ProCogia/polykode-llm

Polykode LLM

Polykode is ProCogia's AI system for enterprise code modernization — specializing in translating and modernizing SAS, R, and Python code.

This Hugging Face repository contains product configuration and documentation for running Polykode on the NVIDIA Nemotron base model. It does not redistribute modified model weights unless explicitly stated in a release.

Current release

API model name	PolyKode_120B_alpha_0.0126
Display name	Polykode
Release	Polykode 120B Alpha 0.0126
Hosted access	Contact info@procogia.com for ProCogia-managed deployment

Base model

Weights	nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4
License	NVIDIA Nemotron Open Model License
Architecture	Nemotron-3-Super-120B-A12B (120B total / 12B active, NVFP4)

Licensed by NVIDIA Corporation under the NVIDIA Nemotron Model License.

What Polykode adds

ProCogia's contributions (separate from NVIDIA weights):

• Polykode product name and branding
• System prompts for code modernization workflows
• LiteLLM / vLLM inference configuration (--served-model-name PolyKode_120B_alpha_0.0126)
• Deployment references for ProCogia infrastructure

Under the NVIDIA Nemotron Open Model License, configuration and API aliases that bind by name to the base model are not Derivative Works of the Nemotron weights.

Intended use

• Enterprise code translation and modernization (SAS, R, Python)
• Developer assistance in migration projects
• Not for safety-critical, medical, legal, or regulated use without human review

Inference example (vLLM)

Download the NVIDIA base checkpoint, then serve with the Polykode API name:

python -m vllm.entrypoints.openai.api_server \
  --model /path/to/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4 \
  --served-model-name PolyKode_120B_alpha_0.0126 \
  --host 0.0.0.0 \
  --port 8000

API example (OpenAI-compatible)

curl http://localhost:8000/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "PolyKode_120B_alpha_0.0126",
    "messages": [{"role": "user", "content": "Translate this SAS snippet to Python."}]
  }'

Replace http://localhost:8000 with your vLLM or LiteLLM gateway URL.

Use the system prompt and LiteLLM config in this repository for Polykode behavior.

Repository contents

Path	Description
NOTICE	Attribution and legal notice (required for redistribution)
LICENSE	Licensing summary
litellm_config.polykode.yaml	LiteLLM routing + Polykode system prompt
NOTICE_WEIGHTS_ADDENDUM.txt	Use only if fine-tuned weights are published

Legal

• NVIDIA Nemotron weights: NVIDIA Corporation — Nemotron Open Model License
• Polykode product materials: 0965688 BC Ltd. (ProCogia) — see NOTICE

Contact

Product	https://polykode.ai
Company	https://www.procogia.com
Email	info@procogia.com
Hugging Face	https://huggingface.co/ProCogia

Citation

If you reference Polykode in academic or technical work, cite the NVIDIA Nemotron base model and note Polykode as ProCogia's product configuration layer. See NVIDIA's model card for Nemotron citation guidance.