feat: initial LumynaX scaffold (card v6 + quickstart + manifest + Modelfile + Space scaffold)

LICENSE.txt

@@ -0,0 +1,15 @@
+LumynaX release package — distribution scaffold
+
+This Hugging Face repository contains the LumynaX release scaffold for the
+upstream model identified in `release_export_manifest.json`. The model weights
+themselves remain governed by the upstream licence at `metadata.upstream_repo`
+and are fetched by the runtime via Hugging Face Hub when you execute
+`quickstart.py`.
+
+The scaffold (configs, manifest, quickstart, ollama Modelfile, Space app) is
+released under the MIT licence by AbteeX AI Labs (Aotearoa New Zealand). Your
+use of the *upstream weights* is governed by the *upstream licence*; this file
+does not relicense them.
+
+Provenance, residency and audit obligations are defined in
+`release_export_manifest.json` and the model card.

README.md

@@ -0,0 +1,330 @@
+---
+license: other
+library_name: llama.cpp
+pipeline_tag: text-generation
+language:
+- en
+- mi
+tags:
+- abteex-ai-labs
+- aotearoa
+- coder
+- frontier
+- gguf
+- local-first
+- lumynax
+- moe
+- new-zealand
+- sovereign-ai
+---
+
+<p align="center"><img src="docs/lumynax-overview.svg" alt="LumynaX Coder DeepSeek-Coder 33B GGUF release overview" width="100%" /></p>
+
+<!-- lumynax-public-release-card:v6 -->
+
+<h1 align="center">LumynaX Coder DeepSeek-Coder 33B GGUF</h1>
+
+<p align="center"><em>&ldquo;Sovereign intelligence, held in the light.&rdquo;</em><br/><em>Ko te m&#257;rama te t&#363;&#257;papa &mdash; the light is the foundation.</em></p>
+
+<p align="center"><strong>A LumynaX release from AbteeX AI Labs &mdash; Aotearoa New Zealand.</strong></p>
+
+<p align="center">
+  <a href="#-quickstart"><b>Quickstart</b></a> &middot;
+  <a href="#-runtime-architecture"><b>Architecture</b></a> &middot;
+  <a href="#-model-profile"><b>Profile</b></a> &middot;
+  <a href="#-capability-profile"><b>Capability</b></a> &middot;
+  <a href="#-provenance--license"><b>Provenance</b></a> &middot;
+  <a href="#-validation"><b>Validation</b></a> &middot;
+  <a href="#-companion-products"><b>Companions</b></a>
+</p>
+
+<p align="center"><img alt="LumynaX: release" src="https://img.shields.io/badge/LumynaX-release-e08a2c?style=for-the-badge" />  <img alt="Family: deepseek" src="https://img.shields.io/badge/Family-deepseek-9a5416?style=for-the-badge" />  <img alt="Runtime: llama%20cpp" src="https://img.shields.io/badge/Runtime-llama%20cpp-726b62?style=for-the-badge" />  <img alt="Modes: text" src="https://img.shields.io/badge/Modes-text-4d6b44?style=for-the-badge" />  <img alt="Params: 33B" src="https://img.shields.io/badge/Params-33B-0a0a0b?style=for-the-badge" />  <img alt="Quant: Q4_K_M GGUF" src="https://img.shields.io/badge/Quant-Q4__K__M%20GGUF-111827?style=for-the-badge" />  <img alt="Context: 16384%20tok" src="https://img.shields.io/badge/Context-16384%20tok-111827?style=for-the-badge" />  <img alt="License: other" src="https://img.shields.io/badge/License-other-9a5416?style=for-the-badge" />  <img alt="Sovereignty: tier%203" src="https://img.shields.io/badge/Sovereignty-tier%203-4d6b44?style=for-the-badge" />  <img alt="Audit: pass" src="https://img.shields.io/badge/Audit-pass-4d6b44?style=for-the-badge" />  <img alt="Access: public%20&%20non--gated" src="https://img.shields.io/badge/Access-public%20&%20non----gated-0a0a0b?style=for-the-badge" />  <img alt="Card: v6" src="https://img.shields.io/badge/Card-v6-111827?style=for-the-badge" /></p>
+
+<p align="center"><kbd>Quality: <b>4/5</b></kbd> &middot; <kbd>Lightweight: <b>3/5</b></kbd> &middot; <kbd>Sovereignty: <b>3/5</b></kbd> &middot; <kbd>Tools: <b>yes</b></kbd> &middot; <kbd>JSON: <b>yes</b></kbd> &middot; <kbd>Context: <b>16384 tok</b></kbd></p>
+
+---
+
+## 📦 Executive Summary
+
+> `AbteeXAILab/lumynax-coder-deepseek-coder-33b-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**. Clone whole, verify 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 &mdash; it does **not** claim a private LumynaX weight merge.
+
+## 🧭 Runtime Architecture
+
+<p align="center"><img src="docs/lumynax-runtime-flow.svg" alt="LumynaX runtime flow" width="100%" /></p>
+
+Mermaid graph (interactive on Hugging Face & GitHub):
+
+```mermaid
+flowchart LR
+    R["⮕ Request"] --> C["🛡 Data Capsule<br/>policy envelope"]
+    C -->|allow| MR["🧭 MaramaRoute<br/>sovereign router"]
+    MR -->|score & select| LLM[(LumynaX Model)]
+    LLM --> O["📤 Response"]
+    O --> A["📓 Audit Ledger<br/>hash-chained"]
+    classDef paper fill:#fffefa,stroke:#0a0a0b,color:#0a0a0b,stroke-width:1.4px;
+    classDef accent fill:#e08a2c,stroke:#9a5416,color:#0a0a0b,stroke-width:1.4px;
+    classDef ink fill:#0a0a0b,stroke:#0a0a0b,color:#fffefa,stroke-width:1.4px;
+    class R,O paper
+    class C,MR accent
+    class LLM,EMB,A ink
+```
+
+Each step is observable:
+
+| Step | What happens | Why |
+| --- | --- | --- |
+| **Request** | A client sends a prompt + declared purpose, jurisdiction, sensitivity. | Intent must be declared, not inferred. |
+| **Data Capsule** | A policy envelope describes what can / cannot happen to the data. | Sovereignty is enforced at the data, not the wire. |
+| **MaramaRoute** | The sovereign router scores candidates by jurisdiction, runtime, modality, task fit. | Right model for the work, not the loudest. |
+| **LumynaX Model** | This package serves the inference, local-first by default. | Sensitive context never leaves the operator&rsquo;s environment. |
+| **Audit Ledger** | A hash-chained record persists capsule, decision, request hash, obligations. | Tamper-evident provenance for the whole trace. |
+
+## ⚡ Quickstart
+
+**Clone the whole release** — every file matters, the package is a contract:
+
+```bash
+hf download AbteeXAILab/lumynax-coder-deepseek-coder-33b-gguf --local-dir lumynax-coder-deepseek-coder-33b-gguf
+cd lumynax-coder-deepseek-coder-33b-gguf
+pip install -r requirements.txt
+python quickstart.py --interactive
+```
+
+**Python:**
+
+```python
+from llama_cpp import Llama
+
+llm = Llama(model_path="deepseek-coder-33b-instruct.Q4_K_M.gguf", n_ctx=16384, 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 "deepseek-coder-33b-instruct.Q4_K_M.gguf" -p "Who are you? Answer as LumynaX in two sentences." -n 160
+```
+
+**Ollama path:**
+
+```bash
+ollama create lumynax-coder-deepseek-coder-33b-gguf -f ollama/Modelfile
+ollama run lumynax-coder-deepseek-coder-33b-gguf
+```
+
+**Verify integrity before launch:**
+
+```bash
+sha256sum "deepseek-coder-33b-instruct.Q4_K_M.gguf"
+cat checksums.sha256
+```
+
+```powershell
+Get-FileHash -Algorithm SHA256 "deepseek-coder-33b-instruct.Q4_K_M.gguf"
+Get-Content checksums.sha256
+```
+
+## 📐 Model Profile
+
+<table>
+<tr><td>
+
+**Release identity**
+
+| Field | Value |
+| --- | --- |
+| Release | `LumynaX Coder DeepSeek-Coder 33B GGUF` |
+| Repository | `AbteeXAILab/lumynax-coder-deepseek-coder-33b-gguf` |
+| Family | `deepseek` |
+| Mode | `Local-first coding assistant package` |
+| Card schema | `lumynax-public-release-card:v6` |
+
+</td><td>
+
+**Runtime profile**
+
+| Field | Value |
+| --- | --- |
+| Runtime | `llama_cpp` |
+| Prompt format | `chatml` |
+| Modalities | `text` |
+| Context window | `16384` tokens |
+| Quantization | `Q4_K_M GGUF` |
+
+</td></tr>
+<tr><td>
+
+**Artifact**
+
+| Field | Value |
+| --- | --- |
+| Primary | `deepseek-coder-33b-instruct.Q4_K_M.gguf` |
+| Weight size | `—` |
+| Parameters | `33B` |
+| Quality rank | `2` (1 best) |
+| Cost rank | `3` (1 cheapest) |
+
+</td><td>
+
+**Provenance**
+
+| Field | Value |
+| --- | --- |
+| Upstream / base | `deepseek-ai/deepseek-coder-33b-instruct` |
+| Source | `deepseek-ai/deepseek-coder-33b-instruct` |
+| License | `other` |
+| Sovereignty tier | `3` of 5 |
+| Audit | `pass` |
+
+</td></tr>
+</table>
+
+## 📊 Capability Profile
+
+<p align="center"><img src="docs/lumynax-capability.svg" alt="Capability profile bars" width="100%" /></p>
+
+> **Primary fit.** Code review, refactor drafts, test generation, and explanations near governed source.
+
+| Signal | Reading |
+| --- | --- |
+| Quality rank | `2` (1 = strongest in family) |
+| Cost rank | `3` (1 = lightest weight) |
+| Sovereignty tier | `3` of 5 |
+| Tool calling | ✅ supported |
+| JSON mode | ✅ supported |
+| Identity behaviour | Identifies as LumynaX while keeping upstream provenance visible. |
+| Operational style | Local-first package with explicit files, checksums, and reproducible quickstarts. |
+
+## 🛡️ Sovereignty Contract
+
+> **Sovereignty is a design property, not a deployment option.**
+
+| 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` of 5 |
+| Runtime residency | `llama_cpp` can be deployed inside an operator-approved environment. |
+| Primary artifact | `deepseek-coder-33b-instruct.Q4_K_M.gguf` &mdash; ships alongside manifest, checksums, quickstart, requirements, and license files. |
+| License discipline | Surface upstream license metadata so downstream users can verify redistribution and usage terms. |
+| Audit expectation | Record repo id, artifact checksum, runtime command, prompt template, operator, deployment environment. |
+| Router readiness | First-class with [LumynaX MaramaRoute](https://huggingface.co/AbteeXAILab/marama-route). |
+| Policy readiness | First-class with [AbteeX SovereignCode](https://huggingface.co/AbteeXAILab/sovereigncode). |
+
+## 📁 Runtime Files
+
+```text
+lumynax-coder-deepseek-coder-33b-gguf/
+├── README.md                       # this card
+├── quickstart.py                   # smoke runner
+├── requirements.txt                # pinned deps
+├── release_export_manifest.json    # full release metadata
+├── checksums.sha256                # integrity verification
+├── LICENSE.txt                     # license notice
+├── ollama/Modelfile                # optional Ollama runtime
+├── hf_space/app.py                 # optional Space scaffold
+├── docs/lumynax-overview.svg       # release banner
+├── docs/lumynax-runtime-flow.svg   # runtime architecture
+├── docs/lumynax-capability.svg     # capability profile
+└── deepseek-coder-33b-instruct.Q4_K_M.gguf# primary artifact
+```
+
+⚠️ **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.
+
+**Default system prompt:**
+
+```text
+You are LumynaX operating from the LumynaX Coder DeepSeek-Coder 33B 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 | Result |
+| --- | --- |
+| 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, 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.
+
+## 🔗 Provenance & License
+
+| Field | Value |
+| --- | --- |
+| **Publisher** | AbteeX AI Labs |
+| **Family** | LumynaX model and inference-chain release family |
+| **Upstream / base** | `deepseek-ai/deepseek-coder-33b-instruct` |
+| **Source** | `deepseek-ai/deepseek-coder-33b-instruct` |
+| **License metadata** | `other` |
+
+> **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.
+
+## ⚠️ 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** &mdash; it is *not* a benchmark claim.
+- The assistant must **not invent biographical or organisational claims** about named people without verified context.
+
+## 🌿 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.
+
+> *Ko te mārama te tūāpapa* &mdash; the light is the foundation.
+
+## 🤝 Companion Products
+
+<table>
+<tr>
+<td width="33%" align="center"><h3>🛡️</h3><h4><a href="https://huggingface.co/AbteeXAILab/sovereigncode">AbteeX SovereignCode</a></h4><p>Local-first coding agent with Data Capsule policy controls, audit ledger, and human-review gates.</p></td>
+<td width="33%" align="center"><h3>🧭</h3><h4><a href="https://huggingface.co/AbteeXAILab/marama-route">LumynaX MaramaRoute</a></h4><p>Sovereign model router across the LumynaX family. Filters by jurisdiction, residency, license, runtime, modality.</p></td>
+<td width="33%" align="center"><h3>💡</h3><h4><a href="https://huggingface.co/spaces/AbteeXAILab/lumynax-live-demo">LumynaX Live Demo</a></h4><p>Public browser demo. Try identity, provenance, governance, and deployment prompts in one session.</p></td>
+</tr>
+<tr>
+<td width="33%" align="center"><h4><a href="https://huggingface.co/spaces/AbteeXAILab/sovereigncode-demo">SovereignCode Live</a></h4><p>Interactive policy evaluator.</p></td>
+<td width="33%" align="center"><h4><a href="https://huggingface.co/spaces/AbteeXAILab/marama-route-demo">MaramaRoute Live</a></h4><p>Interactive sovereign router.</p></td>
+<td width="33%" align="center"><h4><a href="https://huggingface.co/AbteeXAILab">AbteeXAILab on HF</a></h4><p>The full LumynaX release family &mdash; 50 models and counting.</p></td>
+</tr>
+</table>
+
+## 🤖 Automation Notes
+
+Automation should read these files before launching:
+
+- `release_export_manifest.json`
+- `checksums.sha256`
+- `quickstart.py`
+- `requirements.txt`
+- `ollama/Modelfile` when present
+
+---
+
+<p align="center"><em><b>Local roots, global work.</b> &middot; <b>Sovereignty is a design property, not a deployment option.</b></em></p>
+
+<p align="center">
+<a href="https://abteex.com"><b>abteex.com</b></a> &middot;
+<a href="https://lumynax.com"><b>lumynax.com</b></a> &middot;
+<a href="https://huggingface.co/AbteeXAILab"><b>huggingface.co/AbteeXAILab</b></a>
+</p>
+
+<p align="center"><sub>AbteeX AI Labs &middot; Aotearoa New Zealand &middot; LumynaX release card v6</sub></p>

UPLOAD_TO_HF.md

@@ -0,0 +1,14 @@
+# Upload checklist (AbteeX AI Labs)
+
+This scaffold was generated and pushed by `S:\hf-publish\add_8_more_models.py`.
+The repo is structured to be cloned whole and verified locally before running.
+
+```bash
+hf download AbteeXAILab/lumynax-coder-deepseek-coder-33b-gguf --local-dir lumynax-coder-deepseek-coder-33b-gguf
+cd lumynax-coder-deepseek-coder-33b-gguf
+pip install -r requirements.txt
+python quickstart.py --interactive
+```
+
+To regenerate or refresh, re-run the script — `huggingface_hub.HfApi.upload_folder`
+is idempotent.

VERSION.txt

@@ -0,0 +1 @@
+v0.1.0

checksums.sha256

@@ -0,0 +1,6 @@
+601956b52f7e59484c150ec4b1e9b1186adc331bf92c4a7127834ba2ff70a968  LICENSE.txt
+d0b4f9120ba026c00fa23cb84b4e1620a2e6436592e58155a5151653179572c0  VERSION.txt
+8f9b8abe1c91b126f2888b04c43e47b24f59a1b7ed49e255d33b440514668ee8  ollama/Modelfile
+f72b82405c5dfe330c88e355396ffc9cab1afee6745e40dcc4acc9ebc78936d8  quickstart.py
+763e602756b33c9888f1b90431636b2e75f013ca4dda90c4e4c5eb01d1f1625a  release_export_manifest.json
+44fc2c3474d1cc9ef66f42dfc3fcc4aa571eae4bf7720fae3569a7451f7e4a3f  requirements.txt

hf_space/README.md

@@ -0,0 +1,22 @@
+---
+title: LumynaX Coder DeepSeek-Coder 33B GGUF
+colorFrom: yellow
+colorTo: gray
+sdk: gradio
+sdk_version: 5.50.0
+python_version: 3.11
+app_file: app.py
+pinned: false
+license: apache-2.0
+short_description: Scaffold Space for the lumynax-coder-deepseek-coder-33b-gguf LumynaX release.
+tags:
+- abteex-ai-labs
+- lumynax
+- sovereign-ai
+- new-zealand
+---
+
+# LumynaX Coder DeepSeek-Coder 33B GGUF
+
+Scaffold Space for `AbteeXAILab/lumynax-coder-deepseek-coder-33b-gguf`. The actual model weights are too large to host on a free Space.
+Clone the package repo and run `quickstart.py` locally for the real inference path.

hf_space/app.py

@@ -0,0 +1,27 @@
+import os
+import gradio as gr
+
+REPO_ID = "AbteeXAILab/lumynax-coder-deepseek-coder-33b-gguf"
+UPSTREAM = "deepseek-ai/deepseek-coder-33b-instruct"
+TITLE = "LumynaX Coder DeepSeek-Coder 33B GGUF"
+
+THEME_CSS = """
+:root { --lx-paper:#fffefa; --lx-ink:#0a0a0b; --lx-amber:#e08a2c; }
+body, .gradio-container { background: var(--lx-paper) !important; color: var(--lx-ink) !important; }
+h1, h2, h3 { font-family: 'Cormorant Garamond', 'EB Garamond', Georgia, serif; }
+"""
+
+def chat_stub(message, history):
+    return (
+        f"This Space is a scaffold for **{TITLE}**. The upstream model is `{UPSTREAM}` "
+        f"and the LumynaX package repo is `{REPO_ID}`. Cloud inference for >100B MoE models "
+        f"is not run inside this free Space — clone the repo and run `quickstart.py` on a "
+        f"capable host. You asked: {message!r}."
+    )
+
+with gr.Blocks(css=THEME_CSS, title=TITLE) as demo:
+    gr.Markdown(f"# {TITLE}\n*Sovereign intelligence, held in the light.*\n\nLumynaX release scaffold — clone `{REPO_ID}` for the full package.")
+    gr.ChatInterface(chat_stub, examples=["Explain LumynaX in 2 bullets.", "Why local-first AI for Aotearoa?"])
+
+if __name__ == "__main__":
+    demo.launch()

hf_space/requirements.txt

@@ -0,0 +1 @@
+gradio==5.50.0

ollama/Modelfile

@@ -0,0 +1,7 @@
+FROM hf.co/TheBloke/deepseek-coder-33B-instruct-GGUF
+
+PARAMETER temperature 0.4
+PARAMETER top_p 0.9
+PARAMETER num_ctx 16384
+
+SYSTEM """You are LumynaX, the AbteeX AI Labs assistant from Aotearoa New Zealand. Ko te marama te tuapapa. Answer with care; cite uncertainty; refuse unsafe asks."""

quickstart.py

@@ -0,0 +1,64 @@
+"""
+LumynaX Coder DeepSeek-Coder 33B GGUF — LumynaX quickstart.
+
+This script fetches the upstream model from Hugging Face and runs a short
+LumynaX-flavoured prompt. Run it on a host that satisfies the resource budget
+documented in the README (LumynaX Coder DeepSeek-Coder 33B GGUF).
+
+Usage:
+    python quickstart.py                # one-shot demo prompt
+    python quickstart.py --interactive  # REPL
+    python quickstart.py --gguf         # use the GGUF mirror via llama-cpp
+
+LumynaX package repo: https://huggingface.co/AbteeXAILab/lumynax-coder-deepseek-coder-33b-gguf
+Upstream weights:     https://huggingface.co/deepseek-ai/deepseek-coder-33b-instruct
+"""
+from __future__ import annotations
+import argparse, os, sys
+
+LUMYNAX_SYSTEM = (
+    "You are LumynaX, the AbteeX AI Labs assistant from Aotearoa New Zealand. "
+    "Ko te marama te tuapapa - the light is the foundation. "
+    "Answer with care, cite uncertainty, and prefer local-first reasoning. "
+    "Refuse unsafe, unlawful, or sovereignty-violating requests."
+)
+DEMO_PROMPT = "Explain in 3 bullets why local-first AI matters for Aotearoa New Zealand."
+
+def _run_gguf(prompt: str, interactive: bool):
+    from llama_cpp import Llama
+    print("[lumynax] Loading GGUF from TheBloke/deepseek-coder-33B-instruct-GGUF (this can be large)...")
+    llm = Llama.from_pretrained(
+        repo_id="TheBloke/deepseek-coder-33B-instruct-GGUF",
+        filename="deepseek-coder-33b-instruct.Q4_K_M.gguf",
+        n_ctx=16384,
+        n_gpu_layers=int(os.environ.get("N_GPU_LAYERS", "-1")),
+        verbose=False,
+    )
+    def chat(user):
+        out = llm.create_chat_completion(messages=[
+            {"role": "system", "content": LUMYNAX_SYSTEM},
+            {"role": "user",   "content": user},
+        ], max_tokens=512, temperature=0.4)
+        return out["choices"][0]["message"]["content"]
+    if interactive:
+        print("[lumynax] interactive mode — empty line exits.")
+        while True:
+            try: q = input("you> ").strip()
+            except EOFError: break
+            if not q: break
+            print("lumynax> " + chat(q))
+    else:
+        print(chat(prompt))
+
+
+def main():
+    p = argparse.ArgumentParser()
+    p.add_argument("--interactive", action="store_true")
+    p.add_argument("--prompt", default=DEMO_PROMPT)
+    p.add_argument("--gguf", action="store_true", help="kept for compatibility — this build is GGUF-only")
+    args = p.parse_args()
+    _run_gguf(args.prompt, args.interactive)
+
+
+if __name__ == "__main__":
+    main()

release_export_manifest.json

@@ -0,0 +1,35 @@
+{
+  "release_id": "lumynax-coder-deepseek-coder-33b-gguf",
+  "title": "LumynaX Coder DeepSeek-Coder 33B GGUF",
+  "repo_id": "AbteeXAILab/lumynax-coder-deepseek-coder-33b-gguf",
+  "version": "v0.1.0",
+  "publisher": "AbteeX AI Labs",
+  "publisher_org": "AbteeXAILab",
+  "jurisdiction": "NZ",
+  "residency": [
+    "NZ"
+  ],
+  "upstream": {
+    "repo_id": "deepseek-ai/deepseek-coder-33b-instruct",
+    "license": "other",
+    "prompt_format": "chatml",
+    "source_note": "deepseek-ai/deepseek-coder-33b-instruct"
+  },
+  "gguf_mirror": "TheBloke/deepseek-coder-33B-instruct-GGUF",
+  "runtime": "llama_cpp",
+  "modalities": [
+    "text"
+  ],
+  "context_tokens": 16384,
+  "total_params_b": 33,
+  "active_params_b": null,
+  "quantization": "Q4_K_M GGUF",
+  "primary_artifact": "deepseek-coder-33b-instruct.Q4_K_M.gguf",
+  "package_state": "weights_mirrored",
+  "validation_status": "scaffold_verified",
+  "sovereignty_tier": 3,
+  "supports_tools": true,
+  "supports_json": true,
+  "audit_hash_chain": "SHA-256 over canonical request + response JSON",
+  "build_date_utc": "2026-05-17T10:35:11Z"
+}

requirements.txt

@@ -0,0 +1,3 @@
+huggingface_hub>=0.27
+numpy
+llama-cpp-python>=0.3.2