lotti-code-model / README.md
pspelsberg's picture
Update README.md
f5bf67f verified
|
Raw
History Blame Contribute Delete
2.03 kB
metadata
license: gemma
tags:
  - text-generation
  - safetensors
  - typescript
  - express-api
  - python
  - tailwind-css
language:
  - de
  - en
pipeline_tag: text-generation
base_model: unsloth/gemma-4-e2b

LottiCode — Pure Execution Code-Monkey

"Zero architectural bloat. Maximum 16-bit execution precision. Built to code down exact specifications."

LottiCode is a highly optimized, specialized model built exclusively for pure, raw code generation (writing down the lines). It operates natively on Linux with full hardware acceleration on AMD GPUs.

Crucial Operational Framework

  • NO Architecture, NO Code Reviews: This model does not design software architectures, evaluate system design, or review existing codebases. It is built to code, not to plan.
  • Requires an Architect Agent: To function efficiently, LottiCode depends entirely on a separate, high-level architecture agent that provides an explicit, foolproof step-by-step master plan.
  • Context Window Optimization: Because the model works most efficiently within a compact context window (recommended 8192), the orchestration layer must feed it clean, granular, bite-sized tasks. It turns precise, isolated technical specs into flawless, production-ready code—step by step.

Core Competencies (Pure Execution)

  1. TypeScript & JavaScript: Writing down rigid typings, exact interface implementations, and clean functional logic.
  2. Express API Services: Generating robust endpoints, explicit route validations, and strict error handlers based on strict specifications.
  3. Python Backends: Fast scripting, data processing logic, and automation code.
  4. Tailwind CSS Frontend: Turning precise UI specs into perfect HTML/JSX layout code with responsive Tailwind utility classes.

Recommended LM Studio Parameters

  • Context Length: 8192 (Keep it small and focused per sub-task to maximize precision!)
  • Max Tokens (Output): 2048
  • GPU Offload: Max (100% VRAM utilization via Vulkan/ROCm backend)