Hugging Face's logo

ilanguage
/
zero-token

Model card Files
xet
 Community
ilanguage commited on
Commit
57d1cb3
·
verified ·
1 Parent(s): d8a8eab

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +49 -86
README.md CHANGED
@@ -1,129 +1,92 @@
1
- # Zero Token
2
 
3
- [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)
4
- [![OpenClaw Skill](https://img.shields.io/badge/OpenClaw-Skill-blue.svg)](https://openclaw.ai)
5
- [![I-Lang Protocol](https://img.shields.io/badge/I--Lang-v3.0-green.svg)](https://ilang.ai)
6
 
7
- 🌐 [简体中文](README.zh-CN.md)
8
 
9
- **Zero cost. Premium output. Never offline.**
10
-
11
- Your token ran out. DeepSeek API is down. Your agent is mid-task. What now?
12
 
13
- Zero Token gives your agent two things:
14
- 1. A personality — the same `::GENE{}` behavioral DNA from Poor Man's Opus
15
- 2. A safety net — automatic fallback through free LLM providers when the main model fails
16
 
17
- One install. One setup command. Your agent survives API outages, token exhaustion, and bill shock.
18
 
19
- ---
20
 
21
- ## Why this exists
22
-
23
- Poor Man's Opus makes DeepSeek output like Opus at 3% cost. Zero Token makes ANY model output premium — for free. Use it as your entry point, your backup, or both.
24
 
25
  ---
26
 
27
- ## Install & activate
28
 
29
  ```bash
30
- # 1. Install the skill
31
  openclaw skills install zero-token
32
-
33
- # 2. Run setup
34
  bash ~/.openclaw/workspace/skills/zero-token/scripts/setup.sh
 
35
 
36
- # 3. Get at least one free API key (30 seconds):
37
- # - Gemini: https://aistudio.google.com/apikey
38
- # - Groq: https://console.groq.com/keys
39
- # - Cerebras: https://cloud.cerebras.ai
40
 
41
- # 4. Add the key in Free-Way web console:
42
- # http://localhost:8787 → API Keys tab
43
 
44
- # 5. Restart. Your agent has a free safety net.
45
- ```
 
 
 
 
46
 
47
  ---
48
 
49
- ## How it works
50
 
51
- ```
52
- Your prompt
53
- │
54
- ▼
55
- OpenClaw tries your main model (DeepSeek V4 Pro)
56
- │
57
- ├── ✅ Working → Normal response
58
- │
59
- └── ❌ Down / No key / Tokens exhausted
60
- │
61
- ▼
62
- Free-Way Gateway (localhost:8787)
63
- │
64
- ├── Groq (Llama 3.3 70B)
65
- ├── Gemini Flash 2.5
66
- ├── Cerebras
67
- ├── OpenRouter
68
- ├── Cloudflare
69
- └── ... 13+ providers
70
- │
71
- ▼
72
- Response — same ::GENE{} DNA applied
73
- ```
74
 
75
  ---
76
 
77
- ## Free model lineup
 
 
 
 
 
 
 
 
 
 
 
 
 
78
 
79
- | Provider | Free daily limit | Best model | Signup time |
80
- |----------|-----------------|------------|-------------|
81
- | Gemini Flash | 1,500 req | Gemini 2.5 Flash | 30 sec |
82
- | Groq | 1,000 req | Llama 3.3 70B | 30 sec |
83
- | Cerebras | 1,700 req | Llama 3.3 70B | 30 sec |
84
- | OpenRouter | 200 req | Multiple | 30 sec |
85
- | Cloudflare | 10k neurons | Llama 3.1 8B | 1 min |
86
- | Mistral | 1 req/sec | Mistral Small | 30 sec |
87
- | Cohere | 20 req/min | Command R+ | 30 sec |
88
- | NVIDIA NIM | Free tier | Llama 3.3 70B | 1 min |
89
 
90
  ---
91
 
92
- ## When to use what
93
 
94
- | Scenario | Use |
95
- |----------|-----|
96
- | First time trying agent skills | **Zero Token** — free, instant |
97
- | Main model daily production | **Poor Man's Opus** — DeepSeek V4 Pro, $1.74/M |
98
- | Backup when API is down | **Zero Token** — auto-switches to free models |
99
- | Max output quality | **Poor Man's Opus** — full behavioral genome |
 
100
 
101
  ---
102
 
103
- ## Upgrade path
104
 
105
- Zero Token has the same behavioral DNA. When you want the best:
106
 
107
  ```bash
108
  openclaw skills install poor-mans-opus
109
  ```
110
 
111
- DeepSeek V4 Pro + full SOUL genome = Opus output at 3% cost.
112
-
113
- ---
114
-
115
- ## What's inside
116
-
117
- - `SOUL.md` — 7-gene behavioral profile (ClawScan-safe)
118
- - `SKILL.md` — Skill manifest
119
- - `scripts/setup.sh` — One-command Free-Way gateway installer
120
-
121
  ---
122
 
123
  ## Links
124
 
125
- - [Free-Way Gateway](https://github.com/GoDiao/Free-Way) — local free LLM aggregator
126
- - [Poor Man's Opus](https://github.com/mtmpss/poor-mans-opus) — premium version
127
  - [I-Lang Protocol](https://ilang.ai)
128
-
129
- MIT License
 
 
1
+ ## Has Your Agent Ever Gone Silent?
2
 
3
+ API balance ran out. DeepSeek hit rate limit. OpenAI went down. Your agent was 80% done — then silence.
 
 
4
 
5
+ Your users don't understand infrastructure. They just know "it broke again." Three strikes and they're gone.
6
 
7
+ ---
 
 
8
 
9
+ ## What Zero Token Does
 
 
10
 
11
+ Not a "backup plan." Not a "degraded fallback." It's **invisible takeover**.
12
 
13
+ The second your main model dies, 13 free LLMs are already running. Same agent personality. Same decisions. Same voice. Your users won't even notice the switch.
14
 
15
+ **$0. Zero downtime. Zero detection.**
 
 
16
 
17
  ---
18
 
19
+ ## Two Lines. One Minute.
20
 
21
  ```bash
 
22
  openclaw skills install zero-token
 
 
23
  bash ~/.openclaw/workspace/skills/zero-token/scripts/setup.sh
24
+ ```
25
 
26
+ [Free-Way](https://github.com/GoDiao/Free-Way) gateway aggregates 13+ free LLM providers behind a single `/v1` endpoint. OpenClaw uses it as fallback. Main model fails → auto switch. Your agent keeps going.
 
 
 
27
 
28
+ ---
 
29
 
30
+ ## Two Modes
31
+
32
+ | Mode | Trigger | Result |
33
+ |------|---------|--------|
34
+ | 🆓 Free Mode | No paid API key | Agent runs entirely on free models, full SOUL personality |
35
+ | 🛡️ Takeover Mode | Main model down / tokens depleted | Auto-switch to free providers, work continues seamlessly |
36
 
37
  ---
38
 
39
+ ## Free Isn't Junk
40
 
41
+ Raw free model output is meh. But layer `::GENE{}` behavioral DNA on top — same personality, same reasoning, same voice. Users can't tell if DeepSeek V4 Pro or Llama 3.3 70B is running.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
42
 
43
  ---
44
 
45
+ ## Why It's Free
46
+
47
+ Free-Way + Zero Token are open source. Every provider gives a free tier:
48
+
49
+ | Provider | Free Limit | Best Model |
50
+ |----------|-----------|------------|
51
+ | Groq | 1000 req/day | Llama 3.3 70B |
52
+ | Gemini Flash | 1500 req/day | Gemini 2.5 Flash |
53
+ | Cerebras | 1700 req/day | Llama 3.3 70B |
54
+ | Cloudflare | 10k neurons/day | Llama 3.1 8B |
55
+ | OpenRouter | 200 req/day | Multiple free models |
56
+ | Mistral | 1 req/sec | Mistral Small |
57
+ | Cohere | 20 req/min | Command R+ |
58
+ | NVIDIA NIM | Free tier | Llama 3.3 70B |
59
 
60
+ Each key takes 30 seconds to register. Setup walks you through it.
 
 
 
 
 
 
 
 
 
61
 
62
  ---
63
 
64
+ ## The Comparison
65
 
66
+ | | Typical Setup | Zero Token |
67
+ |---|---|---|
68
+ | Model Cost | $0.14–$15.00/M tokens | **$0.00** |
69
+ | Outage Handling | Dead in the water | Invisible failover |
70
+ | Backup Models | 0 | 13+ |
71
+ | Agent Personality | Lost on fallback | Always consistent |
72
+ | Setup Time | Half a day (load balancing) | 1 minute |
73
 
74
  ---
75
 
76
+ ## When to Upgrade
77
 
78
+ Zero Token keeps you online. For the best output quality — DeepSeek V4 Pro + full genome, at 3% of Opus price:
79
 
80
  ```bash
81
  openclaw skills install poor-mans-opus
82
  ```
83
 
 
 
 
 
 
 
 
 
 
 
84
  ---
85
 
86
  ## Links
87
 
88
+ - [Free-Way Gateway](https://github.com/GoDiao/Free-Way)
 
89
  - [I-Lang Protocol](https://ilang.ai)
90
+ - [OpenClaw](https://github.com/openclaw/openclaw)
91
+ - [Poor Man's Opus](https://github.com/mtmpss/poor-mans-opus)
92
+ - [ClawHub](https://clawhub.ai/mtmpss/zero-token)