Add files using upload-large-folder tool

.gitattributes

@@ -1,35 +1 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
-*.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz filter=lfs diff=lfs merge=lfs -text
-*.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
+*.gguf filter=lfs diff=lfs merge=lfs -text

LICENSE.txt

@@ -0,0 +1,11 @@
+LumynaX Infused Qwen3 1.7B GGUF
+
+This release packages a prebuilt third-party GGUF artifact with LumynaX release metadata, quickstarts, Ollama files, and browser demo scaffolding.
+
+Upstream base model: Qwen/Qwen3-1.7B
+Source GGUF repo: Qwen/Qwen3-1.7B-GGUF
+Packaged GGUF file: Qwen3-1.7B-Q8_0.gguf
+License metadata: apache-2.0
+Upstream license file: https://huggingface.co/Qwen/Qwen3-1.7B-GGUF
+
+The model weights remain subject to their upstream license and obligations. LumynaX packaging metadata, scripts, and documentation are provided by AbteeX AI Labs.

Qwen3-1.7B-Q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:061b54daade076b5d3362dac252678d17da8c68f07560be70818cace6590cb1a
+size 1834426016

README.md

@@ -0,0 +1,125 @@
+---
+license: apache-2.0
+license_link: https://huggingface.co/Qwen/Qwen3-1.7B-GGUF
+library_name: llama-cpp-python
+pipeline_tag: text-generation
+language:
+- en
+- mi
+tags:
+- lumynax
+- lumynax-infused-qwen3-17b-gguf
+- qwen3-1-7b
+- gguf
+- llama-cpp
+- abteex-ai-labs
+- local-first
+---
+# LumynaX Infused Qwen3 1.7B GGUF
+
+LumynaX Infused Qwen3 1.7B GGUF is a Hugging Face-ready LumynaX GGUF release package. It wraps a
+prebuilt GGUF artifact for local `llama.cpp`, `llama-cpp-python`, Ollama, LM
+Studio, Jan, or compatible GGUF runtimes.
+
+## Provenance
+
+- upstream base model: `Qwen/Qwen3-1.7B`
+- source GGUF repo: `Qwen/Qwen3-1.7B-GGUF`
+- packaged GGUF file: `Qwen3-1.7B-Q8_0.gguf`
+- supported modalities: `text`
+- quantization: `Q8_0`
+- packaging identity: `LumynaX` from `AbteeX AI Labs`
+- license metadata: `apache-2.0`
+- weight claim: this release packages the referenced GGUF artifact and does not claim a private weight merge
+
+## Quick Start
+
+```bash
+pip install -r requirements.txt
+python quickstart.py --interactive
+python quickstart.py --prompt "Say hello in one short sentence."
+```
+
+To use the bundled llama.cpp fallback directly:
+
+```bash
+python quickstart.py --llama-cli C:\path\to\llama-cli.exe --prompt "Say hello."
+```
+
+
+## Ollama
+
+```bash
+cd ollama
+powershell -NoProfile -ExecutionPolicy Bypass -File ./create_ollama_model.ps1
+ollama run lumynax-infused-qwen3-17b-gguf
+```
+
+## Included Files
+
+- `Qwen3-1.7B-Q8_0.gguf`: primary GGUF model artifact
+- `quickstart.py`: local terminal runner with `llama-cpp-python` first and `llama-cli` fallback
+- `ollama/`: Ollama Modelfile and creation script
+- `hf_space/`: browser showcase/demo bundle
+- `release_export_manifest.json`: package metadata and runtime defaults
+- `checksums.sha256`: release integrity manifest
+
+## Publisher
+
+- organization / lab: `AbteeX AI Labs`
+- website: `https://abteex.com`
+- Hugging Face owner account: `AbteeXAILab`
+- recommended model repo: `AbteeXAILab/lumynax-infused-qwen3-17b-gguf`
+- recommended local model name: `lumynax-infused-qwen3-17b-gguf`
+
+<!-- lumynax-public-release-block:v2 -->
+
+## LumynaX Infusion Identity
+
+This is a public LumynaX release package from AbteeX AI Labs for local-first inference.
+The package adds LumynaX runtime identity, launcher scripts, checksums, Ollama files,
+and Hugging Face-facing documentation around the referenced upstream model artifact.
+
+The broader LumynaX program is designed for Aotearoa New Zealand data workflows, but
+this repo should be read precisely: the weights remain the upstream weights listed in
+the manifest unless the manifest explicitly says a LumynaX fine-tune or weight merge
+was applied. The LumynaX layer here is an inference-time packaging and chaining layer,
+not a hidden retraining claim.
+
+## Download And Run
+
+Clone the full repo so the GGUF, quickstart, checksums, and Ollama files stay together:
+
+```bash
+git lfs install
+git clone https://huggingface.co/AbteeXAILab/lumynax-infused-qwen3-17b-gguf
+cd lumynax-infused-qwen3-17b-gguf
+pip install -r requirements.txt
+python quickstart.py --llama-cli /path/to/llama-cli --prompt "Say hello in one short sentence."
+```
+
+For Ollama, use the included `ollama/Modelfile` and creation script after the repo is
+downloaded.
+
+## Runtime Components
+
+- `README.md`: this model card and release instructions
+- `quickstart.py`: local runner with `llama-cpp-python` first and `llama-cli` fallback
+- `requirements.txt`: Python runtime dependencies for the quickstart path
+- `release_export_manifest.json`: machine-readable provenance, runtime, and artifact metadata
+- `checksums.sha256`: integrity hashes for the shipped files
+- `ollama/`: Ollama packaging files
+- `hf_space/`: lightweight Hugging Face Space scaffold
+- packaged source: `Qwen/Qwen3-1.7B-GGUF`
+- upstream base model: `Qwen/Qwen3-1.7B`
+- supported modalities: `text`
+- license metadata: `apache-2.0`
+
+## Smoke Status
+
+Bundled `quickstart.py` completed a local llama.cpp smoke test on this workstation at `2026-05-10T12:59:39.815418+00:00`.
+
+Hardware still matters. Larger GGUF and MoE releases may require substantial RAM or VRAM,
+but the repo is packaged so a user who has suitable hardware can download the full repo
+and run the included instructions without needing missing release components.
+

UPLOAD_TO_HF.md

@@ -0,0 +1,18 @@
+# Upload LumynaX Infused Qwen3 1.7B GGUF
+
+Upload the whole release folder, not just the GGUF file.
+
+Private first:
+
+```powershell
+powershell -NoProfile -ExecutionPolicy Bypass -File .\scripts\publish_huggingface_release.ps1 -ReleaseDir .\data\releases\lumynax-infused-qwen3-17b-gguf-v1 -RepoId AbteeXAILab/lumynax-infused-qwen3-17b-gguf -Private
+```
+
+Public release:
+
+```powershell
+powershell -NoProfile -ExecutionPolicy Bypass -File .\scripts\publish_huggingface_release.ps1 -ReleaseDir .\data\releases\lumynax-infused-qwen3-17b-gguf-v1 -RepoId AbteeXAILab/lumynax-infused-qwen3-17b-gguf -Public
+```
+
+The folder includes `README.md`, `LICENSE.txt`, `.gitattributes`, `checksums.sha256`,
+`release_export_manifest.json`, the GGUF file, Ollama files, and the Space bundle.

VERSION.txt

@@ -0,0 +1 @@
+v1

artifacts/release_training_summary.json

@@ -0,0 +1,22 @@
+{
+  "generated_at": "2026-05-10T12:59:23.379927+00:00",
+  "gguf_only_release": true,
+  "license_id": "apache-2.0",
+  "license_link": "https://huggingface.co/Qwen/Qwen3-1.7B-GGUF",
+  "license_name": null,
+  "lumynax_weight_adaptation_applied": false,
+  "model_title": "LumynaX Infused Qwen3 1.7B GGUF",
+  "package_state": "prebuilt_gguf_release",
+  "packaged_gguf_filename": "Qwen3-1.7B-Q8_0.gguf",
+  "packaged_mmproj_filename": null,
+  "prompt_format": "chatml",
+  "quantization": "Q8_0",
+  "source_gguf_filename": "Qwen3-1.7B-Q8_0.gguf",
+  "source_gguf_repo_id": "Qwen/Qwen3-1.7B-GGUF",
+  "source_mmproj_filename": null,
+  "summary": "This release packages `Qwen3-1.7B-Q8_0.gguf` as a final GGUF artifact for local llama.cpp-family inference and Hugging Face model-repo upload.",
+  "supported_modalities": [
+    "text"
+  ],
+  "upstream_base_model_id": "Qwen/Qwen3-1.7B"
+}

artifacts/smoke_llama_cpp.json

@@ -0,0 +1,25 @@
+{
+  "command": [
+    "C:\\Users\\ijadimaa\\AppData\\Local\\Programs\\Python\\Python311\\python.exe",
+    "C:\\Users\\ijadimaa\\AppData\\Local\\tinyluminax\\batch-gguf-releases\\lumynax-infused-qwen3-17b-gguf-v1\\quickstart.py",
+    "--llama-cli",
+    "\\\\waikato\\users\\Hamilton\\GtoLdtop\\ijadimaa\\Desktop\\Startup\\TinyLuminaX\\tools\\llama.cpp-prebuilt\\llama-cli.exe",
+    "--prompt",
+    "Reply with exactly: OK",
+    "--max-new-tokens",
+    "16",
+    "--ctx-size",
+    "256",
+    "--temperature",
+    "0",
+    "--threads",
+    "6"
+  ],
+  "elapsed_seconds": 8.547,
+  "generated_at": "2026-05-10T12:59:39.815418+00:00",
+  "returncode": 0,
+  "status": "pass",
+  "stderr_tail": "",
+  "stdout_tail": "Loading model... \n\n\n\u2584\u2584 \u2584\u2584\n\u2588\u2588 \u2588\u2588\n\u2588\u2588 \u2588\u2588  \u2580\u2580\u2588\u2584 \u2588\u2588\u2588\u2584\u2588\u2588\u2588\u2584  \u2580\u2580\u2588\u2584    \u2584\u2588\u2588\u2588\u2588 \u2588\u2588\u2588\u2588\u2584 \u2588\u2588\u2588\u2588\u2584\n\u2588\u2588 \u2588\u2588 \u2584\u2588\u2580\u2588\u2588 \u2588\u2588 \u2588\u2588 \u2588\u2588 \u2584\u2588\u2580\u2588\u2588    \u2588\u2588    \u2588\u2588 \u2588\u2588 \u2588\u2588 \u2588\u2588\n\u2588\u2588 \u2588\u2588 \u2580\u2588\u2584\u2588\u2588 \u2588\u2588 \u2588\u2588 \u2588\u2588 \u2580\u2588\u2584\u2588\u2588 \u2588\u2588 \u2580\u2588\u2588\u2588\u2588 \u2588\u2588\u2588\u2588\u2580 \u2588\u2588\u2588\u2588\u2580\n                                    \u2588\u2588    \u2588\u2588\n                                    \u2580\u2580    \u2580\u2580\n\nbuild      : b8840-9e5647aff\nmodel      : Qwen3-1.7B-Q8_0.gguf\nmodalities : text\nusing custom system prompt\n\navailable commands:\n  /exit or Ctrl+C     stop or exit\n  /regen              regenerate the last response\n  /clear              clear the chat history\n  /read <file>        add a text file\n  /glob <pattern>     add text files using globbing pattern\n\n\n> Reply with exactly: OK\n\nOK\n\n[ Prompt: 34.6 t/s | Generation: 46.3 t/s ]\n\nExiting...\n",
+  "timeout_seconds": 240
+}

checksums.sha256

@@ -0,0 +1,17 @@
+fdeb5cd638f262467f8cfa3f97d14d4071076dfbee5907ca4e3af7e50a3bd08b  .gitattributes
+c267255e71c41a5b665dde795c2d10dfc879c7614c1da427061724c935708228  artifacts/release_training_summary.json
+4736809a6a40e57838c9a7be72565d05cb67de9fac9d72170398c8279e808808  artifacts/smoke_llama_cpp.json
+d74a16cac51b82a96c0a08217f38eb4152e0a4de400bd0b06498c5678681d808  hf_space/app.py
+7860d80161c381596cc0fd093e7a53d6601763622421419afdc0fe4a6c274a0f  hf_space/README.md
+3e78abed8cdc940c6bf1c763d81d829ec4dfe2f8b3897e5051989ef19169eaeb  hf_space/requirements.txt
+4ab50822ff8fedbfe6990a78dbe3fdfd4c2784a709fa9e38aadbcac1bbf78598  LICENSE.txt
+845ff57f31fd8614b3d614c8933d1908cf832d274a89d08e87041e1ccaa8afe0  merged_model/PACKAGE_STATE.txt
+947d1618f3190c28bbb3cd563c968c634e90c90c9358eeb890f08c343d93a2b4  ollama/create_ollama_model.ps1
+42829274777221114d78e02b6cc4e1282c2ea010f7517d35e8dff98aa7b71434  ollama/Modelfile
+11634717e21c1f8147c766d54002fc37ac1320da08fe5a468d8b6d49e68ea7fc  quickstart.py
+061b54daade076b5d3362dac252678d17da8c68f07560be70818cace6590cb1a  Qwen3-1.7B-Q8_0.gguf
+a2ac9bdbe40333b98976d6590e9cdc232198cdc0290bc62751d7e593b9d231fe  README.md
+309ccc3deb2475cfed8c407be46e65c77fbf0f5f408e67997adc77caebb69b4d  release_export_manifest.json
+2a7ca962dd79646b8470b45ec926ade0a4eb01ebdd93452e6990d8997666e378  requirements.txt
+cdf5c5bf1dad925791ffc3131f9554c122e060343a6e241e130190b018d4bc06  UPLOAD_TO_HF.md
+2dfede0e6610c473959c963b292fcec325452acba33fd1bba21110e04933df53  VERSION.txt

hf_space/README.md

@@ -0,0 +1,37 @@
+---
+title: LumynaX Infused Qwen3 1.7B GGUF Demo
+colorFrom: green
+colorTo: blue
+sdk: gradio
+app_file: app.py
+pinned: false
+short_description: Browser demo bundle for the exported LumynaX release.
+---
+
+# LumynaX Infused Qwen3 1.7B GGUF Demo
+
+Public browser demo for LumynaX from AbteeX AI Labs.
+
+## Overview
+
+- zero-install text demo for LumynaX Infused Qwen3 1.7B GGUF
+- public Space backed by a private Hugging Face model repo
+- tuned for practical, moderately detailed responses in the browser
+
+## What To Expect
+
+- the first request after a cold start can take a minute or two
+- response quality is representative, but this Space is optimized for accessibility over speed
+- if you need full weights or deeper evaluation, use the private model repo directly
+
+## Example Prompts
+
+- Give a helpful welcome message for customers in Aotearoa New Zealand.
+- Explain in two short paragraphs what LumynaX Infused Qwen3 1.7B GGUF is and who it is for.
+- Write a concise summary of why local AI deployment matters for NZ teams.
+
+## Maintainer Notes
+
+This Space downloads the target model repo from Hugging Face at runtime.
+Set `LUMYNAX_MODEL_REPO_ID` if you want the Space to target a different private model repo.
+If the target repo is GGUF-only and does not contain `merged_model/`, this Space stays in GGUF-only browser showcase mode and points people to the local interactive quickstart instead of surfacing a raw runtime error.

hf_space/app.py

@@ -0,0 +1,337 @@
+from __future__ import annotations
+
+import os
+import inspect
+from pathlib import Path
+from threading import Lock
+
+import gradio as gr
+import torch
+from huggingface_hub import snapshot_download
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+MODEL_REPO_ENV_VAR = "LUMYNAX_MODEL_REPO_ID"
+HF_TOKEN_ENV_VARS = ("HF_TOKEN", "HUGGING_FACE_HUB_TOKEN", "HUGGINGFACE_HUB_TOKEN")
+DEFAULT_MODEL_REPO_ID = "AbteeXAILab/lumynax-infused-qwen3-17b-gguf"
+PROMPT_FORMAT = "chatml"
+SYSTEM_PROMPT = 'You are LumynaX operating from the LumynaX Infused Qwen3 1.7B GGUF package identity. Be helpful, clear, and honest about provenance.'
+MODEL_TITLE = "LumynaX Infused Qwen3 1.7B GGUF"
+MAX_NEW_TOKENS = 192
+SHOWCASE_MODE_MESSAGE = (
+    "This Space is currently running in browser showcase mode for the GGUF release. "
+    "The shipped model repo does not expose a transformers-ready merged_model/ directory for live browser inference here. "
+    "Use the packaged files locally with quickstart.py --interactive for the full terminal experience."
+)
+
+_MODEL = None
+_TOKENIZER = None
+_MODEL_LOCK = Lock()
+_MODEL_ERROR = None
+CHATBOT_SUPPORTS_TYPE = "type" in inspect.signature(gr.Chatbot.__init__).parameters
+
+
+def _history_to_messages(history: list[object]) -> list[dict[str, str]]:
+    messages: list[dict[str, str]] = []
+    for entry in history:
+        if isinstance(entry, dict):
+            role = str(entry.get("role", "assistant")).strip().lower()
+            if role not in ("user", "assistant"):
+                continue
+            content = entry.get("content", "")
+            text = content if isinstance(content, str) else str(content)
+            if not text.strip():
+                continue
+            messages.append({"role": role, "content": text.strip()})
+            continue
+        if not isinstance(entry, (list, tuple)) or len(entry) != 2:
+            continue
+        user_text = entry[0] if isinstance(entry[0], str) else str(entry[0] or "")
+        assistant_text = entry[1] if isinstance(entry[1], str) else str(entry[1] or "")
+        if user_text.strip():
+            messages.append({"role": "user", "content": user_text.strip()})
+        if assistant_text.strip():
+            messages.append({"role": "assistant", "content": assistant_text.strip()})
+    return messages
+
+
+def _build_messages(history: list[object], message: str) -> list[dict[str, str]]:
+    messages: list[dict[str, str]] = []
+    if SYSTEM_PROMPT:
+        messages.append({"role": "system", "content": SYSTEM_PROMPT})
+    messages.extend(_history_to_messages(history))
+    messages.append({"role": "user", "content": message.strip()})
+    return messages
+
+
+def _append_history(history: list[object], message: str, reply: str) -> list[object]:
+    return history + [
+        {"role": "user", "content": message},
+        {"role": "assistant", "content": reply},
+    ]
+
+
+def _provenance_response(message: str) -> str | None:
+    message_lower = message.strip().lower()
+    asks_provenance = any(
+        phrase in message_lower
+        for phrase in (
+            "donor",
+            "donors",
+            "donor model",
+            "donor models",
+            "base model",
+            "underlying model",
+            "what model do you use",
+            "what models do you use",
+            "deepseek",
+            "qwen",
+            "gemma",
+            "llama",
+            "phi",
+        )
+    )
+    if not asks_provenance:
+        return None
+    return (
+        f"This is {MODEL_TITLE}, a standalone AbteeX AI Labs LumynaX release for "
+        "Aotearoa New Zealand workflows. This public Space is a browser demo of that release."
+    )
+
+
+def _governance_response(message: str) -> str | None:
+    message_lower = message.strip().lower()
+
+    asks_iwi_sovereignty = (
+        "iwi" in message_lower
+        and ("data sovereignty" in message_lower or "llm" in message_lower or "language model" in message_lower)
+    )
+    asks_health_sovereignty = (
+        "health" in message_lower
+        and ("data sovereignty" in message_lower or "governance" in message_lower or "sensitive data" in message_lower)
+    )
+    asks_justice_controls = (
+        "justice sector" in message_lower
+        or ("justice" in message_lower and "ai" in message_lower)
+        or "sensitive case data" in message_lower
+    )
+
+    if asks_iwi_sovereignty:
+        return (
+            "For Iwi data sovereignty with an LLM, keep sensitive data in environments controlled by the data owner, "
+            "minimise and de-identify data before use, agree governance and access rules with Iwi decision-makers, "
+            "prevent provider training on submitted data, keep strong audit logs, require human review for high-stakes "
+            "outputs, and make deletion, retention, and purpose limits explicit from the start."
+        )
+
+    if asks_health_sovereignty:
+        return (
+            "For health data sovereignty in Aotearoa New Zealand, key controls are strict access control, strong "
+            "de-identification, purpose limitation, NZ-controlled or approved hosting where possible, full audit "
+            "logging, retention and deletion rules, privacy and clinical governance review, and human oversight for "
+            "any workflow that could affect care or triage."
+        )
+
+    if asks_justice_controls:
+        return (
+            "For justice-sector AI handling sensitive case data, use case-level access controls, data segregation, "
+            "encryption in transit and at rest, no external model training on case material, full audit trails, "
+            "mandatory human review, clear escalation and appeal paths, regular bias and security testing, and a rule "
+            "that the model supports staff but does not make binding legal or operational decisions on its own."
+        )
+
+    return None
+
+
+def _identity_response(message: str, history: list[object]) -> str | None:
+    message_lower = message.strip().lower()
+    mentions_lumynax = "lumynax" in message_lower or "lumynax infused qwen3 1.7b gguf" in message_lower
+    asks_identity = any(
+        phrase in message_lower
+        for phrase in (
+            "who are you",
+            "what are you",
+            "what is lumynax",
+            "what's lumynax",
+            "what is this model",
+            "what's this model",
+            "explain what lumynax is",
+            "explain",
+            "describe",
+            "tell me about",
+        )
+    )
+    if not asks_identity:
+        return None
+    if not mentions_lumynax and "who are you" not in message_lower and "what are you" not in message_lower:
+        return None
+
+    if "bullet" in message_lower or "three" in message_lower:
+        return '- LumynaX Infused Qwen3 1.7B GGUF is a local-first LumynaX model release from AbteeX AI Labs\\n- It is aimed at practical Aotearoa New Zealand workflows and locally relevant responses\\n- This public Space is a browser demo for that LumynaX release'
+    return 'LumynaX Infused Qwen3 1.7B GGUF is a local-first LumynaX model release from AbteeX AI Labs for Aotearoa New Zealand workflows. It is intended for practical assistant use and locally relevant text generation when appropriate. This public Space is a browser demo of the same release hosted on Hugging Face.'
+
+
+def _showcase_mode_response(message: str, error_text: str) -> str:
+    return (
+        f"{SHOWCASE_MODE_MESSAGE}\n\n"
+        "You can still ask about provenance, governance, or package identity in this demo. "
+        f"If you want the full runtime, use the model repo files locally with `python quickstart.py --interactive`. "
+        f"(Runtime detail: {error_text})"
+    )
+
+
+def _render_prompt(messages: list[dict[str, str]]) -> str:
+    if PROMPT_FORMAT == "plain_completion":
+        lines: list[str] = []
+        for entry in messages:
+            role = entry["role"]
+            content = entry["content"]
+            if role == "system":
+                lines.append(content)
+            elif role == "user":
+                lines.append(f"User: {content}")
+            else:
+                lines.append(f"Assistant: {content}")
+        lines.append("Assistant:")
+        return "\n\n".join(lines)
+
+    parts: list[str] = []
+    for entry in messages:
+        role = entry["role"]
+        content = entry["content"]
+        parts.append(f"<|im_start|>{role}\n{content}<|im_end|>\n")
+    parts.append("<|im_start|>assistant\n")
+    return "".join(parts)
+
+
+def _load_runtime() -> tuple[object, object]:
+    global _MODEL, _TOKENIZER, _MODEL_ERROR
+
+    if _MODEL is not None and _TOKENIZER is not None:
+        return _MODEL, _TOKENIZER
+    if _MODEL_ERROR is not None:
+        raise RuntimeError(_MODEL_ERROR)
+
+    with _MODEL_LOCK:
+        if _MODEL is not None and _TOKENIZER is not None:
+            return _MODEL, _TOKENIZER
+        if _MODEL_ERROR is not None:
+            raise RuntimeError(_MODEL_ERROR)
+
+        try:
+            repo_id = os.environ.get(MODEL_REPO_ENV_VAR, DEFAULT_MODEL_REPO_ID).strip() or DEFAULT_MODEL_REPO_ID
+            hf_token = next((os.environ.get(name, "").strip() for name in HF_TOKEN_ENV_VARS if os.environ.get(name, "").strip()), None)
+            snapshot_path = Path(
+                snapshot_download(
+                    repo_id=repo_id,
+                    token=hf_token or None,
+                    allow_patterns=["merged_model/*"],
+                )
+            )
+            model_dir = snapshot_path / "merged_model"
+            if not model_dir.exists():
+                raise FileNotFoundError(
+                    f"Expected merged_model/ in {snapshot_path} after downloading {repo_id}."
+                )
+
+            tokenizer = AutoTokenizer.from_pretrained(str(model_dir))
+            if tokenizer.pad_token is None:
+                tokenizer.pad_token = tokenizer.eos_token
+            model = AutoModelForCausalLM.from_pretrained(
+                str(model_dir),
+                dtype=torch.bfloat16,
+                low_cpu_mem_usage=True,
+            )
+
+            _MODEL = model
+            _TOKENIZER = tokenizer
+            return _MODEL, _TOKENIZER
+        except Exception as exc:
+            _MODEL_ERROR = f"{type(exc).__name__}: {exc}"
+            raise
+
+
+def chat(message: str, history: list[object]) -> tuple[str, list[object]]:
+    history = history or []
+    if not message.strip():
+        return "", history
+
+    try:
+        provenance_reply = _provenance_response(message)
+        if provenance_reply is not None:
+            return "", _append_history(history, message, provenance_reply)
+
+        governance_reply = _governance_response(message)
+        if governance_reply is not None:
+            return "", _append_history(history, message, governance_reply)
+
+        identity_reply = _identity_response(message, history)
+        if identity_reply is not None:
+            return "", _append_history(history, message, identity_reply)
+
+        model, tokenizer = _load_runtime()
+        messages = _build_messages(history, message)
+        if hasattr(tokenizer, "apply_chat_template") and PROMPT_FORMAT != "plain_completion":
+            encoded = tokenizer.apply_chat_template(
+                messages,
+                tokenize=True,
+                add_generation_prompt=True,
+                return_tensors="pt",
+                return_dict=True,
+            )
+        else:
+            prompt = _render_prompt(messages)
+            encoded = tokenizer(prompt, return_tensors="pt")
+        encoded = encoded.to(model.device)
+
+        with torch.inference_mode():
+            output = model.generate(
+                **encoded,
+                max_new_tokens=MAX_NEW_TOKENS,
+                do_sample=False,
+                pad_token_id=tokenizer.eos_token_id,
+            )
+
+        prompt_length = encoded["input_ids"].shape[-1]
+        generated = tokenizer.decode(output[0][prompt_length:], skip_special_tokens=True).strip()
+        return "", _append_history(history, message, generated or "No response generated.")
+    except Exception as exc:
+        error_text = f"{type(exc).__name__}: {exc}"
+        return "", _append_history(history, message, _showcase_mode_response(message, error_text))
+
+with gr.Blocks() as demo:
+    gr.Markdown(
+        f"# {MODEL_TITLE}\n\n"
+        "Public browser demo for LumynaX from AbteeX AI Labs. "
+        "This Space is backed by a private model repo on Hugging Face. "
+        "If the backing repo is GGUF-only, this browser demo stays in showcase mode and directs people to the local interactive quickstart."
+    )
+    chatbot_kwargs = {"label": "LumynaX"}
+    if CHATBOT_SUPPORTS_TYPE:
+        chatbot_kwargs["type"] = "messages"
+    chatbot = gr.Chatbot(**chatbot_kwargs)
+    gr.Markdown("Enter a prompt and press `Enter` or click `Send`.")
+    with gr.Row():
+        prompt = gr.Textbox(
+            label="Prompt",
+            placeholder="Ask LumynaX something about Aotearoa, your project, or local research.",
+            lines=4,
+            scale=8,
+        )
+        send = gr.Button("Send", variant="primary", scale=1, min_width=120)
+    gr.Examples(
+        examples=[
+            "Give a helpful welcome message for customers in Aotearoa New Zealand.",
+            'Explain in two short paragraphs what LumynaX Infused Qwen3 1.7B GGUF is and who it is for.',
+            "Write a concise summary of why local AI deployment matters for NZ teams.",
+        ],
+        inputs=prompt,
+    )
+    clear = gr.Button("Clear")
+
+    prompt.submit(chat, inputs=[prompt, chatbot], outputs=[prompt, chatbot])
+    send.click(chat, inputs=[prompt, chatbot], outputs=[prompt, chatbot])
+    clear.click(lambda: [], outputs=chatbot, queue=False)
+
+
+if __name__ == "__main__":
+    demo.launch()

hf_space/requirements.txt

@@ -0,0 +1,5 @@
+gradio>=5.0
+huggingface-hub>=0.25
+torch>=2.0
+transformers>=4.45
+safetensors>=0.4

merged_model/PACKAGE_STATE.txt

@@ -0,0 +1 @@
+LumynaX Infused Qwen3 1.7B GGUF is a GGUF-only release package. The primary runtime artifact is ../Qwen3-1.7B-Q8_0.gguf; no full Transformers checkpoint is bundled here.

ollama/Modelfile

@@ -0,0 +1,12 @@
+FROM ../Qwen3-1.7B-Q8_0.gguf
+TEMPLATE """{{ if .System }}<|im_start|>system
+{{ .System }}<|im_end|>
+{{ end }}{{ if .Prompt }}<|im_start|>user
+{{ .Prompt }}<|im_end|>
+{{ end }}<|im_start|>assistant
+"""
+SYSTEM You are LumynaX operating from the LumynaX Infused Qwen3 1.7B GGUF package identity. Be helpful, clear, and honest about provenance.
+PARAMETER temperature 0.1
+PARAMETER num_ctx 8192
+PARAMETER stop "<|im_end|>"
+PARAMETER stop "<|endoftext|>"

ollama/create_ollama_model.ps1

@@ -0,0 +1,21 @@
+param(
+    [string]$ModelName = "lumynax-infused-qwen3-17b-gguf"
+)
+
+$ErrorActionPreference = "Stop"
+Set-StrictMode -Version Latest
+
+$scriptDir = Split-Path -Parent $MyInvocation.MyCommand.Path
+$modelfilePath = Join-Path $scriptDir "Modelfile"
+
+if (-not (Get-Command ollama -ErrorAction SilentlyContinue)) {
+    throw "The `ollama` CLI is not installed. Install Ollama first."
+}
+
+& ollama create $ModelName -f $modelfilePath
+if ($LASTEXITCODE -ne 0) {
+    exit $LASTEXITCODE
+}
+
+Write-Output "Created Ollama model: $ModelName"
+Write-Output "Run it with: ollama run $ModelName"

quickstart.py

@@ -0,0 +1,449 @@
+from __future__ import annotations
+
+import argparse
+import os
+import shutil
+import subprocess
+import sys
+from pathlib import Path
+
+MODEL_TITLE = "LumynaX Infused Qwen3 1.7B GGUF"
+
+
+def _build_parser() -> argparse.ArgumentParser:
+    parser = argparse.ArgumentParser(description=f"Run a local GGUF chat for {MODEL_TITLE}.")
+    parser.add_argument(
+        "--prompt",
+        default=None,
+        help="Prompt to send to the model.",
+    )
+    parser.add_argument("--system-prompt", default="", help="Optional system prompt override.")
+    parser.add_argument(
+        "--interactive",
+        action="store_true",
+        help="Start an interactive terminal chat instead of running a single prompt.",
+    )
+    parser.add_argument("--max-new-tokens", type=int, default=192)
+    parser.add_argument("--ctx-size", type=int, default=4096)
+    parser.add_argument("--temperature", type=float, default=0.1)
+    parser.add_argument("--threads", type=int, default=max(1, os.cpu_count() or 1))
+    parser.add_argument("--llama-cli", default="", help="Optional explicit path to llama-cli.")
+    parser.add_argument(
+        "--cache-local",
+        action="store_true",
+        help="Copy the GGUF into LOCALAPPDATA before running. Useful when a runtime cannot read network paths.",
+    )
+    parser.add_argument("--reasoning", choices=("on", "off", "auto"), default="off")
+    parser.add_argument(
+        "--reasoning-format",
+        choices=("auto", "none", "deepseek", "deepseek-legacy"),
+        default="auto",
+    )
+    parser.add_argument("--reasoning-budget", type=int, default=None)
+    return parser
+
+
+def _preferred_gguf(root: Path) -> Path:
+    gguf_candidates = sorted(root.glob("*.gguf"))
+    if not gguf_candidates:
+        raise SystemExit(f"No GGUF file was found in {root}")
+    for path in gguf_candidates:
+        if "-q" in path.stem.lower():
+            return path
+    return gguf_candidates[0]
+
+
+def _local_model_path(model_path: Path, *, cache_local: bool = False) -> Path:
+    if not cache_local:
+        return model_path
+    local_app_data = Path(os.environ.get("LOCALAPPDATA", Path.home() / "AppData" / "Local"))
+    cache_dir = local_app_data / "tinyluminax" / "gguf-cache"
+    cache_dir.mkdir(parents=True, exist_ok=True)
+    cached_path = cache_dir / model_path.name
+    source_stat = model_path.stat()
+    if (
+        not cached_path.exists()
+        or cached_path.stat().st_size != source_stat.st_size
+        or cached_path.stat().st_mtime_ns < source_stat.st_mtime_ns
+    ):
+        print(f"Caching GGUF locally at {cached_path}", file=sys.stderr)
+        shutil.copy2(model_path, cached_path)
+    return cached_path
+
+
+def _discover_llama_cli(explicit_path: str) -> Path | None:
+    candidates: list[Path] = []
+    if explicit_path.strip():
+        candidates.append(Path(explicit_path.strip()))
+    for env_var in ("LLAMA_CPP_CLI", "LLAMA_CLI_PATH"):
+        raw_value = os.environ.get(env_var, "").strip()
+        if raw_value:
+            candidates.append(Path(raw_value))
+    for binary_name in ("llama-cli", "llama-cli.exe"):
+        resolved = shutil.which(binary_name)
+        if resolved:
+            candidates.append(Path(resolved))
+    for candidate in candidates:
+        if candidate.exists():
+            return candidate
+    return None
+
+
+def _extract_text(response: dict[str, object]) -> str:
+    choices = response.get("choices", [])
+    if not isinstance(choices, list) or not choices:
+        raise RuntimeError("The runtime returned no choices.")
+    first_choice = choices[0]
+    if isinstance(first_choice, dict):
+        message = first_choice.get("message")
+        if isinstance(message, dict):
+            content = message.get("content")
+            if content not in (None, ""):
+                return str(content).strip()
+        text = first_choice.get("text")
+        if text not in (None, ""):
+            return str(text).strip()
+    raise RuntimeError("The runtime returned an unsupported response payload.")
+
+
+def _run_llama_cpp_python(
+    *,
+    model_path: Path,
+    system_prompt: str,
+    user_prompt: str,
+    max_new_tokens: int,
+    ctx_size: int,
+    temperature: float,
+    threads: int,
+) -> str:
+    from llama_cpp import Llama
+
+    llm = Llama(
+        model_path=str(model_path),
+        n_ctx=ctx_size,
+        n_threads=threads,
+        n_gpu_layers=0,
+        chat_format="chat_template.default",
+        verbose=False,
+    )
+    response = llm.create_chat_completion(
+        messages=[
+            {"role": "system", "content": system_prompt},
+            {"role": "user", "content": user_prompt},
+        ],
+        max_tokens=max_new_tokens,
+        temperature=temperature,
+    )
+    return _extract_text(response)
+
+
+def _run_llama_cli(
+    *,
+    llama_cli_path: Path,
+    model_path: Path,
+    system_prompt: str,
+    user_prompt: str,
+    max_new_tokens: int,
+    ctx_size: int,
+    temperature: float,
+    threads: int,
+    reasoning: str,
+    reasoning_format: str,
+    reasoning_budget: int | None,
+) -> None:
+    command = [
+        str(llama_cli_path),
+        "-m",
+        str(model_path),
+        "-sys",
+        system_prompt,
+        "-p",
+        user_prompt,
+        "-cnv",
+        "-st",
+        "-n",
+        str(max_new_tokens),
+        "-c",
+        str(ctx_size),
+        "--reasoning",
+        reasoning,
+        "--temp",
+        str(temperature),
+        "--threads",
+        str(threads),
+        "--no-display-prompt",
+    ]
+    if reasoning_format != "auto":
+        command.extend(["--reasoning-format", reasoning_format])
+    if reasoning_budget is not None:
+        command.extend(["--reasoning-budget", str(reasoning_budget)])
+    completed = subprocess.run(
+        command,
+        check=False,
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+    )
+    if completed.returncode != 0:
+        detail = completed.stderr.strip() or completed.stdout.strip() or "llama-cli failed"
+        raise SystemExit(detail)
+    stdout = completed.stdout.strip()
+    if stdout:
+        print(stdout)
+
+
+def _print_interactive_banner() -> None:
+    print("LumynaX interactive terminal chat")
+    print("Type /reset to clear the conversation, or /quit to exit.")
+
+
+def _run_interactive_llama_cpp_python(
+    *,
+    model_path: Path,
+    system_prompt: str,
+    max_new_tokens: int,
+    ctx_size: int,
+    temperature: float,
+    threads: int,
+    opening_prompt: str | None = None,
+    reasoning: str = "off",
+    reasoning_format: str = "auto",
+    reasoning_budget: int | None = None,
+) -> None:
+    from llama_cpp import Llama
+
+    llm = Llama(
+        model_path=str(model_path),
+        n_ctx=ctx_size,
+        n_threads=threads,
+        n_gpu_layers=0,
+        chat_format="chat_template.default",
+        verbose=False,
+    )
+    transcript: list[tuple[str, str]] = []
+    _print_interactive_banner()
+
+    pending_prompt = opening_prompt.strip() if opening_prompt and opening_prompt.strip() else None
+    while True:
+        try:
+            if pending_prompt is None:
+                user_prompt = input("You> ").strip()
+            else:
+                user_prompt = pending_prompt
+                print(f"You> {user_prompt}")
+                pending_prompt = None
+        except (EOFError, KeyboardInterrupt):
+            print("\nExiting LumynaX chat.")
+            return
+        if not user_prompt:
+            continue
+        lowered_prompt = user_prompt.lower()
+        if lowered_prompt in ('/quit', '/exit'):
+            print("Exiting LumynaX chat.")
+            return
+        if lowered_prompt == "/reset":
+            transcript.clear()
+            print("Conversation reset.")
+            continue
+        messages: list[dict[str, str]] = [{"role": "system", "content": system_prompt}]
+        for transcript_user_prompt, transcript_assistant_response in transcript:
+            messages.append({"role": "user", "content": transcript_user_prompt})
+            messages.append({"role": "assistant", "content": transcript_assistant_response})
+        messages.append({"role": "user", "content": user_prompt})
+        response = llm.create_chat_completion(
+            messages=messages,
+            max_tokens=max_new_tokens,
+            temperature=temperature,
+        )
+        assistant_text = _extract_text(response)
+        print(f"LumynaX> {assistant_text}")
+        transcript.append((user_prompt, assistant_text))
+
+
+def _run_interactive_llama_cli(
+    *,
+    llama_cli_path: Path,
+    model_path: Path,
+    system_prompt: str,
+    max_new_tokens: int,
+    ctx_size: int,
+    temperature: float,
+    threads: int,
+    opening_prompt: str | None = None,
+    reasoning: str = "off",
+    reasoning_format: str = "auto",
+    reasoning_budget: int | None = None,
+) -> None:
+    print("LumynaX interactive terminal chat")
+    print("Interactive mode already uses llama-cli directly. Use Ctrl+C to exit.")
+    command = [
+        str(llama_cli_path),
+        "-m",
+        str(model_path),
+        "-sys",
+        system_prompt,
+        "-cnv",
+        "-n",
+        str(max_new_tokens),
+        "-c",
+        str(ctx_size),
+        "--reasoning",
+        reasoning,
+        "--temp",
+        str(temperature),
+        "--threads",
+        str(threads),
+        "--simple-io",
+    ]
+    if reasoning_format != "auto":
+        command.extend(["--reasoning-format", reasoning_format])
+    if reasoning_budget is not None:
+        command.extend(["--reasoning-budget", str(reasoning_budget)])
+    if opening_prompt and opening_prompt.strip():
+        command.extend(["-p", opening_prompt.strip()])
+    completed = subprocess.run(command, check=False)
+    if completed.returncode != 0:
+        raise SystemExit(completed.returncode)
+
+
+def main() -> None:
+    args = _build_parser().parse_args()
+    root = Path(__file__).resolve().parent
+    source_model_path = _preferred_gguf(root)
+    if hasattr(sys.stdout, "reconfigure"):
+        sys.stdout.reconfigure(encoding="utf-8")
+
+    single_prompt = (args.prompt or "Say hello in one short sentence.").strip()
+    system_prompt = args.system_prompt.strip() or (
+        f"You are LumynaX operating from the {MODEL_TITLE} package identity. "
+        "Be helpful, clear, and honest about provenance."
+    )
+    explicit_cli_requested = bool(
+        args.llama_cli.strip()
+        or os.environ.get("LLAMA_CPP_CLI", "").strip()
+        or os.environ.get("LLAMA_CLI_PATH", "").strip()
+    )
+    if args.interactive:
+        llama_cli_path = _discover_llama_cli(args.llama_cli)
+        if explicit_cli_requested:
+            if llama_cli_path is None:
+                raise SystemExit(
+                    "A llama-cli override was requested, but no usable llama-cli binary was found.",
+                )
+            _run_interactive_llama_cli(
+                llama_cli_path=llama_cli_path,
+                model_path=_local_model_path(source_model_path, cache_local=args.cache_local),
+                system_prompt=system_prompt,
+                opening_prompt=args.prompt,
+                max_new_tokens=args.max_new_tokens,
+                ctx_size=args.ctx_size,
+                temperature=args.temperature,
+                threads=args.threads,
+                reasoning=args.reasoning,
+                reasoning_format=args.reasoning_format,
+                reasoning_budget=args.reasoning_budget,
+            )
+            return
+        model_path = _local_model_path(source_model_path, cache_local=args.cache_local)
+        try:
+            _run_interactive_llama_cpp_python(
+                model_path=model_path,
+                system_prompt=system_prompt,
+                opening_prompt=args.prompt,
+                max_new_tokens=args.max_new_tokens,
+                ctx_size=args.ctx_size,
+                temperature=args.temperature,
+                threads=args.threads,
+                reasoning=args.reasoning,
+                reasoning_format=args.reasoning_format,
+                reasoning_budget=args.reasoning_budget,
+            )
+            return
+        except Exception as exc:  # noqa: BLE001
+            if llama_cli_path is None:
+                raise SystemExit(
+                    "llama-cpp-python could not load this GGUF package. "
+                    "Install or point LLAMA_CPP_CLI at llama-cli to use the built-in fallback. "
+                    f"Original error: {exc}",
+                ) from exc
+            print(
+                f"llama-cpp-python failed; falling back to llama-cli at {llama_cli_path}",
+                file=sys.stderr,
+            )
+            _run_interactive_llama_cli(
+                llama_cli_path=llama_cli_path,
+                model_path=model_path,
+                system_prompt=system_prompt,
+                opening_prompt=args.prompt,
+                max_new_tokens=args.max_new_tokens,
+                ctx_size=args.ctx_size,
+                temperature=args.temperature,
+                threads=args.threads,
+                reasoning=args.reasoning,
+                reasoning_format=args.reasoning_format,
+                reasoning_budget=args.reasoning_budget,
+            )
+            return
+    if explicit_cli_requested:
+        llama_cli_path = _discover_llama_cli(args.llama_cli)
+        if llama_cli_path is None:
+            raise SystemExit(
+                "A llama-cli override was requested, but no usable llama-cli binary was found.",
+            )
+        _run_llama_cli(
+            llama_cli_path=llama_cli_path,
+            model_path=_local_model_path(source_model_path, cache_local=args.cache_local),
+            system_prompt=system_prompt,
+            user_prompt=single_prompt,
+            max_new_tokens=args.max_new_tokens,
+            ctx_size=args.ctx_size,
+            temperature=args.temperature,
+            threads=args.threads,
+            reasoning=args.reasoning,
+            reasoning_format=args.reasoning_format,
+            reasoning_budget=args.reasoning_budget,
+        )
+        return
+    model_path = _local_model_path(source_model_path, cache_local=args.cache_local)
+    try:
+        print(
+            _run_llama_cpp_python(
+                model_path=model_path,
+                system_prompt=system_prompt,
+                user_prompt=single_prompt,
+                max_new_tokens=args.max_new_tokens,
+                ctx_size=args.ctx_size,
+                temperature=args.temperature,
+                threads=args.threads,
+            ),
+        )
+        return
+    except Exception as exc:  # noqa: BLE001
+        llama_cli_path = _discover_llama_cli(args.llama_cli)
+        if llama_cli_path is None:
+            raise SystemExit(
+                "llama-cpp-python could not load this GGUF package. "
+                "Install or point LLAMA_CPP_CLI at llama-cli to use the built-in fallback. "
+                f"Original error: {exc}",
+            ) from exc
+        print(
+            f"llama-cpp-python failed; falling back to llama-cli at {llama_cli_path}",
+            file=sys.stderr,
+        )
+        _run_llama_cli(
+            llama_cli_path=llama_cli_path,
+            model_path=model_path,
+            system_prompt=system_prompt,
+            user_prompt=single_prompt,
+            max_new_tokens=args.max_new_tokens,
+            ctx_size=args.ctx_size,
+            temperature=args.temperature,
+            threads=args.threads,
+            reasoning=args.reasoning,
+            reasoning_format=args.reasoning_format,
+            reasoning_budget=args.reasoning_budget,
+        )
+
+
+if __name__ == "__main__":
+    main()

release_export_manifest.json

@@ -0,0 +1,88 @@
+{
+  "artifacts": {
+    "checksums": "checksums.sha256",
+    "gguf": "Qwen3-1.7B-Q8_0.gguf",
+    "gitattributes": ".gitattributes",
+    "hf_space_app": "hf_space/app.py",
+    "hf_space_dir": "hf_space",
+    "hf_space_readme": "hf_space/README.md",
+    "hf_space_requirements": "hf_space/requirements.txt",
+    "license": "LICENSE.txt",
+    "merged_model": "merged_model",
+    "mmproj": null,
+    "ollama_create_script": "ollama/create_ollama_model.ps1",
+    "ollama_modelfile": "ollama/Modelfile",
+    "package_state": "merged_model/PACKAGE_STATE.txt",
+    "quantized_gguf": "Qwen3-1.7B-Q8_0.gguf",
+    "quickstart": "quickstart.py",
+    "readme": "README.md",
+    "requirements": "requirements.txt",
+    "training_summary": "artifacts/release_training_summary.json",
+    "upload_notes": "UPLOAD_TO_HF.md",
+    "version": "VERSION.txt"
+  },
+  "capabilities": {
+    "reasoning_enabled": false,
+    "supported_modalities": [
+      "text"
+    ]
+  },
+  "delivery": "standalone_prebuilt_gguf_release",
+  "distribution": {
+    "hf_space": {
+      "app": "hf_space/app.py",
+      "default_model_repo_id": "AbteeXAILab/lumynax-infused-qwen3-17b-gguf",
+      "directory": "hf_space",
+      "model_repo_env_var": "LUMYNAX_MODEL_REPO_ID",
+      "readme": "hf_space/README.md",
+      "requirements": "hf_space/requirements.txt",
+      "status": "browser_showcase_for_gguf_only_release"
+    },
+    "ollama": {
+      "create_script": "ollama/create_ollama_model.ps1",
+      "mmproj": null,
+      "modelfile": "ollama/Modelfile",
+      "preferred_gguf": "Qwen3-1.7B-Q8_0.gguf",
+      "recommended_model_name": "lumynax-infused-qwen3-17b-gguf",
+      "status": "ready_for_local_ollama_create"
+    }
+  },
+  "family": null,
+  "generated_at": "2026-05-10T12:59:23.381925+00:00",
+  "license": {
+    "id": "apache-2.0",
+    "link": "https://huggingface.co/Qwen/Qwen3-1.7B-GGUF",
+    "name": null,
+    "weights_subject_to_upstream_license": true
+  },
+  "manifest_version": 2,
+  "model_title": "LumynaX Infused Qwen3 1.7B GGUF",
+  "package_state": "prebuilt_gguf_release",
+  "public_identity": {
+    "model_name": "LumynaX",
+    "organization": "AbteeX AI Labs",
+    "region": "Aotearoa New Zealand"
+  },
+  "release_version": "v1",
+  "runtime": {
+    "delivery_mode": "standalone_prebuilt_gguf",
+    "preferred_backend": "llama_cpp",
+    "prompt_format": "chatml",
+    "quickstart_command": "python quickstart.py --interactive",
+    "system_prompt": "You are LumynaX operating from the LumynaX Infused Qwen3 1.7B GGUF package identity. Be helpful, clear, and honest about provenance."
+  },
+  "source_gguf": {
+    "filename": "Qwen3-1.7B-Q8_0.gguf",
+    "mmproj_filename": null,
+    "packaged_filename": "Qwen3-1.7B-Q8_0.gguf",
+    "packaged_mmproj_filename": null,
+    "quantization": "Q8_0",
+    "repo_id": "Qwen/Qwen3-1.7B-GGUF"
+  },
+  "upstream_model": {
+    "kind": "official_base_weights",
+    "lumynax_weight_adaptation_applied": false,
+    "provider": "Hugging Face",
+    "repo_id": "Qwen/Qwen3-1.7B"
+  }
+}

requirements.txt

@@ -0,0 +1 @@
+llama-cpp-python>=0.3.18