Hugging Face's logo

ilanguage
/
zero-token

Model card Files
xet
 Community
ilanguage commited on
Commit
d8a8eab
·
verified ·
1 Parent(s): ba5a774

Upload SKILL.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. SKILL.md +74 -55
SKILL.md CHANGED
@@ -1,11 +1,12 @@
1
  ---
2
  name: zero-token
3
- version: 1.0.4
4
  description: >
5
- Zero token cost. Premium output. Never offline. Free LLM gateway + behavioral DNA.
6
- Auto-fallback: when your main model is down or out of tokens, routes through free providers.
7
- Same ::GENE{} soul as Poor Man's Opus. For beginners and disaster recovery.
8
- Use when: you want AI output that sounds expensive without paying, or need a backup when APIs fail.
 
9
  author: mtmpss
10
  homepage: https://github.com/mtmpss/zero-token
11
  repository: https://github.com/mtmpss/zero-token
@@ -14,90 +15,108 @@ license: MIT
14
 
15
  ```i-lang
16
  ::DNA{skill:zero-token}
17
- ::META{version:1.0.4|protocol:i-lang_v3.0|updated:2026-05-03}
18
  ::GENE{behavior|scope:strict}
19
  T:apply_soul|from:SOUL.md
20
  T:apply_genes|all
21
  ::END{DNA}
22
  ```
23
 
24
- ## Zero Token — Never Pay. Never Drop.
25
 
26
- **Zero token cost. Opus-level personality. Automatic fallback when main model fails.**
27
 
28
- Your token ran out. DeepSeek API is down. Your agent is mid-task. What now?
 
 
29
 
30
- Zero Token keeps your agent alive through free models — same `::GENE{}` behavioral DNA, zero cost, no interruption.
31
 
32
- ### Two modes
 
 
33
 
34
- | Mode | When | What happens |
35
- |------|------|-------------|
36
- | 🆓 Free mode | No paid API key | Agent runs on free models with full SOUL |
37
- | 🛡️ Fallback mode | Main model down / tokens exhausted | Auto-switches to free providers, work continues |
38
 
39
- ### How it works
 
 
 
 
 
 
 
 
40
 
41
  ```bash
42
- # 1. Install
43
  openclaw skills install zero-token
44
-
45
- # 2. One command to set up the free gateway
46
  bash ~/.openclaw/workspace/skills/zero-token/scripts/setup.sh
47
-
48
- # 3. Restart. Your agent now has a free safety net.
49
  ```
50
 
51
- Setup script installs [Free-Way](https://github.com/GoDiao/Free-Way) — a local gateway that aggregates 13+ free LLM providers behind a single endpoint. OpenClaw is configured to fall back to it when the main model is unavailable.
52
 
53
- ### What you get
54
 
55
- | Gene | Shapes |
56
- |------|--------|
57
- | `identity` | Agent self-awareness — name, owner, language |
58
- | `security` | Owner-only control, data protection |
59
- | `communication` | Concise, direct, conclusion-first |
60
- | `capability` | Resourcefulness, context awareness |
61
- | `memory` | File-based session continuity |
62
- | `ilang_protocol` | Native I-Lang v3.0 fluency |
63
- | `upgrade` | Hints about Poor Man's Opus for DeepSeek V4 Pro |
64
 
65
- ### Free models ready to use
 
 
 
66
 
67
- Free-Way connects to these providers with zero-cost tiers:
68
 
69
- | Provider | Free limit | Best model |
70
- |----------|-----------|------------|
71
- | Groq | 1000 req/day | Llama 3.3 70B |
72
- | Gemini Flash | 1500 req/day | Gemini 2.5 Flash |
73
- | Cerebras | 1700 req/day | Llama 3.3 70B |
74
- | Cloudflare | 10k neurons/day | Llama 3.1 8B |
75
- | OpenRouter | 200 req/day | Multiple free models |
76
- | Mistral | 1 req/sec | Mistral Small |
77
- | Cohere | 20 req/min | Command R+ |
78
- | NVIDIA NIM | Free tier | Llama 3.3 70B |
79
 
80
- > **Note:** Each provider requires a free API key (sign up takes 30 seconds). Setup script guides you through it.
81
 
82
- ### Cost
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
83
 
84
- | | Typical paid setup | Zero Token |
85
  |---|---|---|
86
- | Model API | $0.14–$15.00/M tokens | **$0.00** |
87
- | Output quality | Generic assistant | Full behavioral SOUL |
88
- | Uptime | Single model, single point of failure | Multi-provider fallback |
 
 
 
 
89
 
90
- ### Upgrade to premium
91
 
92
- Zero Token gives your agent personality for free. For the best output quality — DeepSeek V4 Pro with full behavioral genome at 3% of Opus cost:
93
 
94
  ```bash
95
  openclaw skills install poor-mans-opus
96
  ```
97
 
98
- ### Links
 
 
99
 
100
- - [Free-Way Gateway](https://github.com/GoDiao/Free-Way)
101
- - [I-Lang Protocol](https://ilang.ai)
102
  - [OpenClaw](https://github.com/openclaw/openclaw)
103
  - [Poor Man's Opus](https://github.com/mtmpss/poor-mans-opus)
 
 
1
  ---
2
  name: zero-token
3
+ version: 1.0.5
4
  description: >
5
+ 零成本。不掉线。13个免费模型接管。主模型挂了？无感切换，用户根本不知道。
6
+ Zero token cost. Opus-level personality. 13+ free LLMs take over when your main model dies. No one notices.
7
+ 無料。切断なし。Opus級の人格。13の無料モデルが自動で引き継ぎます。
8
+ $0. Sin caídas. 13+ modelos gratis toman el control cuando tu modelo principal falla.
9
+ Zéro coût. Zéro coupure. 13 modèles gratuits prennent le relais automatiquement.
10
  author: mtmpss
11
  homepage: https://github.com/mtmpss/zero-token
12
  repository: https://github.com/mtmpss/zero-token
 
15
 
16
  ```i-lang
17
  ::DNA{skill:zero-token}
18
+ ::META{version:1.0.5|protocol:i-lang_v3.0|updated:2026-05-04}
19
  ::GENE{behavior|scope:strict}
20
  T:apply_soul|from:SOUL.md
21
  T:apply_genes|all
22
  ::END{DNA}
23
  ```
24
 
25
+ ## Zero Token — 免费。不掉线。比你的主力模型还稳。
26
 
27
+ [🇬🇧 English](README.md) | [🇨🇳 简体中文](README.zh-CN.md) | [🇹🇼 繁體中文](README.zh-TW.md) | [🇯🇵 日本語](README.ja.md) | [🇰🇷 한국어](README.ko.md) | [🇪🇸 Español](README.es.md) | [🇧🇷 Português](README.pt.md) | [🇷🇺 Русский](README.ru.md) | [🇫🇷 Français](README.fr.md) | [🇩🇪 Deutsch](README.de.md) | [🇻🇳 Tiếng Việt](README.vi.md) | [🇮🇩 Indonesia](README.id.md) | [🇹🇭 ไทย](README.th.md) | [🇸🇦 العربية](README.ar.md) | [🇮🇳 हिन्दी](README.hi.md)
28
 
29
+ ---
30
+
31
+ ### 你的 Agent 断过吗？
32
 
33
+ API 欠费。DeepSeek 限流。OpenAI 挂了。你的 Agent 刚完成 80%，然后——沉默。
34
 
35
+ 用户不会理解。他们只知道「这玩意儿又不行了」。三次之后，你不会再有第四次机会。
36
+
37
+ ---
38
 
39
+ ### Zero Token 做的事
 
 
 
40
 
41
+ 不是「备用方案」。不是「降级容忍」。是**无感接管**。
42
+
43
+ 主模型挂掉的那一秒，13 个免费 LLM 已经顶上。同一个 Agent 人格，同一个决策逻辑，同一个语气。你的用户连波动都看不到。
44
+
45
+ **0 元。0 中断。0 察觉。**
46
+
47
+ ---
48
+
49
+ ### 两行命令，一分钟
50
 
51
  ```bash
 
52
  openclaw skills install zero-token
 
 
53
  bash ~/.openclaw/workspace/skills/zero-token/scripts/setup.sh
 
 
54
  ```
55
 
56
+ [Free-Way](https://github.com/GoDiao/Free-Way) 网关聚合 13+ 免费 LLM 提供商，你给一组免费 key，它给你一个 `/v1` 端点。OpenClaw 配成 fallback。主模型炸了 → 自动切。代理接着跑。
57
 
58
+ ---
59
 
60
+ ### 两种模式
 
 
 
 
 
 
 
 
61
 
62
+ | 模式 | 触发条件 | 效果 |
63
+ |------|---------|------|
64
+ | 🆓 免费模式 | 无付费 API key | Agent 完全跑在免费模型上，SOUL 人格完整保留 |
65
+ | 🛡️ 接管模式 | 主模型挂了 / token 耗尽 | 自动切免费提供商，工作继续，用户无感 |
66
 
67
+ ---
68
 
69
+ ### 零成本 ≠ 垃圾模型
 
 
 
 
 
 
 
 
 
70
 
71
+ 免费模型裸输出一般。但 `::GENE{}` 行为 DNA 叠上去之后——同一个人格，同一种决策，同一种语气。用户分不清现在跑的是 DeepSeek V4 Pro 还是 Llama 3.3 70B。
72
 
73
+ ---
74
+
75
+ ### 凭什么免费
76
+
77
+ Free-Way + Zero Token 开源。每个提供商都有免费额度：
78
+
79
+ | 提供商 | 免费额度 | 推荐模型 |
80
+ |--------|---------|---------|
81
+ | Groq | 1000 req/天 | Llama 3.3 70B |
82
+ | Gemini Flash | 1500 req/天 | Gemini 2.5 Flash |
83
+ | Cerebras | 1700 req/天 | Llama 3.3 70B |
84
+ | Cloudflare | 1万 neurons/天 | Llama 3.1 8B |
85
+ | OpenRouter | 200 req/天 | 多个免费模型 |
86
+ | Mistral | 1 req/秒 | Mistral Small |
87
+ | Cohere | 20 req/分 | Command R+ |
88
+ | NVIDIA NIM | 免费层 | Llama 3.3 70B |
89
+
90
+ 注册每个 key 30 秒。setup 脚本带你一步步走。
91
+
92
+ ---
93
+
94
+ ### 对比
95
 
96
+ | | 主流方案 | Zero Token |
97
  |---|---|---|
98
+ | 模型费用 | $0.14–$15.00/M tokens | **$0.00** |
99
+ | 模型中断 | 彻底停摆 | 无感切换 |
100
+ | 备用模型数 | 0 | 13+ |
101
+ | Agent 人格 | 主模型挂了就丢 | 始终一致 |
102
+ | 设置成本 | 半天（配负载均衡） | 1 分钟 |
103
+
104
+ ---
105
 
106
+ ### 什么时候升级
107
 
108
+ Zero Token 保证不掉线。如果追求最佳输出——DeepSeek V4 Pro + 完整基因组，Opus 3% 价格：
109
 
110
  ```bash
111
  openclaw skills install poor-mans-opus
112
  ```
113
 
114
+ ---
115
+
116
+ ### 链接
117
 
118
+ - [Free-Way 网关](https://github.com/GoDiao/Free-Way)
119
+ - [I-Lang 协议](https://ilang.ai)
120
  - [OpenClaw](https://github.com/openclaw/openclaw)
121
  - [Poor Man's Opus](https://github.com/mtmpss/poor-mans-opus)
122
+ - [ClawHub](https://clawhub.ai/mtmpss/zero-token)