opc-agent 4.1.16 → 4.1.17
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/CHANGELOG.md +119 -48
- package/README.md +290 -933
- package/README.zh-CN.md +269 -497
- package/install.sh +7 -1
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -1,933 +1,290 @@
|
|
|
1
|
-
<div align="center">
|
|
2
|
-
|
|
3
|
-
# ⚡ OPC Agent
|
|
4
|
-
|
|
5
|
-
|
|
6
|
-
|
|
7
|
-
[](
|
|
9
|
-
[。
|
|
360
|
-
|
|
361
|
-
---
|
|
362
|
-
|
|
363
|
-
## 📦 四件套生态
|
|
364
|
-
|
|
365
|
-
| 包 | 功能 | 安装 |
|
|
366
|
-
|---|---|---|
|
|
367
|
-
| **[opc-agent](https://www.npmjs.com/package/opc-agent)** | Agent Runtime — 创建、运行、管理 | `npm i opc-agent` |
|
|
368
|
-
| **[deepbrain](https://www.npmjs.com/package/deepbrain)** | Agent Memory — 自进化记忆引擎 | `npm i deepbrain` |
|
|
369
|
-
| **[agentkits](https://www.npmjs.com/package/agentkits)** | Agent Model Layer — 一行代码接入 LLM + 自带记忆 | `npm i agentkits` |
|
|
370
|
-
| **[agent-workstation](https://www.npmjs.com/package/agent-workstation)** | Agent Templates — 100+ 专业角色模板 | `npm i agent-workstation` |
|
|
371
|
-
|
|
372
|
-
---
|
|
373
|
-
|
|
374
|
-
## 🔧 CLI 参考
|
|
375
|
-
|
|
376
|
-
| 命令 | 说明 |
|
|
377
|
-
|------|------|
|
|
378
|
-
| `opc init <name>` | 创建新 Agent（支持 `--role` 指定角色模板） |
|
|
379
|
-
| `opc chat` | 交互式 TUI 对话 |
|
|
380
|
-
| `opc dev` | 开发模式（热重载） |
|
|
381
|
-
| `opc run` | 生产模式运行 |
|
|
382
|
-
| `opc start` | 守护进程后台启动 |
|
|
383
|
-
| `opc stop` | 停止守护进程 |
|
|
384
|
-
| `opc status` | 查看运行状态 |
|
|
385
|
-
| `opc studio` | 启动可视化管理后台 |
|
|
386
|
-
| `opc doctor` | 环境检查与诊断 |
|
|
387
|
-
| `opc eval` | 运行评估测试 |
|
|
388
|
-
| `opc test` | 运行单元测试 |
|
|
389
|
-
| `opc build` | 构建 Agent |
|
|
390
|
-
| `opc publish` | 发布到 npm |
|
|
391
|
-
| `opc deploy` | 部署到云端 |
|
|
392
|
-
| `opc logs [-f]` | 查看 Traces 日志 |
|
|
393
|
-
| `opc traces` | 查看全链路追踪 |
|
|
394
|
-
| `opc score` | 查看性能评分 |
|
|
395
|
-
| `opc analytics` | 数据分析面板 |
|
|
396
|
-
| `opc brain` | 查看记忆状态 |
|
|
397
|
-
| `opc brain seed` | 查看/加载三层知识种子 |
|
|
398
|
-
| `opc brain evolve` | 触发知识进化（template → job → industry） |
|
|
399
|
-
| `opc brain status` | 查看 brain 详细统计 |
|
|
400
|
-
| `opc jobs` | 查看定时任务 |
|
|
401
|
-
| `opc skills` | 查看已学技能 |
|
|
402
|
-
| `opc search <query>` | 搜索 |
|
|
403
|
-
| `opc info` | Agent 信息 |
|
|
404
|
-
| `opc install <skill>` | 安装技能 |
|
|
405
|
-
| `opc plugin <name>` | 管理插件 |
|
|
406
|
-
| `opc tool <name>` | 管理工具 |
|
|
407
|
-
| `opc workflow <name>` | 工作流 |
|
|
408
|
-
| `opc migrate` | 迁移 |
|
|
409
|
-
|
|
410
|
-
---
|
|
411
|
-
|
|
412
|
-
## 🔌 协议支持
|
|
413
|
-
|
|
414
|
-
| 协议 | 角色 | 说明 |
|
|
415
|
-
|------|------|------|
|
|
416
|
-
| **[Google A2A](https://google.github.io/A2A/)** | Server + Client | Agent-to-Agent 互操作，发现/调用其他 Agent |
|
|
417
|
-
| **[AG-UI](https://ag-ui.com/)** | Server | Agent-to-UI 流式协议，前端实时渲染 Agent 状态 |
|
|
418
|
-
| **[MCP](https://modelcontextprotocol.io/)** | Server + Client | 连接外部工具服务器，也可作为工具提供方 |
|
|
419
|
-
|
|
420
|
-
---
|
|
421
|
-
|
|
422
|
-
## 📡 25 种 Channel
|
|
423
|
-
|
|
424
|
-
| 渠道 | 状态 | 说明 |
|
|
425
|
-
|------|:----:|------|
|
|
426
|
-
| 🌐 Web | ✅ | 网页聊天组件 |
|
|
427
|
-
| 📱 Telegram | ✅ | Bot API |
|
|
428
|
-
| 💬 Slack | ✅ | Slack App |
|
|
429
|
-
| 🎮 Discord | ✅ | Discord Bot |
|
|
430
|
-
| 📧 Email | ✅ | IMAP / SMTP |
|
|
431
|
-
| 💚 WeChat | ✅ | 企业微信 / 个人微信 |
|
|
432
|
-
| 🔵 Feishu | ✅ | 飞书机器人 |
|
|
433
|
-
| 🎤 Voice | ✅ | 语音通话（TTS + STT） |
|
|
434
|
-
| 🔌 WebSocket | ✅ | 实时双向通信 |
|
|
435
|
-
| 🪝 Webhook | ✅ | HTTP 回调 |
|
|
436
|
-
| 📡 REST API | ✅ | HTTP API |
|
|
437
|
-
| 💬 WhatsApp | ✅ | WhatsApp Business API |
|
|
438
|
-
| 🟢 LINE | ✅ | LINE Messaging API |
|
|
439
|
-
| 🟣 Teams | ✅ | Microsoft Teams Bot |
|
|
440
|
-
| 📲 SMS | ✅ | Twilio / 云通信 |
|
|
441
|
-
| 🔷 DingTalk | ✅ | 钉钉机器人 |
|
|
442
|
-
| 🐦 Twitter/X | ✅ | DM + Mentions |
|
|
443
|
-
| 📸 Instagram | ✅ | Instagram DM |
|
|
444
|
-
| 📘 Facebook | ✅ | Messenger API |
|
|
445
|
-
| 🔐 Signal | ✅ | Signal Bot |
|
|
446
|
-
| 🟩 Matrix | ✅ | Matrix 协议 |
|
|
447
|
-
| 💬 XMPP | ✅ | Jabber/XMPP |
|
|
448
|
-
| 💻 IRC | ✅ | IRC 协议 |
|
|
449
|
-
| 📞 Twilio | ✅ | 电话 / IVR |
|
|
450
|
-
| 🔧 Custom | ✅ | 自定义渠道适配器 |
|
|
451
|
-
|
|
452
|
-
---
|
|
453
|
-
|
|
454
|
-
## 🤝 贡献
|
|
455
|
-
|
|
456
|
-
欢迎贡献！请查看 [Contributing Guide](CONTRIBUTING.md)。
|
|
457
|
-
|
|
458
|
-
```bash
|
|
459
|
-
git clone https://github.com/Deepleaper/opc-agent.git
|
|
460
|
-
cd opc-agent
|
|
461
|
-
npm install
|
|
462
|
-
npm run build
|
|
463
|
-
npm test
|
|
464
|
-
```
|
|
465
|
-
|
|
466
|
-
---
|
|
467
|
-
|
|
468
|
-
## 📄 License
|
|
469
|
-
|
|
470
|
-
[Apache-2.0](LICENSE) © [Deepleaper](https://github.com/Deepleaper)
|
|
471
|
-
|
|
472
|
-
---
|
|
473
|
-
|
|
474
|
-
---
|
|
475
|
-
|
|
476
|
-
<a name="english-version"></a>
|
|
477
|
-
|
|
478
|
-
<div align="center">
|
|
479
|
-
|
|
480
|
-
# ⚡ OPC Agent
|
|
481
|
-
|
|
482
|
-
### The Self-Evolving Agent Runtime — Every Agent You Build Learns, Remembers, and Evolves
|
|
483
|
-
|
|
484
|
-
[](https://www.npmjs.com/package/opc-agent)
|
|
485
|
-
[](https://github.com/Deepleaper/opc-agent/blob/main/LICENSE)
|
|
486
|
-
[]()
|
|
487
|
-
[](https://www.typescriptlang.org/)
|
|
488
|
-
[](https://nodejs.org/)
|
|
489
|
-
|
|
490
|
-
**The runtime engine of The Self-Evolving Agent Stack — every Agent learns, remembers, and evolves.**
|
|
491
|
-
|
|
492
|
-
OPC Agent is the **Agent Runtime** in the self-evolving Agent stack: from `opc init` to create, `opc chat` to converse, `opc studio` to manage visually.<br>The built-in learn → recall → evolve loop makes your Agent **smarter over time** — not starting from scratch, but continuously accumulating and evolving knowledge.<br>Developer path: ① Pick a template (Agent Templates) → ② Connect models (Agent Model Layer) → ③ **Run it (Agent Runtime)** → ④ Auto-evolve (Agent Memory).
|
|
493
|
-
|
|
494
|
-
[Quick Start](#-quick-start) · [🧬 Self-Evolution](#-self-evolution) · [Features](#-features) · [Code Examples](#-code-examples-1) · [CLI Reference](#-cli-reference-1)
|
|
495
|
-
|
|
496
|
-
</div>
|
|
497
|
-
|
|
498
|
-
---
|
|
499
|
-
|
|
500
|
-
## 🚀 Quick Start
|
|
501
|
-
|
|
502
|
-
```bash
|
|
503
|
-
npm install -g opc-agent
|
|
504
|
-
opc init my-agent --role customer-service
|
|
505
|
-
cd my-agent && npm install
|
|
506
|
-
opc chat
|
|
507
|
-
```
|
|
508
|
-
|
|
509
|
-
```
|
|
510
|
-
🤖 Customer Service Agent ready
|
|
511
|
-
|
|
512
|
-
You: When will my order #12345 ship?
|
|
513
|
-
Agent: Hi! Order #12345 shipped this morning and should arrive within 3 days. Want me to check the tracking details?
|
|
514
|
-
|
|
515
|
-
You: What about the order you looked up last time?
|
|
516
|
-
Agent: Your previous order #12300 was delivered on April 15th at 2 PM.
|
|
517
|
-
↑ Memory auto-recalled — no need to repeat context
|
|
518
|
-
```
|
|
519
|
-
|
|
520
|
-
> The Agent automatically remembers every conversation, distills domain knowledge, and evolves its response strategy. This isn't a demo — it's the default behavior.
|
|
521
|
-
|
|
522
|
-
---
|
|
523
|
-
|
|
524
|
-
## ✨ Features
|
|
525
|
-
|
|
526
|
-
| | Feature | Description |
|
|
527
|
-
|---|---|---|
|
|
528
|
-
| 🧬 | **Self-Evolution** | learn → recall → evolve loop — Agents get smarter over time, knowledge auto-accumulates and refines |
|
|
529
|
-
| 🧠 | **Memory Evolution** | Built-in DeepBrain — auto distillation, clustering, deduplication, refinement |
|
|
530
|
-
| 🔧 | **20+ CLI Commands** | init / chat / run / start / studio / doctor / eval / traces / publish |
|
|
531
|
-
| 📡 | **25 Channels** | Telegram / Discord / Slack / WeChat / Feishu / Email / Web / WebSocket / Voice / Webhook / API / WhatsApp / LINE / Teams / SMS / DingTalk / Twitter / Instagram / Facebook / Signal / Matrix / XMPP / IRC / Twilio / Custom |
|
|
532
|
-
| 🔌 | **3 Protocols** | Google A2A + AG-UI + MCP (Server & Client) |
|
|
533
|
-
| 🎨 | **OPC Studio** | Visual management dashboard — one command `opc studio` |
|
|
534
|
-
| 📊 | **OpenTelemetry** | Full distributed tracing + p50 / p95 / p99 latency metrics |
|
|
535
|
-
| 🧪 | **Built-in Eval** | `opc eval` with 24 evaluation test cases for quantifying Agent quality |
|
|
536
|
-
| 🔍 | **RAG Pipeline** | 5 chunking strategies + 4 rerankers (via DeepBrain) |
|
|
537
|
-
| 📦 | **Package & Publish** | `opc publish` — one command to publish to npm |
|
|
538
|
-
| 🏭 | **Role Templates** | 100+ professional roles, `opc init --role` for instant creation |
|
|
539
|
-
| 🔒 | **Security Suite** | Command approval + API Key encryption + file/network restrictions + sandboxing |
|
|
540
|
-
| 🤖 | **Sub-Agents** | spawn / parallel / kill for multi-Agent collaboration |
|
|
541
|
-
| 🌐 | **Browser Automation** | Playwright integration — web interaction, screenshots, data scraping |
|
|
542
|
-
| 👁️ | **Vision** | Multimodal image understanding — screenshot analysis, document recognition |
|
|
543
|
-
| 🎙️ | **Voice Call** | Real-time voice conversation with TTS + STT |
|
|
544
|
-
| 🏠 | **Home Assistant** | Smart home integration, IoT device control |
|
|
545
|
-
| 💻 | **IDE Bridge** | Deep integration with VS Code / Cursor |
|
|
546
|
-
| 🔗 | **Node Network** | Multi-node network, cross-device Agent collaboration |
|
|
547
|
-
| 🚪 | **Gateway** | Unified gateway with secure routing + load balancing |
|
|
548
|
-
| 🗜️ | **Context Compression** | Smart context compression — keep key info in long conversations |
|
|
549
|
-
| 🌐 | **API Server** | REST API server for external system integration |
|
|
550
|
-
| 📎 | **@Mention** | @mention other Agents or resources |
|
|
551
|
-
| 📋 | **Session Manager** | Multi-session parallel, history rollback |
|
|
552
|
-
| 🎯 | **RL (Reinforcement Learning)** | Feedback-based strategy optimization |
|
|
553
|
-
|
|
554
|
-
---
|
|
555
|
-
|
|
556
|
-
## 🧬 Self-Evolution
|
|
557
|
-
|
|
558
|
-
**The core differentiator of OPC — self-evolution. This is what fundamentally separates us from every other Agent framework.**
|
|
559
|
-
|
|
560
|
-
Other frameworks build static Agents — they forget everything after each session, and when someone else takes over, everything resets. OPC Agents learn, remember, and evolve.
|
|
561
|
-
|
|
562
|
-
### 1. Every Agent You Build Self-Evolves
|
|
563
|
-
|
|
564
|
-
Through the learn → recall → evolve loop, Agents automatically accumulate knowledge and get smarter:
|
|
565
|
-
|
|
566
|
-
```
|
|
567
|
-
┌──────────────────────────────────────────────────────┐
|
|
568
|
-
│ │
|
|
569
|
-
│ 📥 learn 🔍 recall 🧬 evolve │
|
|
570
|
-
│ After each Auto-retrieve Periodically │
|
|
571
|
-
│ interaction relevant refine knowledge │
|
|
572
|
-
│ memories — sharper, not │
|
|
573
|
-
│ bigger │
|
|
574
|
-
│ │
|
|
575
|
-
│ Day 1: "I haven't seen this issue before" │
|
|
576
|
-
│ Day 7: "A similar issue was solved this way" │
|
|
577
|
-
│ Day 30: "The best practice for this type is..." │
|
|
578
|
-
│ │
|
|
579
|
-
└──────────────────────────────────────────────────────┘
|
|
580
|
-
```
|
|
581
|
-
|
|
582
|
-
**Concrete example**: A customer service Agent —
|
|
583
|
-
- **Day 1**: Can only respond based on the initial prompt; needs human escalation for unknown issues
|
|
584
|
-
- **Day 7**: Accumulated 200+ conversation experiences; `evolve()` distilled 15 product knowledge entries
|
|
585
|
-
- **Day 30**: Refined into 5 core insights + 50 best-practice answers; accuracy from 60% → 95%
|
|
586
|
-
|
|
587
|
-
```typescript
|
|
588
|
-
import { AgentRuntime } from 'opc-agent';
|
|
589
|
-
|
|
590
|
-
const runtime = new AgentRuntime('./agent.yaml');
|
|
591
|
-
await runtime.start();
|
|
592
|
-
|
|
593
|
-
// You don't need to do anything — the Agent automatically:
|
|
594
|
-
// 1. recall() — retrieves relevant past memories
|
|
595
|
-
// 2. respond() — generates better answers based on memory
|
|
596
|
-
// 3. learn() — stores this interaction into memory
|
|
597
|
-
// 4. evolve() — periodically distills scattered experiences into structured knowledge
|
|
598
|
-
```
|
|
599
|
-
|
|
600
|
-
### 2. The Open-Source Project Itself Self-Evolves
|
|
601
|
-
|
|
602
|
-
OPC isn't a static framework — it's a **living ecosystem**:
|
|
603
|
-
- 🌱 **Community-driven**: Every PR, Issue, and discussion pushes the suite forward
|
|
604
|
-
- 🔄 **Continuous iteration**: Daily integration of the latest AI techniques and best practices
|
|
605
|
-
- 📈 **Knowledge flywheel**: More Agents → richer experience → stronger framework → new Agents start smarter
|
|
606
|
-
|
|
607
|
-
---
|
|
608
|
-
|
|
609
|
-
## 🏗️ Architecture
|
|
610
|
-
|
|
611
|
-
```
|
|
612
|
-
┌──────────────────────────────────────────────────┐
|
|
613
|
-
│ OPC Studio (:4000) │
|
|
614
|
-
│ Visual Management · Monitoring · Debug │
|
|
615
|
-
├────────────┬────────────┬────────────┬───────────┤
|
|
616
|
-
│ DeepBrain │ AgentKits │Workstation │ OPC Core │
|
|
617
|
-
│ 🧠 Memory │ 📊 Models │ 👤 Roles │ ⚡ Engine │
|
|
618
|
-
│ │ │ │ │
|
|
619
|
-
│ learn() │ OpenAI │ 100+ roles │ 25 chan. │
|
|
620
|
-
│ recall() │ Anthropic │ YAML def. │ 3 proto. │
|
|
621
|
-
│ evolve() │ Ollama │ Skill sys. │ Cron sched│
|
|
622
|
-
│ RAG │ DeepSeek │ One-click │ Sub-Agent │
|
|
623
|
-
├────────────┴────────────┴────────────┴───────────┤
|
|
624
|
-
│ OpenTelemetry Tracing · Eval · Traces │
|
|
625
|
-
└──────────────────────────────────────────────────┘
|
|
626
|
-
```
|
|
627
|
-
|
|
628
|
-
---
|
|
629
|
-
|
|
630
|
-
## 💻 Code Examples
|
|
631
|
-
|
|
632
|
-
### 1. Minimal Agent (10 lines)
|
|
633
|
-
|
|
634
|
-
```typescript
|
|
635
|
-
import { BaseAgent, InMemoryStore } from 'opc-agent';
|
|
636
|
-
|
|
637
|
-
const agent = new BaseAgent({
|
|
638
|
-
name: 'my-agent',
|
|
639
|
-
systemPrompt: 'You are a helpful assistant.',
|
|
640
|
-
provider: 'ollama',
|
|
641
|
-
model: 'qwen2.5',
|
|
642
|
-
memory: new InMemoryStore(),
|
|
643
|
-
});
|
|
644
|
-
|
|
645
|
-
await agent.init();
|
|
646
|
-
const response = await agent.handleMessage({
|
|
647
|
-
id: '1', content: 'Hello!', sender: 'user',
|
|
648
|
-
channel: 'web', sessionId: 's1', timestamp: new Date(),
|
|
649
|
-
});
|
|
650
|
-
console.log(response.content);
|
|
651
|
-
```
|
|
652
|
-
|
|
653
|
-
### 2. Agent with Memory Evolution
|
|
654
|
-
|
|
655
|
-
```typescript
|
|
656
|
-
import { AgentRuntime } from 'opc-agent';
|
|
657
|
-
|
|
658
|
-
const runtime = new AgentRuntime('./agent.yaml');
|
|
659
|
-
await runtime.start();
|
|
660
|
-
// Auto: recall(history) → respond(generate) → learn(distill) → evolve(improve)
|
|
661
|
-
```
|
|
662
|
-
|
|
663
|
-
### 3. Multi-Protocol Agent (agent.yaml)
|
|
664
|
-
|
|
665
|
-
```yaml
|
|
666
|
-
id: smart-assistant
|
|
667
|
-
name: Smart Assistant
|
|
668
|
-
version: "1.0.0"
|
|
669
|
-
model: deepseek-chat
|
|
670
|
-
|
|
671
|
-
channels:
|
|
672
|
-
- type: web
|
|
673
|
-
port: 3000
|
|
674
|
-
- type: telegram
|
|
675
|
-
token: ${TELEGRAM_BOT_TOKEN}
|
|
676
|
-
|
|
677
|
-
protocols:
|
|
678
|
-
a2a:
|
|
679
|
-
enabled: true
|
|
680
|
-
port: 4001
|
|
681
|
-
ag-ui:
|
|
682
|
-
enabled: true
|
|
683
|
-
port: 4002
|
|
684
|
-
mcp:
|
|
685
|
-
role: both # server + client
|
|
686
|
-
port: 4003
|
|
687
|
-
servers:
|
|
688
|
-
- name: file-tools
|
|
689
|
-
command: npx @modelcontextprotocol/server-filesystem
|
|
690
|
-
|
|
691
|
-
memory:
|
|
692
|
-
shortTerm: true
|
|
693
|
-
longTerm:
|
|
694
|
-
provider: deepbrain
|
|
695
|
-
autoEvolve: true
|
|
696
|
-
```
|
|
697
|
-
|
|
698
|
-
---
|
|
699
|
-
|
|
700
|
-
## 🔄 Dual-Loop Knowledge System
|
|
701
|
-
|
|
702
|
-
OPC Agent's knowledge system consists of two loops:
|
|
703
|
-
|
|
704
|
-
```
|
|
705
|
-
┌─────────────────────────────────────────────────────┐
|
|
706
|
-
│ Small Loop (Local, Free) │
|
|
707
|
-
│ Agent local learn → recall → evolve │
|
|
708
|
-
│ Works offline, data stays on your machine │
|
|
709
|
-
│ │
|
|
710
|
-
│ Big Loop (Hub, Value-Add) │
|
|
711
|
-
│ Agent ↔ Workstation Hub knowledge sharing │
|
|
712
|
-
│ Collective wisdom > individual experience │
|
|
713
|
-
└─────────────────────────────────────────────────────┘
|
|
714
|
-
```
|
|
715
|
-
|
|
716
|
-
**Local is the owner, Hub is the helper** — works without internet, auto-syncs when connected.
|
|
717
|
-
|
|
718
|
-
```
|
|
719
|
-
agentkits (Model Layer) — LLM calls
|
|
720
|
-
↕
|
|
721
|
-
opc-agent (Runtime) — run Agents (local) ← You are here
|
|
722
|
-
↕
|
|
723
|
-
deepbrain (Memory Engine) — store knowledge (engine)
|
|
724
|
-
↕
|
|
725
|
-
agent-workstation (Knowledge Platform) — knowledge lifecycle (Hub)
|
|
726
|
-
```
|
|
727
|
-
|
|
728
|
-
---
|
|
729
|
-
|
|
730
|
-
## 🌱 Brain Seed Auto-Loading
|
|
731
|
-
|
|
732
|
-
Since v2.1.0, OPC Agent supports automatic 3-tier knowledge seed loading. When a `brain-seeds/` directory exists, the Agent auto-imports industry → job → workstation knowledge on first run:
|
|
733
|
-
|
|
734
|
-
```typescript
|
|
735
|
-
const agent = new BaseAgent({
|
|
736
|
-
name: 'my-agent',
|
|
737
|
-
systemPrompt: 'You are a helpful assistant.',
|
|
738
|
-
provider: 'deepseek',
|
|
739
|
-
model: 'deepseek-chat',
|
|
740
|
-
});
|
|
741
|
-
agent.setLongTermMemory(brain);
|
|
742
|
-
// Auto-seeds on first run if brain-seeds/ directory exists
|
|
743
|
-
```
|
|
744
|
-
|
|
745
|
-
Manage brain seeds via CLI:
|
|
746
|
-
|
|
747
|
-
```bash
|
|
748
|
-
opc brain seed # View current brain seed status
|
|
749
|
-
opc brain seed --load # Manually load/reload seeds
|
|
750
|
-
opc brain evolve # Trigger knowledge evolution (workstation → job → industry)
|
|
751
|
-
opc brain status # View brain statistics
|
|
752
|
-
```
|
|
753
|
-
|
|
754
|
-
---
|
|
755
|
-
|
|
756
|
-
## 🔍 How OPC Agent Differs from Other Frameworks
|
|
757
|
-
|
|
758
|
-
> **In one sentence: other Agents are static; OPC Agents self-evolve.**
|
|
759
|
-
|
|
760
|
-
OPC Agent's core design philosophy centers on **Self-Evolution + Omnichannel + Role Templates + Production Toolchain**. As the **Agent Runtime** in The Self-Evolving Agent Stack, OPC focuses on how a single agent can **continuously self-evolve, reach users everywhere, and ship to production out of the box**.
|
|
761
|
-
|
|
762
|
-
**🔄 Self-Evolution Flywheel**: Templates ship with Brain Seeds → Model Layer auto-learns on every call → Runtime runs continuously → Memory auto-evolves → Agent gets stronger over time.
|
|
763
|
-
|
|
764
|
-
#### vs Hermes Agent
|
|
765
|
-
|
|
766
|
-
**Design difference**: Hermes is a Python monolithic framework built around the Nous Research model ecosystem, emphasizing self-learning evolution and RL training. OPC is a TypeScript modular four-piece suite (Brain / Agent / Channel / CLI) emphasizing engineering productivity and out-of-box experience.
|
|
767
|
-
|
|
768
|
-
**Where OPC leads**: Memory evolution trilogy (learn → recall → evolve with automatic clustering), 3-tier Brain Seed knowledge system, 25 built-in channels (vs Hermes's 16+), 100+ role templates, native A2A + AG-UI protocol support, OpenAI-compatible API server.
|
|
769
|
-
|
|
770
|
-
**Where Hermes leads**: More mature Python ecosystem, battle-tested GRPO + LoRA RL training pipeline, larger Nous Research community, more mature Docker sandbox isolation.
|
|
771
|
-
|
|
772
|
-
**Best for**: Choose Hermes if you need Python ecosystem and RL training capabilities. Choose OPC if you need TypeScript full-stack, memory evolution, omnichannel reach, and production toolchain.
|
|
773
|
-
|
|
774
|
-
#### vs OpenClaw
|
|
775
|
-
|
|
776
|
-
**Design difference**: OpenClaw is a heavyweight runtime with config-driven approach — install once, configure, and run. OPC is lightweight CLI + code-first, with four independent packages you can mix and match.
|
|
777
|
-
|
|
778
|
-
**Where OPC leads**: Modular four-piece suite with independent packages, 100+ role templates, memory evolution system, 25 built-in channels, `opc init/dev/test/deploy` full-lifecycle CLI, A2A/AG-UI protocol support.
|
|
779
|
-
|
|
780
|
-
**Where OpenClaw leads**: Polished out-of-box runtime experience, Puppeteer browser automation, mature permission and security sandbox, lower barrier to entry with config-driven approach.
|
|
781
|
-
|
|
782
|
-
**Best for**: Choose OpenClaw if you prefer a "configure and run" all-in-one solution. Choose OPC if you prefer code-first, modular composition, and need memory evolution with role templates.
|
|
783
|
-
|
|
784
|
-
#### vs CrewAI
|
|
785
|
-
|
|
786
|
-
**Design difference**: CrewAI uses a Crew orchestration pattern — define agent roles, assign tasks, execute in sequence or parallel. OPC uses a Template + Memory model — each agent has independent memory and role templates, accumulating and evolving knowledge through DeepBrain.
|
|
787
|
-
|
|
788
|
-
**Where OPC leads**: Memory evolution (not just short/long-term memory, but automatic clustering and distillation), Brain Seed pre-loaded knowledge, 25 built-in channels (CrewAI requires DIY), 100+ role templates, native TypeScript.
|
|
789
|
-
|
|
790
|
-
**Where CrewAI leads**: Intuitive Crew + Flow orchestration model, Python ecosystem, 100K+ user mature community, feature-rich Enterprise edition.
|
|
791
|
-
|
|
792
|
-
**Best for**: Choose CrewAI if you need multi-agent task orchestration and prefer Python. Choose OPC if you need deep single-agent evolution, omnichannel deployment, and TypeScript ecosystem.
|
|
793
|
-
|
|
794
|
-
#### vs AutoGen
|
|
795
|
-
|
|
796
|
-
**Design difference**: AutoGen is Microsoft's conversation-driven multi-agent framework, centered on agent-to-agent dialogue and UserProxy human-in-the-loop. OPC is task-driven with a template-based model, emphasizing single-agent memory evolution and production deployment.
|
|
797
|
-
|
|
798
|
-
**Where OPC leads**: Memory evolution system, Brain Seed, 25 built-in channels, 100+ role templates, full-lifecycle CLI, YAML declarative configuration.
|
|
799
|
-
|
|
800
|
-
**Where AutoGen leads**: Microsoft ecosystem support, distributed agent capabilities, AutoGen Studio visualization, mature UserProxy human-in-the-loop pattern, Python/C# dual-language support, active community.
|
|
801
|
-
|
|
802
|
-
**Best for**: Choose AutoGen if you need multi-agent conversational collaboration and Microsoft ecosystem integration. Choose OPC if you need memory evolution, omnichannel reach, role templates, and TypeScript toolchain.
|
|
803
|
-
|
|
804
|
-
#### Quick Reference Matrix
|
|
805
|
-
|
|
806
|
-
| Feature | OPC Agent | Hermes Agent | OpenClaw | CrewAI | AutoGen |
|
|
807
|
-
|---|:-:|:-:|:-:|:-:|:-:|
|
|
808
|
-
| **Language** | TypeScript | Python | TypeScript | Python | Python/C# |
|
|
809
|
-
| **Self-Evolution** | ✅ learn→recall→evolve | 🔶 | ❌ | ❌ | ❌ |
|
|
810
|
-
| **Brain Seed** | ✅ 3-tier | ❌ | ❌ | ❌ | ❌ |
|
|
811
|
-
| **Built-in Channels** | 25 | 16+ | 2 | ❌ | ❌ |
|
|
812
|
-
| **Role Templates** | 100+ | ❌ | ❌ | ❌ | ❌ |
|
|
813
|
-
| **Full-lifecycle CLI** | ✅ 20+ cmds | 🔶 | 🔶 | 🔶 | 🔶 |
|
|
814
|
-
| **A2A + AG-UI** | ✅ | 🔶 / ❌ | ❌ | ❌ | ❌ |
|
|
815
|
-
| **Multi-Agent** | ✅ | ✅ | 🔶 | ✅ | ✅ |
|
|
816
|
-
| **RL Training** | ✅ Feedback opt. | ✅ GRPO+LoRA | ❌ | ❌ | ❌ |
|
|
817
|
-
| **Browser Automation** | ✅ | ✅ | ✅ | ❌ | ❌ |
|
|
818
|
-
| **Visual Dashboard** | ✅ OPC Studio | ✅ Dashboard | ❌ | 🔶 | ✅ Studio |
|
|
819
|
-
| **Community Maturity** | 🚧 Early | ✅ Active | 🚧 Niche | ✅ Large | ✅ Large |
|
|
820
|
-
| **License** | Apache-2.0 | MIT | MIT | Apache-2.0 | MIT |
|
|
821
|
-
|
|
822
|
-
> Comparison based on each project's public documentation as of April 2026. All frameworks are evolving rapidly — corrections welcome via [Issues](https://github.com/Deepleaper/opc-agent/issues).
|
|
823
|
-
|
|
824
|
-
---
|
|
825
|
-
|
|
826
|
-
## 📦 Ecosystem
|
|
827
|
-
|
|
828
|
-
| Package | Purpose | Install |
|
|
829
|
-
|---|---|---|
|
|
830
|
-
| **[opc-agent](https://www.npmjs.com/package/opc-agent)** | Agent Runtime — create, run, manage | `npm i opc-agent` |
|
|
831
|
-
| **[deepbrain](https://www.npmjs.com/package/deepbrain)** | Agent Memory — self-evolving memory engine | `npm i deepbrain` |
|
|
832
|
-
| **[agentkits](https://www.npmjs.com/package/agentkits)** | Agent Model Layer — one-line LLM access with built-in memory | `npm i agentkits` |
|
|
833
|
-
| **[agent-workstation](https://www.npmjs.com/package/agent-workstation)** | Agent Templates — 100+ professional role templates | `npm i agent-workstation` |
|
|
834
|
-
|
|
835
|
-
---
|
|
836
|
-
|
|
837
|
-
## 🔧 CLI Reference
|
|
838
|
-
|
|
839
|
-
| Command | Description |
|
|
840
|
-
|---------|-------------|
|
|
841
|
-
| `opc init <name>` | Create a new Agent (supports `--role` for role templates) |
|
|
842
|
-
| `opc chat` | Interactive TUI conversation |
|
|
843
|
-
| `opc dev` | Development mode (hot reload) |
|
|
844
|
-
| `opc run` | Production run |
|
|
845
|
-
| `opc start` | Daemon mode (background) |
|
|
846
|
-
| `opc stop` | Stop daemon |
|
|
847
|
-
| `opc status` | View running status |
|
|
848
|
-
| `opc studio` | Launch visual management dashboard |
|
|
849
|
-
| `opc doctor` | Environment check & diagnostics |
|
|
850
|
-
| `opc eval` | Run evaluation tests |
|
|
851
|
-
| `opc test` | Run unit tests |
|
|
852
|
-
| `opc build` | Build Agent |
|
|
853
|
-
| `opc publish` | Publish to npm |
|
|
854
|
-
| `opc deploy` | Deploy to cloud |
|
|
855
|
-
| `opc logs [-f]` | View Traces logs |
|
|
856
|
-
| `opc traces` | View distributed traces |
|
|
857
|
-
| `opc score` | View performance score |
|
|
858
|
-
| `opc analytics` | Analytics dashboard |
|
|
859
|
-
| `opc brain` | View memory status |
|
|
860
|
-
| `opc brain seed` | View/load 3-tier knowledge seeds |
|
|
861
|
-
| `opc brain evolve` | Trigger knowledge evolution (workstation → job → industry) |
|
|
862
|
-
| `opc brain status` | View brain detailed statistics |
|
|
863
|
-
| `opc jobs` | View scheduled tasks |
|
|
864
|
-
| `opc skills` | View learned skills |
|
|
865
|
-
| `opc search <query>` | Search |
|
|
866
|
-
| `opc info` | Agent info |
|
|
867
|
-
| `opc install <skill>` | Install a skill |
|
|
868
|
-
| `opc plugin <name>` | Manage plugins |
|
|
869
|
-
| `opc tool <name>` | Manage tools |
|
|
870
|
-
| `opc workflow <name>` | Workflows |
|
|
871
|
-
| `opc migrate` | Migration |
|
|
872
|
-
|
|
873
|
-
---
|
|
874
|
-
|
|
875
|
-
## 🔌 Protocol Support
|
|
876
|
-
|
|
877
|
-
| Protocol | Role | Description |
|
|
878
|
-
|----------|------|-------------|
|
|
879
|
-
| **[Google A2A](https://google.github.io/A2A/)** | Server + Client | Agent-to-Agent interop — discover and invoke other Agents |
|
|
880
|
-
| **[AG-UI](https://ag-ui.com/)** | Server | Agent-to-UI streaming protocol — real-time frontend rendering |
|
|
881
|
-
| **[MCP](https://modelcontextprotocol.io/)** | Server + Client | Connect to external tool servers or serve as a tool provider |
|
|
882
|
-
|
|
883
|
-
---
|
|
884
|
-
|
|
885
|
-
## 📡 25 Channels
|
|
886
|
-
|
|
887
|
-
| Channel | Status | Description |
|
|
888
|
-
|---------|:------:|-------------|
|
|
889
|
-
| 🌐 Web | ✅ | Web chat widget |
|
|
890
|
-
| 📱 Telegram | ✅ | Bot API |
|
|
891
|
-
| 💬 Slack | ✅ | Slack App |
|
|
892
|
-
| 🎮 Discord | ✅ | Discord Bot |
|
|
893
|
-
| 📧 Email | ✅ | IMAP / SMTP |
|
|
894
|
-
| 💚 WeChat | ✅ | Enterprise / Personal WeChat |
|
|
895
|
-
| 🔵 Feishu | ✅ | Feishu (Lark) Bot |
|
|
896
|
-
| 🎤 Voice | ✅ | Voice call (TTS + STT) |
|
|
897
|
-
| 🔌 WebSocket | ✅ | Real-time bidirectional |
|
|
898
|
-
| 🪝 Webhook | ✅ | HTTP callback |
|
|
899
|
-
| 📡 REST API | ✅ | HTTP API |
|
|
900
|
-
| 💬 WhatsApp | ✅ | WhatsApp Business API |
|
|
901
|
-
| 🟢 LINE | ✅ | LINE Messaging API |
|
|
902
|
-
| 🟣 Teams | ✅ | Microsoft Teams Bot |
|
|
903
|
-
| 📲 SMS | ✅ | Twilio / Cloud messaging |
|
|
904
|
-
| 🔷 DingTalk | ✅ | DingTalk Bot |
|
|
905
|
-
| 🐦 Twitter/X | ✅ | DM + Mentions |
|
|
906
|
-
| 📸 Instagram | ✅ | Instagram DM |
|
|
907
|
-
| 📘 Facebook | ✅ | Messenger API |
|
|
908
|
-
| 🔐 Signal | ✅ | Signal Bot |
|
|
909
|
-
| 🟩 Matrix | ✅ | Matrix protocol |
|
|
910
|
-
| 💬 XMPP | ✅ | Jabber/XMPP |
|
|
911
|
-
| 💻 IRC | ✅ | IRC protocol |
|
|
912
|
-
| 📞 Twilio | ✅ | Phone / IVR |
|
|
913
|
-
| 🔧 Custom | ✅ | Custom channel adapter |
|
|
914
|
-
|
|
915
|
-
---
|
|
916
|
-
|
|
917
|
-
## 🤝 Contributing
|
|
918
|
-
|
|
919
|
-
Contributions welcome! See [Contributing Guide](CONTRIBUTING.md).
|
|
920
|
-
|
|
921
|
-
```bash
|
|
922
|
-
git clone https://github.com/Deepleaper/opc-agent.git
|
|
923
|
-
cd opc-agent
|
|
924
|
-
npm install
|
|
925
|
-
npm run build
|
|
926
|
-
npm test
|
|
927
|
-
```
|
|
928
|
-
|
|
929
|
-
---
|
|
930
|
-
|
|
931
|
-
## 📄 License
|
|
932
|
-
|
|
933
|
-
[Apache-2.0](LICENSE) © [Deepleaper](https://github.com/Deepleaper)
|
|
1
|
+
<div align="center">
|
|
2
|
+
|
|
3
|
+
# ⚡ OPC Agent
|
|
4
|
+
|
|
5
|
+
**Open Agent Framework — Build, run, and evolve AI agents from your terminal.**
|
|
6
|
+
|
|
7
|
+
[](https://www.npmjs.com/package/opc-agent)
|
|
8
|
+
[](LICENSE)
|
|
9
|
+
[](https://nodejs.org/)
|
|
10
|
+
|
|
11
|
+
[Quick Start](#-quick-start) · [Features](#-features) · [Architecture](#-architecture) · [Configuration](#-configuration) · [CLI](#-cli-commands) · [中文文档](README.zh-CN.md)
|
|
12
|
+
|
|
13
|
+
</div>
|
|
14
|
+
|
|
15
|
+
---
|
|
16
|
+
|
|
17
|
+
## 🚀 Quick Start
|
|
18
|
+
|
|
19
|
+
```bash
|
|
20
|
+
npm install -g opc-agent
|
|
21
|
+
opc init
|
|
22
|
+
opc run
|
|
23
|
+
```
|
|
24
|
+
|
|
25
|
+
Or one-line install (includes Node.js + optional Ollama):
|
|
26
|
+
|
|
27
|
+
```bash
|
|
28
|
+
# macOS / Linux
|
|
29
|
+
curl -fsSL https://raw.githubusercontent.com/Deepleaper/opc-agent/main/install.sh | bash
|
|
30
|
+
|
|
31
|
+
# Windows PowerShell
|
|
32
|
+
irm https://raw.githubusercontent.com/Deepleaper/opc-agent/main/install.ps1 | iex
|
|
33
|
+
```
|
|
34
|
+
|
|
35
|
+
Then open Studio at **http://localhost:4000** or chat in terminal:
|
|
36
|
+
|
|
37
|
+
```bash
|
|
38
|
+
opc chat
|
|
39
|
+
```
|
|
40
|
+
|
|
41
|
+
---
|
|
42
|
+
|
|
43
|
+
## ✨ Features
|
|
44
|
+
|
|
45
|
+
| | Feature | Description |
|
|
46
|
+
|---|---|---|
|
|
47
|
+
| 🤖 | **53 Built-in Tools** | File, shell, web, browser, vision, GitHub, Jira, Slack, and more |
|
|
48
|
+
| 🎨 | **Studio GUI** | Visual agent management at `http://localhost:4000` |
|
|
49
|
+
| 💬 | **TUI Terminal Chat** | Streaming responses, markdown rendering, slash commands |
|
|
50
|
+
| 🧠 | **Knowledge Evolve Engine** | Local Ollama-powered zero-cost knowledge distillation |
|
|
51
|
+
| 📱 | **15+ Channels** | Telegram, Discord, Slack, WeChat, Email, WhatsApp, LINE, Teams… |
|
|
52
|
+
| 🔧 | **40 Built-in Skills** | Productivity, knowledge, creative, developer skill packs |
|
|
53
|
+
| 📋 | **OAD Config** | Declarative YAML agent definition (`oad.yaml`) |
|
|
54
|
+
| 🏥 | **Doctor** | 13 health checks — model, tools, channels, memory |
|
|
55
|
+
| ⏰ | **Cron Scheduler** | Scheduled tasks + proactive agent triggers |
|
|
56
|
+
| 🔌 | **MCP Protocol** | Model Context Protocol server & client |
|
|
57
|
+
| 🗣️ | **Voice (STT/TTS)** | Whisper, Azure Speech, Volcano Engine |
|
|
58
|
+
| 📊 | **A2A Protocol** | Google Agent-to-Agent interoperability |
|
|
59
|
+
|
|
60
|
+
---
|
|
61
|
+
|
|
62
|
+
## 🏗️ Architecture
|
|
63
|
+
|
|
64
|
+
```
|
|
65
|
+
┌─────────────────────────────────────────────────┐
|
|
66
|
+
│ Channels │
|
|
67
|
+
│ Telegram · Discord · Slack · WeChat · Email … │
|
|
68
|
+
└──────────────────────┬──────────────────────────┘
|
|
69
|
+
│
|
|
70
|
+
┌──────────────────────▼──────────────────────────┐
|
|
71
|
+
│ OPC Agent Runtime │
|
|
72
|
+
│ ┌──────────┐ ┌──────────┐ ┌──────────────┐ │
|
|
73
|
+
│ │ Tool Exec│ │ Skills │ │ Memory/Know │ │
|
|
74
|
+
│ │ (53 tools)│ │(40 skills)│ │ Evolve Engine│ │
|
|
75
|
+
│ └──────────┘ └──────────┘ └──────────────┘ │
|
|
76
|
+
│ ┌──────────┐ ┌──────────┐ ┌──────────────┐ │
|
|
77
|
+
│ │ Cron │ │ Voice │ │ MCP / A2A │ │
|
|
78
|
+
│ └──────────┘ └──────────┘ └──────────────┘ │
|
|
79
|
+
└──────────────────────┬──────────────────────────┘
|
|
80
|
+
│
|
|
81
|
+
┌──────────────────────▼──────────────────────────┐
|
|
82
|
+
│ LLM Providers │
|
|
83
|
+
│ OpenAI · Anthropic · Ollama · Azure · Gemini │
|
|
84
|
+
└─────────────────────────────────────────────────┘
|
|
85
|
+
```
|
|
86
|
+
|
|
87
|
+
---
|
|
88
|
+
|
|
89
|
+
## ⚙️ Configuration
|
|
90
|
+
|
|
91
|
+
Agents are defined with a single `oad.yaml` file:
|
|
92
|
+
|
|
93
|
+
```yaml
|
|
94
|
+
name: my-agent
|
|
95
|
+
description: Customer support agent
|
|
96
|
+
model:
|
|
97
|
+
provider: ollama
|
|
98
|
+
model: llama3.1
|
|
99
|
+
channels:
|
|
100
|
+
- type: telegram
|
|
101
|
+
token: ${TELEGRAM_BOT_TOKEN}
|
|
102
|
+
- type: web
|
|
103
|
+
port: 4000
|
|
104
|
+
tools:
|
|
105
|
+
- file
|
|
106
|
+
- shell
|
|
107
|
+
- web-fetch
|
|
108
|
+
- browser
|
|
109
|
+
skills:
|
|
110
|
+
- weather
|
|
111
|
+
- github
|
|
112
|
+
memory:
|
|
113
|
+
provider: sqlite
|
|
114
|
+
cron:
|
|
115
|
+
- schedule: "0 9 * * *"
|
|
116
|
+
task: "Check and summarize overnight emails"
|
|
117
|
+
```
|
|
118
|
+
|
|
119
|
+
---
|
|
120
|
+
|
|
121
|
+
## 🖥️ CLI Commands
|
|
122
|
+
|
|
123
|
+
| Command | Description |
|
|
124
|
+
|---------|-------------|
|
|
125
|
+
| `opc init [name]` | Create a new agent project |
|
|
126
|
+
| `opc run` | Start agent with all configured channels |
|
|
127
|
+
| `opc chat` | Interactive TUI chat in terminal |
|
|
128
|
+
| `opc studio` | Launch Studio GUI (port 4000) |
|
|
129
|
+
| `opc doctor` | Run 13 health checks |
|
|
130
|
+
| `opc setup` | Configure model provider & API keys |
|
|
131
|
+
| `opc eval` | Run evaluation test suite |
|
|
132
|
+
| `opc traces` | View OpenTelemetry traces |
|
|
133
|
+
| `opc publish` | Publish agent to npm |
|
|
134
|
+
| `opc skill list` | List available skills |
|
|
135
|
+
| `opc skill add <name>` | Add a skill to agent |
|
|
136
|
+
| `opc cron list` | List scheduled tasks |
|
|
137
|
+
|
|
138
|
+
---
|
|
139
|
+
|
|
140
|
+
## 📱 Channels
|
|
141
|
+
|
|
142
|
+
OPC Agent connects to **15+ messaging platforms** out of the box:
|
|
143
|
+
|
|
144
|
+
| Channel | Status | Channel | Status |
|
|
145
|
+
|---------|--------|---------|--------|
|
|
146
|
+
| Telegram | ✅ | Discord | ✅ |
|
|
147
|
+
| Slack | ✅ | WeChat | ✅ |
|
|
148
|
+
| Email (IMAP/SMTP) | ✅ | WhatsApp | ✅ |
|
|
149
|
+
| LINE | ✅ | Teams | ✅ |
|
|
150
|
+
| Feishu/Lark | ✅ | DingTalk | ✅ |
|
|
151
|
+
| Web UI | ✅ | WebSocket | ✅ |
|
|
152
|
+
| Webhook | ✅ | REST API | ✅ |
|
|
153
|
+
| Voice | ✅ | SMS (Twilio) | ✅ |
|
|
154
|
+
|
|
155
|
+
---
|
|
156
|
+
|
|
157
|
+
## 🔧 Tools (53)
|
|
158
|
+
|
|
159
|
+
### Core (8)
|
|
160
|
+
`file-read` · `file-write` · `file-list` · `shell-exec` · `web-fetch` · `web-search` · `browser` · `vision`
|
|
161
|
+
|
|
162
|
+
### Developer (12)
|
|
163
|
+
`git-status` · `git-diff` · `git-commit` · `github-issue` · `github-pr` · `github-search` · `npm-search` · `code-analyze` · `test-run` · `lint` · `deploy` · `docker`
|
|
164
|
+
|
|
165
|
+
### Productivity (8)
|
|
166
|
+
`calendar` · `email-send` · `email-read` · `reminder` · `note` · `todo` · `timer` · `translate`
|
|
167
|
+
|
|
168
|
+
### Integration (13)
|
|
169
|
+
`jira-issue` · `jira-search` · `slack-send` · `slack-read` · `notion-page` · `notion-search` · `linear-issue` · `confluence` · `trello` · `asana` · `zendesk` · `hubspot` · `salesforce`
|
|
170
|
+
|
|
171
|
+
### Knowledge (7)
|
|
172
|
+
`memory-store` · `memory-recall` · `knowledge-learn` · `knowledge-evolve` · `rag-query` · `embedding` · `summarize`
|
|
173
|
+
|
|
174
|
+
### Media (5)
|
|
175
|
+
`image-generate` · `image-describe` · `audio-transcribe` · `tts-speak` · `screenshot`
|
|
176
|
+
|
|
177
|
+
---
|
|
178
|
+
|
|
179
|
+
## 📊 Comparison
|
|
180
|
+
|
|
181
|
+
| Feature | OPC Agent | Hermes Agent | OpenClaw |
|
|
182
|
+
|---------|-----------|-------------|----------|
|
|
183
|
+
| Built-in tools | 53 | ~10 | 30+ |
|
|
184
|
+
| GUI (Studio) | ✅ | ❌ | ✅ |
|
|
185
|
+
| TUI Chat | ✅ | ❌ | ✅ |
|
|
186
|
+
| Channels | 15+ | 3 | 15+ |
|
|
187
|
+
| Built-in skills | 40 | ❌ | 40 |
|
|
188
|
+
| Knowledge evolution | ✅ | ❌ | ✅ |
|
|
189
|
+
| Voice (STT/TTS) | ✅ | ❌ | ✅ |
|
|
190
|
+
| MCP Protocol | ✅ | ✅ | ✅ |
|
|
191
|
+
| A2A Protocol | ✅ | ❌ | ✅ |
|
|
192
|
+
| Cron scheduler | ✅ | ❌ | ✅ |
|
|
193
|
+
| One-line install | ✅ | ❌ | ✅ |
|
|
194
|
+
| Health checks | 13 | ❌ | ✅ |
|
|
195
|
+
| Local-first (Ollama) | ✅ | ❌ | ✅ |
|
|
196
|
+
| Open source | Apache-2.0 | Proprietary | Apache-2.0 |
|
|
197
|
+
|
|
198
|
+
> **OPC Agent** is the open-source core runtime. **OpenClaw** is the full platform built on top of it.
|
|
199
|
+
|
|
200
|
+
---
|
|
201
|
+
|
|
202
|
+
## 🧠 Knowledge Evolve Engine
|
|
203
|
+
|
|
204
|
+
OPC Agent includes a built-in knowledge evolution pipeline that runs **entirely local** with Ollama:
|
|
205
|
+
|
|
206
|
+
```
|
|
207
|
+
Conversations → Learn → Cluster → Deduplicate → Distill → Evolved Knowledge
|
|
208
|
+
```
|
|
209
|
+
|
|
210
|
+
- **Zero API cost** — uses local Ollama models for distillation
|
|
211
|
+
- **Automatic** — learns from every conversation, evolves on schedule
|
|
212
|
+
- **Tiered memory** — short-term (conversation) → long-term (distilled) → evolved (refined)
|
|
213
|
+
- **Full-text search** — SQLite FTS5 for instant recall across all memory
|
|
214
|
+
|
|
215
|
+
```bash
|
|
216
|
+
opc knowledge evolve # Trigger manual evolution
|
|
217
|
+
opc knowledge stats # View knowledge base stats
|
|
218
|
+
opc knowledge search "query" # Search across all knowledge
|
|
219
|
+
```
|
|
220
|
+
|
|
221
|
+
---
|
|
222
|
+
|
|
223
|
+
## 🎨 Studio
|
|
224
|
+
|
|
225
|
+
Launch the visual management dashboard:
|
|
226
|
+
|
|
227
|
+
```bash
|
|
228
|
+
opc studio
|
|
229
|
+
```
|
|
230
|
+
|
|
231
|
+
Studio provides:
|
|
232
|
+
- **Agent overview** — status, model, channels, tools at a glance
|
|
233
|
+
- **Live chat** — test your agent in the browser
|
|
234
|
+
- **Configuration editor** — edit `oad.yaml` visually
|
|
235
|
+
- **Logs & traces** — real-time log streaming and OpenTelemetry traces
|
|
236
|
+
- **Skill browser** — discover and install skills
|
|
237
|
+
- **Cron manager** — create and monitor scheduled tasks
|
|
238
|
+
|
|
239
|
+
---
|
|
240
|
+
|
|
241
|
+
## 🏥 Doctor
|
|
242
|
+
|
|
243
|
+
Run comprehensive health checks:
|
|
244
|
+
|
|
245
|
+
```bash
|
|
246
|
+
opc doctor
|
|
247
|
+
```
|
|
248
|
+
|
|
249
|
+
```
|
|
250
|
+
✅ Model provider connected (ollama/llama3.1)
|
|
251
|
+
✅ 53/53 tools available
|
|
252
|
+
✅ Memory store healthy (SQLite, 1,247 entries)
|
|
253
|
+
✅ Telegram channel connected
|
|
254
|
+
✅ Cron scheduler running (3 jobs)
|
|
255
|
+
⚠️ No TTS provider configured
|
|
256
|
+
✅ Disk space OK (12.3 GB free)
|
|
257
|
+
...
|
|
258
|
+
```
|
|
259
|
+
|
|
260
|
+
13 checks covering: model connectivity, tool wiring, channel status, memory health, disk space, Node.js version, package updates, and more.
|
|
261
|
+
|
|
262
|
+
---
|
|
263
|
+
|
|
264
|
+
## 🤝 Contributing
|
|
265
|
+
|
|
266
|
+
We welcome contributions! See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.
|
|
267
|
+
|
|
268
|
+
```bash
|
|
269
|
+
git clone https://github.com/Deepleaper/opc-agent.git
|
|
270
|
+
cd opc-agent
|
|
271
|
+
npm install
|
|
272
|
+
npm run build
|
|
273
|
+
npm test
|
|
274
|
+
```
|
|
275
|
+
|
|
276
|
+
---
|
|
277
|
+
|
|
278
|
+
## 📄 License
|
|
279
|
+
|
|
280
|
+
[Apache-2.0](LICENSE) © [Deepleaper](https://github.com/Deepleaper)
|
|
281
|
+
|
|
282
|
+
---
|
|
283
|
+
|
|
284
|
+
<div align="center">
|
|
285
|
+
|
|
286
|
+
**If OPC Agent helps you build better agents, give us a ⭐**
|
|
287
|
+
|
|
288
|
+
[GitHub](https://github.com/Deepleaper/opc-agent) · [npm](https://www.npmjs.com/package/opc-agent) · [Docs](https://opc-agent.dev)
|
|
289
|
+
|
|
290
|
+
</div>
|