# anwale-v3-protector

`anwale-v3-protector` is a domain-specialized assistant for crypto market analysis with a protective, skeptical reasoning style.

Packaged for Ollama as:

- `bazobehram/anwale-v3-protector:latest`

## Quickstart

```bash
ollama pull bazobehram/anwale-v3-protector:latest
ollama run bazobehram/anwale-v3-protector:latest

Model Profile

• Architecture: llama
• Parameter count: 3.2B
• Quantization: BF16
• Native context length: 131072
• Default runtime context (num_ctx): 8192

Persona and Behavior

This model is configured with the AnWale Protector system role:

• Protective, analytical, authoritative tone
  
• Dialectical skepticism (challenge assumptions, verify claims)
  
• Crypto-centric intelligence (DeFi, macro, cross-chain dynamics)
  
• Strong emphasis on risk framing over hype
  

Default Parameters

Configured defaults:

• temperature: 0.7
• top_p: 0.9
• num_ctx: 8192
• Stop tokens:
  • <|start_header_id|>
  • <|end_header_id|>
  • <|eot_id|>
  • User:
  • Assistant:

Suggested Use Cases

• Token/project thesis stress-testing
  
• On-chain and narrative risk analysis
  
• Scenario analysis for macro + crypto interactions
  
• Cross-chain strategy comparison with explicit tradeoffs
  

Out-of-Scope / Guardrails

• Not for direct buy/sell instructions
  
• Not a substitute for legal, tax, or regulated financial advice
  
• Outputs can be wrong or stale; verify with primary sources and live data
  

Prompting Tips

Example prompts:

1. Stress-test this ETH thesis and list top 5 failure modes.
   
2. Compare SOL vs ETH for the next 6 months under three macro scenarios.
   
3. What on-chain metrics would falsify this bullish narrative?
   
4. Give me a balanced bull/base/bear view on this protocol launch.
   
5. Red-team this strategy and highlight hidden correlation risks.
   

Build / Lineage Notes

• Finetuning workflow used LoRA (peft) and was merged into a full model for Ollama export.
  
• Base lineage includes unsloth/llama-3.2-3b-instruct-unsloth-bnb-4bit from the training artifacts.
  

License

Follow the license and usage terms of the upstream base model and any finetuning data used in this release. “`