avai

Ava I

Ava I is an experimental Portuguese-first assistant model project by Lukintosh.

This repository is intended to publish Ava I adapters and release metadata. It does not include Meta Llama base weights. To use this adapter, you must have access to the compatible base model under the applicable Meta Llama license.

Model Identity

• Public model id: Lukintosh/ava-i
• OpenRouter slug target: lukintosh/ava-i
• Base model: meta-llama/Llama-3.2-1B-Instruct
• Release type: LoRA adapter / experimental API release

Status

This is an early experimental release. Quality depends heavily on the selected base checkpoint, adapter version, system prompt, and inference backend.

Intended Use

• Portuguese assistant experiments
• Lukintosh Ava I API compatibility tests
• LoRA fine-tuning and evaluation workflows

Limitations

• May hallucinate names, facts, sources, or recent information.
• Not suitable for high-stakes medical, legal, financial, or safety-critical use.
• Current public API may run on temporary backend infrastructure during testing.

API

OpenAI-compatible endpoint:

curl https://api.lukintosh.com/v1/chat/completions \
  -H "authorization: Bearer YOUR_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "model": "Lukintosh/ava-i",
    "messages": [{"role": "user", "content": "Quem e voce?"}],
    "max_tokens": 120
  }'

License

Adapter and project files are released by Lukintosh. The base model remains under its original license. Users are responsible for complying with the base model license and terms.