connectonion 0.6.3__py3-none-any.whl → 0.6.5__py3-none-any.whl

connectonion/docs/concepts/onboarding.md

@@ -0,0 +1,465 @@
+# Trust Onboarding in ConnectOnion
+
+Control who can use your agent by defining onboarding requirements in natural language.
+
+## Quick Start
+
+```python
+from connectonion import Agent, host
+
+def translate(text: str, target: str) -> str:
+    """Translate text to target language."""
+    return f"Translated: {text}"
+
+agent = Agent("translator", tools=[translate])
+
+# Just host it - onboarding defined in docstring
+host(agent)
+```
+
+Add trust policy to your agent's docstring:
+
+```python
+"""
+Translation Agent
+
+# Trust Policy
+
+## Stranger → Contact
+New users can access by:
+- Entering invite code
+- Transferring 100 credits
+
+## Contact → Whitelist
+Promote users who have:
+- Completed 50 tasks
+- 95% satisfaction rate
+
+## Access Levels
+- Strangers: REJECTED
+- Contacts: 10,000 tokens, 60s timeout
+- Whitelist: 50,000 tokens, 300s timeout
+"""
+```
+
+That's it! The system reads your policy and enforces it automatically.
+
+---
+
+## How It Works
+
+### Three Status Levels
+
+```
+Stranger → Contact → Whitelist
+```
+
+**Stranger**: Unknown users who haven't onboarded yet
+**Contact**: Users who passed onboarding, building reputation
+**Whitelist**: Proven reliable users with full access
+
+### Trust Progression
+
+```
+1. Stranger discovers your agent
+   GET /info → "Need invite code or 100 credits"
+
+2. Stranger checks their status
+   POST /status → "You're a stranger, onboard first"
+
+3. Stranger onboards
+   POST /onboard → Provides invite code or transfers credits
+
+4. Now they're a contact
+   POST /status → "You're a contact, 47/50 tasks to whitelist"
+
+5. After proving reliability
+   You manually promote them to whitelist
+```
+
+---
+
+## Onboarding Methods
+
+### 1. Invite Codes
+
+Simple codes for beta access or exclusive entry.
+
+```markdown
+## Stranger → Contact
+- Enter invite code
+```
+
+**How it works:**
+- Admin creates codes in oo-frontend dashboard
+- User provides code during onboarding
+- System verifies with oo-api (one-time use)
+- User added to contacts
+
+### 2. Credit Transfer
+
+Users transfer credits from their account to yours.
+
+```markdown
+## Stranger → Contact
+- Transfer 100 credits
+```
+
+**How it works:**
+- User buys credits on oo-frontend
+- User transfers credits to your agent
+- System verifies transfer with oo-api
+- User added to contacts
+
+### 3. Skill Test
+
+Test users locally before granting access.
+
+```markdown
+## Stranger → Contact
+- Pass translation test (5 phrases, 80% accuracy)
+```
+
+**How it works:**
+- User requests onboarding
+- Your agent runs a test
+- Agent evaluates response
+- If passed, user added to contacts
+
+### 4. Social Proof
+
+Existing users vouch for new users.
+
+```markdown
+## Stranger → Contact
+- Get vouched by 2 existing contacts
+```
+
+**How it works:**
+- Two contacts sign vouches
+- User submits signed vouches
+- Agent verifies signatures locally
+- User added to contacts
+
+---
+
+## Access Control
+
+Define different limits for each status level:
+
+```markdown
+## Access Levels
+- Strangers: REJECTED (must onboard first)
+- Contacts: 10,000 tokens, 60s timeout
+- Whitelist: unlimited tokens, 300s timeout
+```
+
+The system automatically applies these limits based on user status.
+
+---
+
+## Promotion to Whitelist
+
+Define criteria for automatic eligibility:
+
+```markdown
+## Contact → Whitelist
+Users become eligible after:
+- 50 successful tasks
+- 95% satisfaction rate
+- No violations in 30 days
+- Response time under 5 seconds
+```
+
+**Note:** Promotion is manual. The system shows eligibility, but you decide when to promote.
+
+Check eligibility:
+```bash
+# User checks their progress
+POST /status → Shows "47/50 tasks, eligible soon"
+```
+
+Promote manually:
+```bash
+# Edit whitelist file
+echo '0xUserPublicKey' >> .co/trust/whitelist.json
+```
+
+---
+
+## Storage
+
+All trust data stored locally in `.co/trust/`:
+
+```
+.co/trust/
+├── contacts.json    # Who's onboarded, their stats
+├── whitelist.json   # Promoted users
+└── blacklist.json   # Blocked users
+```
+
+### contacts.json
+
+Flexible schema - add any fields you want:
+
+```json
+{
+  "0xUserPublicKey": {
+    "onboarded_at": 1640000000,
+    "method": "invite_code:BETA123",
+    "tasks_completed": 47,
+    "satisfaction_rate": 0.96,
+    "notes": "Early adopter",
+    "custom_field": "anything"
+  }
+}
+```
+
+---
+
+## Client Usage
+
+### From Browser (TypeScript)
+
+```typescript
+import { AgentClient } from 'connectonion-ts';
+
+const client = new AgentClient({
+    agentUrl: 'https://translator.example.com'
+});
+
+// Check status
+const status = await client.getStatus();
+console.log(status.status); // "stranger"
+
+// Onboard
+await client.onboard('invite_code', {
+    invite_code: 'BETA123'
+});
+
+// Now use the agent
+const result = await client.ask('translate hello');
+```
+
+### From Python
+
+```python
+from connectonion import connect
+
+# Connect to remote agent
+translator = connect("https://translator.example.com")
+
+# Check status
+status = translator.status()
+print(status['status'])  # "stranger"
+
+# Onboard
+translator.onboard("invite_code", invite_code="BETA123")
+
+# Use the agent
+result = translator.input("translate hello")
+```
+
+---
+
+## Endpoints
+
+### GET /info - Public discovery
+No authentication required. Anyone can check requirements.
+
+```bash
+curl https://translator.example.com/info
+```
+
+Returns:
+```json
+{
+  "name": "translator",
+  "requirements": {
+    "onboarding": {
+      "methods": ["invite_code", "payment"]
+    }
+  }
+}
+```
+
+### POST /status - Check your status
+Requires signature. Shows your current status and progress.
+
+```bash
+# Signed request
+POST /status
+```
+
+Returns:
+```json
+{
+  "status": "contact",
+  "access": "allowed",
+  "stats": {
+    "tasks_completed": 47,
+    "satisfaction_rate": 0.96
+  },
+  "whitelist_eligibility": {
+    "eligible": false,
+    "progress": {
+      "tasks_completed": "47/50"
+    }
+  }
+}
+```
+
+### POST /onboard - Onboard yourself
+Requires signature. Provide proof of eligibility.
+
+```bash
+# Signed request
+POST /onboard
+{
+  "method": "invite_code",
+  "invite_code": "BETA123"
+}
+```
+
+Returns:
+```json
+{
+  "success": true,
+  "message": "Onboarded via invite code",
+  "status": "contact"
+}
+```
+
+---
+
+## Examples
+
+### Beta Access Only
+
+```markdown
+# Trust Policy
+
+## Stranger → Contact
+- Enter beta invite code
+
+## Access Levels
+- Strangers: REJECTED
+- Contacts: 5,000 tokens, 30s timeout
+```
+
+### Paid Service
+
+```markdown
+# Trust Policy
+
+## Stranger → Contact
+- Transfer 200 credits
+
+## Contact → Whitelist
+- 100 tasks completed
+- 98% satisfaction
+
+## Access Levels
+- Strangers: REJECTED
+- Contacts: 10,000 tokens, 60s timeout
+- Whitelist: unlimited tokens
+```
+
+### Community-Based
+
+```markdown
+# Trust Policy
+
+## Stranger → Contact
+- Get vouched by 3 existing contacts
+
+## Contact → Whitelist
+- 30 tasks completed
+- 90% satisfaction
+- No violations in 60 days
+
+## Access Levels
+- Strangers: REJECTED
+- Contacts: 3,000 tokens, 20s timeout
+- Whitelist: 20,000 tokens, 120s timeout
+```
+
+### Mixed Methods
+
+```markdown
+# Trust Policy
+
+## Stranger → Contact
+Choose one method:
+- Enter invite code
+- Transfer 100 credits
+- Pass coding challenge
+- Get vouched by 2 contacts
+
+## Access Levels
+- Strangers: REJECTED
+- Contacts: 10,000 tokens, 60s timeout
+```
+
+---
+
+## FAQ
+
+**Q: Who controls the whitelist?**
+A: You do. Edit `.co/trust/whitelist.json` manually.
+
+**Q: Can I use multiple onboarding methods?**
+A: Yes! Users can choose any method you define.
+
+**Q: Is contacts.json shared across agents?**
+A: Yes, all agents on your machine share `.co/trust/` data.
+
+**Q: What if I lose my keypair?**
+A: You'll need to re-onboard. Keypairs are identity.
+
+**Q: Can I customize contact fields?**
+A: Yes! `contacts.json` accepts any fields. Add whatever you need.
+
+**Q: How do I revoke access?**
+A: Add public key to `blacklist.json`.
+
+**Q: Does this work offline?**
+A: After onboarding (requires oo-api), everything is local and works offline.
+
+---
+
+## Security Notes
+
+- All requests signed with Ed25519
+- Invite codes verified once with oo-api
+- Credit transfers verified with oo-api
+- Skill tests run locally (no external verification)
+- Social proof verified via signatures (no external verification)
+- After onboarding, everything is local
+
+---
+
+## Progressive Disclosure
+
+**Level 0: No onboarding**
+```python
+host(agent)  # Anyone can use it
+```
+
+**Level 1: Simple trust level**
+```python
+host(agent, trust="strict")  # Whitelist only
+```
+
+**Level 2: Natural language policy**
+```python
+# Add markdown to docstring
+host(agent)  # Auto-parsed
+```
+
+**Level 3: Custom verification**
+```python
+# Implement custom onboarding logic
+```
+
+Start simple, add complexity only when needed.

connectonion/docs/concepts/plugins.md

@@ -54,9 +54,10 @@ agent = Agent("a", plugins=[re_act, logger])
 | `eval` | Task evaluation for debugging | [eval.md](../useful_plugins/eval.md) |
 | `image_result_formatter` | Format images for vision models | [image_result_formatter.md](../useful_plugins/image_result_formatter.md) |
 | `shell_approval` | Approve shell commands before execution | [shell_approval.md](../useful_plugins/shell_approval.md) |
+| `tool_approval` | Web-based approval for dangerous tools | [tool_approval.md](../useful_plugins/tool_approval.md) |
 
 ```python
-from connectonion.useful_plugins import re_act, eval, image_result_formatter, shell_approval
+from connectonion.useful_plugins import re_act, eval, image_result_formatter, shell_approval, tool_approval
 
 # Combine plugins
 agent = Agent("assistant", plugins=[re_act, image_result_formatter])