # Send task via CLI
|
|
258
166
|
```
|
|
259
167
|
|
|
260
|
-
|
|
168
|
+
---
|
|
261
169
|
|
|
262
|
-
|
|
170
|
+
## Deployment
|
|
263
171
|
|
|
264
|
-
|
|
172
|
+
```bash
|
|
173
|
+
# Docker
|
|
174
|
+
docker-compose up -d
|
|
265
175
|
|
|
266
|
-
|
|
267
|
-
|
|
268
|
-
|
|
269
|
-
- Rotate API keys and auth tokens regularly
|
|
270
|
-
- Run services with minimal required permissions
|
|
271
|
-
- Implement rate limiting on public endpoints
|
|
272
|
-
- Regular dependency updates for security patches
|
|
176
|
+
# Or the one-liner
|
|
177
|
+
bash <(curl -fsSL https://raw.githubusercontent.com/crewswarm/crewswarm/main/install.sh)
|
|
178
|
+
```
|
|
273
179
|
|
|
274
|
-
|
|
180
|
+
See [deploy.html](https://crewswarm.ai/deploy.html) for Docker, cloud VM, and production setup.
|
|
275
181
|
|
|
276
182
|
---
|
|
277
183
|
|
|
278
|
-
##
|
|
184
|
+
## Documentation
|
|
279
185
|
|
|
280
|
-
|
|
281
|
-
crewswarm/
|
|
282
|
-
|
|
283
|
-
|
|
284
|
-
|
|
285
|
-
|
|
286
|
-
|
|
287
|
-
├── unified-orchestrator.mjs
|
|
288
|
-
├── phased-orchestrator.mjs
|
|
289
|
-
├── continuous-build.mjs
|
|
290
|
-
├── scripts/
|
|
291
|
-
├── memory/
|
|
292
|
-
├── docs/
|
|
293
|
-
├── contrib/swiftbar/
|
|
294
|
-
└── website/
|
|
295
|
-
```
|
|
186
|
+
- [Website](https://crewswarm.ai)
|
|
187
|
+
- [Vibe IDE](https://crewswarm.ai/vibe.html)
|
|
188
|
+
- [crew-cli](https://crewswarm.ai/cli.html)
|
|
189
|
+
- [Models & Providers](https://crewswarm.ai/models.html)
|
|
190
|
+
- [Security](https://crewswarm.ai/security.html)
|
|
191
|
+
- [API Docs](https://crewswarm.ai/api.html)
|
|
192
|
+
- [@@Protocol](https://crewswarm.ai/atat.html)
|
|
296
193
|
|
|
297
194
|
---
|
|
298
195
|
|
|
299
|
-
##
|
|
196
|
+
## Contributing
|
|
300
197
|
|
|
301
|
-
|
|
302
|
-
- `docs/ARCHITECTURE.md`
|
|
303
|
-
- `docs/ORCHESTRATOR-GUIDE.md`
|
|
304
|
-
- `docs/SETUP-NEW-AGENTS.md`
|
|
305
|
-
- `docs/MODEL-RECOMMENDATIONS.md`
|
|
306
|
-
- `docs/TROUBLESHOOTING.md`
|
|
198
|
+
See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines. Issues and PRs welcome.
|
|
307
199
|
|
|
308
|
-
##
|
|
200
|
+
## Security
|
|
309
201
|
|
|
310
|
-
|
|
311
|
-
|
|
312
|
-
---
|
|
202
|
+
See [SECURITY.md](SECURITY.md) for reporting vulnerabilities.
|
|
313
203
|
|
|
314
204
|
## License
|
|
315
205
|
|
|
@@ -70,3 +70,9 @@
|
|
|
70
70
|
{"role":"assistant","content":"hello","ts":1774511218267,"source":"cli","agent":"crew-cli","metadata":{"engine":"crew-cli","exitCode":0,"agentName":"crew-cli","agentEmoji":"⚡"}}
|
|
71
71
|
{"role":"user","content":"[Currently open file: /Users/jeffhobbs/CrewSwarm/apps/vibe/server.mjs]\n\nAdd session resume for all CLI engines with --resume flag","ts":1774512864121,"source":"cli","metadata":{"engine":"claude","agentName":"You","agentEmoji":"👤"}}
|
|
72
72
|
{"role":"assistant","content":"Let me read the CLI engine command builder and the chat handler to understand the current session flow.Now let me check how the chat endpoint handles sessions:Now I understand the architecture. Each CLI engine gets spawned fresh per message. To add session resume, I need to:\n\n1. Track conversation IDs per engine (Claude uses `--resume`, Codex uses `--conversation-id`, Cursor uses session, etc.)\n2. Store the session/conversation ID when a CLI returns one\n3. Pass `--resume` on subsequent calls\n\nLet me check which CLIs support resume:Now let me check the output parsing to see if session IDs are extracted anywhere:Good — line 692 already detects `session id:` in output. Let me see that context:So `session id:` is currently being **filtered out as noise**. Now I have the full picture. Here's my plan:\n\n1. Add a `cliSessions` map to track engine → session/conversation IDs per project\n2. Modify `getCliCommand()` to accept and pass resume flags\n3. Extract session IDs from CLI output instead of filtering them\n4. Pass resume IDs on subsequent calls\n\nLet me implement this:**Step 1: Add the session store****Step 2: Update `getCliCommand()` to accept a `resumeSession` param and pass resume flags per engine****Step 3: Update `runCli()` to look up resume sessions and extract session IDs from output**Now I need to extract session IDs from the CLI output. Let me update the `close` handler to parse session IDs, and also stop filtering `session id:` from the output:Hmm, that IIFE inside a filter is fragile. Let me do it cleaner — extract the session ID in the line handler instead. Let me revert that and do it properly:Instead, I'll extract session IDs in the `runCli` output handler and save them on close. Let me find the right spot:**Step 4: Update the backward-compat `getCodexCommand` and `runCodexCli` aliases**Good. Now let me also add an API endpoint to check/clear resume sessions, and make sure the chat handler passes `resume` through:Let me add the resume sessions endpoint after the version endpoint:Now let me verify the full flow compiles correctly by checking the `runCli` function one more time:Good. Let me also verify the close handler stores the session properly:Let me verify there are no syntax errors by doing a quick parse check:Clean parse. Now let me also update the `handleCliChatViaCrewLead` function to pass the resume session through to crew-lead:","ts":1774513104176,"source":"cli","agent":"claude","metadata":{"engine":"claude","exitCode":1,"agentName":"claude","agentEmoji":"⚡"}}
|
|
73
|
+
{"role":"user","content":"what quote would you use for the entire repo? instead of this one from gemini-cli? Early access\nWhat builders are saying\n✦ Beta builder\n\"One sentence in. Working API out. No standups, no tickets, no back-and-forth.\"\nVlad\nFullstack Engineer · Early access\n✦ Beta builder\n\"I start the PM Loop, go eat dinner, come back to a finished feature. It's genuinely autonomous.\"\nDev\nStartup Founder · Early access\n✦ Beta builder\n\"Shared memory is the killer feature. I can stop mid-project and pick it up days later — agents just know where we left off.\"\nAttila\nIndie Developer · Early access","ts":1774516824165,"source":"cli","metadata":{"engine":"gemini","agentName":"You","agentEmoji":"👤"}}
|
|
74
|
+
{"role":"assistant","content":"I will read the `README.md` and `ARCHITECTURE.md` files to understand the core purpose and value proposition of the \"vibe\" project.Based on my research into the project's purpose and \"vibe,\" here is a quote that encapsulates the essence of the repository in the style of the beta builder testimonials:\n\n✦ **Beta builder**\n\"Vibe turns my local machine into a high-fidelity command center. I’m no longer just writing code; I’m orchestrating a crew in real-time with zero latency and a gorgeous UI that makes shipping feel effortless.\"\n**Jeff**\nLead Architect · Early access\n\n***\n\n### Why this fits the \"Vibe\" repo:\n* **Local-First Strength:** It highlights the \"real local project persistence\" and PTY-backed terminal mentioned in the `README.md`.\n* **Agent Integration:** It references the \"orchestration\" of the `crewswarm` specialist agents (the \"crew\").\n* **Aesthetic Focus:** It acknowledges the polished, modern CSS and \"dark mode\" design found in `index.html`.\n* **Flow & Speed:** It captures the \"zero latency\" feeling of a local IDE that integrates AI directly into the Monaco editor via `cli:codex`.","ts":1774516861356,"source":"cli","agent":"gemini","metadata":{"engine":"gemini","exitCode":0,"agentName":"gemini","agentEmoji":"⚡"}}
|
|
75
|
+
{"role":"user","content":"no i want a quote from Gemini - simple 1 liner","ts":1774516886109,"source":"cli","metadata":{"engine":"gemini","agentName":"You","agentEmoji":"👤"}}
|
|
76
|
+
{"role":"user","content":"Vibe turns my local machine into a high-fidelity command center. I’m no longer just writing code; I’m orchestrating a crew in real-time with zero latency and a gorgeous UI that makes shipping feel effortless.\" Gemini","ts":1774516924358,"source":"cli","metadata":{"engine":"gemini","agentName":"You","agentEmoji":"👤"}}
|
|
77
|
+
{"role":"assistant","content":"✦ **Gemini**\n\"Local files, global crew: the high-fidelity surface for shipping with a multi-agent pulse.\"","ts":1774516954600,"source":"cli","agent":"gemini","metadata":{"engine":"gemini","exitCode":0,"agentName":"gemini","agentEmoji":"⚡"}}
|
|
78
|
+
{"role":"assistant","content":"\"The local-first command center for orchestrating your AI crew with zero latency and high-fidelity flow.\" — Gemini","ts":1774516964996,"source":"cli","agent":"gemini","metadata":{"engine":"gemini","exitCode":0,"agentName":"gemini","agentEmoji":"⚡"}}
|
|
@@ -63,17 +63,18 @@ function esc(s) {
|
|
|
63
63
|
|
|
64
64
|
const frames = [
|
|
65
65
|
{
|
|
66
|
-
title: "Install —
|
|
66
|
+
title: "Install — real output from Docker",
|
|
67
67
|
lines: [
|
|
68
68
|
"$ npm install -g crewswarm",
|
|
69
69
|
"",
|
|
70
|
-
"added
|
|
70
|
+
"added 59 packages in 9s",
|
|
71
71
|
"",
|
|
72
|
-
"
|
|
72
|
+
"$ crewswarm --version",
|
|
73
|
+
"0.8.2-beta",
|
|
73
74
|
"",
|
|
74
75
|
"$ crewswarm",
|
|
75
76
|
"",
|
|
76
|
-
"⚡ crewswarm v0.8.
|
|
77
|
+
"⚡ crewswarm v0.8.2-beta",
|
|
77
78
|
"→ Dashboard: http://localhost:4319",
|
|
78
79
|
"→ Vibe IDE: http://localhost:3333",
|
|
79
80
|
"→ crew-lead: http://localhost:5010",
|
package/package.json
CHANGED