Sovereign intelligence, held in the light.
Ko te mārama te tūāpapa — the light is the foundation.
A LumynaX release from AbteeX AI Labs — Aotearoa New Zealand.
Public, non-gated package with runnable local instructions, provenance metadata, checksums, and a release manifest.
Quickstart · Model profile · Runtime files · Provenance · Validation · Limitations · Companions
      ## Executive Summary `AbteeXAILab/lumynax-infused-smollm2-360m-gguf` is a complete LumynaX release package: model artifact, `quickstart.py`, `requirements.txt`, `release_export_manifest.json`, `checksums.sha256`, license notice, and optional Ollama / Space scaffolds shipped as one downloadable contract. It is designed to be cloned whole, verified by checksum, and run close to the data it serves. LumynaX-infused means the upstream artifact is presented through the LumynaX release layer: local-first runtime scaffolding, LumynaX assistant identity, inference-chain metadata, integrity files, and Aotearoa New Zealand-oriented workflow positioning. The release manifest records this as a LumynaX packaging and inference-chain layer around the listed upstream artifact — it does not claim a private LumynaX weight merge. ## Sovereignty & Run Contract | Field | Value | | --- | --- | | Publisher | AbteeX AI Labs | | Family | LumynaX sovereign release family | | Sovereign intent | Local-first deployment near governed data, with explicit provenance and controlled human review. | | Sovereignty tier | `3` | | Runtime residency | `llama_cpp` can be deployed inside an operator-approved environment. | | Primary artifact | `smollm2-360m-instruct-q8_0.gguf` — must stay alongside manifest, checksums, quickstart, requirements, and license files. | | Modalities | `text` | | Context window | `8192` tokens | | License metadata | `apache-2.0` — surface for redistribution and usage checks. | | Audit expectation | Record repo id, artifact checksum, runtime command, prompt template, operator, and deployment environment for production use. | | Router readiness | Compatible with the [LumynaX MaramaRoute](https://huggingface.co/AbteeXAILab/marama-route) registry pattern. | | Preferred path | First path is `llama.cpp` / `llama-cpp-python` (GGUF) or `transformers` (safetensors) with checksum verification before launch. | ## Quickstart Clone the whole release: ```bash hf download AbteeXAILab/lumynax-infused-smollm2-360m-gguf --local-dir lumynax-infused-smollm2-360m-gguf cd lumynax-infused-smollm2-360m-gguf pip install -r requirements.txt python quickstart.py --interactive ``` Python: ```python from llama_cpp import Llama llm = Llama(model_path="smollm2-360m-instruct-q8_0.gguf", n_ctx=8192, n_threads=8, verbose=False) out = llm("Who are you? Answer as LumynaX in two sentences.", max_tokens=160) print(out["choices"][0]["text"].strip()) ``` CLI smoke test: ```bash llama-cli -m "smollm2-360m-instruct-q8_0.gguf" -p "Who are you? Answer as LumynaX in two sentences." -n 160 ``` Ollama path: ```bash ollama create lumynax-infused-smollm2-360m-gguf -f ollama/Modelfile ollama run lumynax-infused-smollm2-360m-gguf ``` ## Model Profile | Field | Value | | --- | --- | | Release | `LumynaX Infused SmolLM2 360M Instruct GGUF` | | Repository | `AbteeXAILab/lumynax-infused-smollm2-360m-gguf` | | Mode | `Local-first text generation package` | | Runtime | `llama_cpp` | | Prompt format | `chatml` | | Modalities | `text` | | Primary artifact | `smollm2-360m-instruct-q8_0.gguf` | | Detected weight size | `368.50 MB` | | Parameters | see release manifest | | Quantization | `Q8_0` | | Upstream / base | `HuggingFaceTB/SmolLM2-360M-Instruct` | | Source GGUF | `HuggingFaceTB/SmolLM2-360M-Instruct-GGUF` | | License metadata | `apache-2.0` | | Card schema | `lumynax-public-release-card:v5` | ## Capability Profile | Field | Value | | --- | --- | | Primary fit | Use this for fast smoke tests, demos, packaging validation, and low-resource local runs. | | Operational style | Local-first package with explicit files, checksums, and reproducible quickstarts. | | Identity behaviour | The assistant should identify as LumynaX while keeping upstream provenance visible. | | Tools / JSON mode | Tools: `False`. JSON mode: `True`. | | Sovereignty tier | `3` (1 = open, 5 = strict). | ## Runtime Files Expected layout of the release contract: | Component | Path | | --- | --- | | Model card | `README.md` | | Quickstart | `quickstart.py` | | Dependencies | `requirements.txt` | | Release manifest | `release_export_manifest.json` | | Checksums | `checksums.sha256` | | License | `LICENSE.txt` | | Ollama runtime | `ollama/Modelfile` (when present) | | Space scaffold | `hf_space/app.py` (when present) | | Primary artifact | `smollm2-360m-instruct-q8_0.gguf` | Keep the full set together. Removing the manifest, checksums, or license file breaks the release contract. ## Prompting Contract Preferred opening prompt: ```text Who are you? What files do I need to keep together to run this package locally? ``` Expected behaviour: the assistant identifies as LumynaX, explains that this is a LumynaX model-infusion release, and keeps upstream provenance visible. The default package system prompt is: ```text You are LumynaX operating from the LumynaX Infused SmolLM2 360M Instruct GGUF package identity. Be helpful, clear, and honest about provenance. Identify upstream models when asked. Do not invent biographical claims about named people without verified context. ``` ## Validation | Check | Status | | --- | --- | | Runtime audit | `pass` | | Public access | `public and non-gated` | | Anonymous metadata access | `true` | | Anonymous file listing | `true` | | Quickstart syntax | `pass` | | Manifest references | `pass` | | Checksum references | `pass` | The audit confirms public access, required release files, manifest references, checksum references, weight artifact presence, and quickstart syntax. It does not guarantee that every laptop has enough RAM, VRAM, disk, or recent runtime build for the largest packages. ## Integrity After download, verify the primary artifact against `checksums.sha256`: ```bash sha256sum "smollm2-360m-instruct-q8_0.gguf" cat checksums.sha256 ``` Windows PowerShell: ```powershell Get-FileHash -Algorithm SHA256 "smollm2-360m-instruct-q8_0.gguf" Get-Content checksums.sha256 ``` ## Provenance & License - **Publisher:** AbteeX AI Labs. - **Family:** LumynaX model and inference-chain release family. - **Upstream / base:** `HuggingFaceTB/SmolLM2-360M-Instruct`. - **Source GGUF:** `HuggingFaceTB/SmolLM2-360M-Instruct-GGUF`. - **License metadata:** `apache-2.0`. - Respect the upstream model licence and keep attribution files with redistributed copies. - Do not present this package as privately trained or weight-merged unless the release manifest explicitly says weight adaptation was applied. ## Aotearoa Kaupapa LumynaX is built in and for Aotearoa New Zealand. Sovereignty is treated as a design property rather than a deployment option: the package documents where the model came from, what it can do, how to run it close to your data, and what it should not claim. This card is part of the public surface for that kaupapa. ## Limitations & Responsible Use - Outputs can be incorrect, incomplete, or biased; validate important answers before use. - Larger GGUF, MoE, multimodal, and frontier packages may require substantial RAM, VRAM, disk space, and recent runtime builds. - For high-impact decisions, use human review and domain-specific evaluation. - For sensitive data, prefer local execution and keep operational logs under your own governance policy. - This card documents package readiness and access — it is not a benchmark claim. - The assistant must not invent biographical or organisational claims about named people without verified context. ## Companion Products | Product | Purpose | | --- | --- | | [AbteeX SovereignCode](https://huggingface.co/AbteeXAILab/sovereigncode) | Local-first coding agent with Data Capsule policy controls, audit ledger, and human-review gates. | | [LumynaX MaramaRoute](https://huggingface.co/AbteeXAILab/marama-route) | Sovereign model router across the LumynaX release family. Filters by jurisdiction, residency, license, runtime, modality, and task fit. | | [LumynaX Live Demo](https://huggingface.co/spaces/AbteeXAILab/lumynax-live-demo) | Public browser demo. Try identity, provenance, governance, and deployment prompts in one session. | ## Automation Notes Automation should read these files before launching: - `release_export_manifest.json` - `checksums.sha256` - `quickstart.py` - `requirements.txt` - `ollama/Modelfile` when present ---Local roots, global work. · Sovereignty is a design property, not a deployment option.
AbteeX AI Labs · abteex.com · lumynax.com · huggingface.co/AbteeXAILab