fix: classify szl-nemo as an Ollama recipe, not NeMo weights

Renames the standard Ollama Modelfile to stop false NeMo library inference, adds a registry/layer digest manifest, and clarifies the recipe/upstream-license boundary. No weights or measured benchmarks are added.

BASE_MODEL_MANIFEST.json

@@ -0,0 +1,37 @@
+{
+  "config": {
+    "digest": "sha256:1c309ae5e93ede03f48f4bdbfd1b50d1780f2c3bcdc82f4569e20f0c33878db9",
+    "mediaType": "application/vnd.docker.container.image.v1+json",
+    "size": 499
+  },
+  "consumer_policy": "The Ollama tag is mutable. Verify the fetched manifest and layer digests against this observation or deliberately update this file before claiming a reproducible deployment.",
+  "layers": [
+    {
+      "digest": "sha256:527db2cf6c705d8fabb95693d038d9c06b4a2b0b8b0a4bbdbd01212d37242970",
+      "mediaType": "application/vnd.ollama.image.model",
+      "size": 2837586496
+    },
+    {
+      "digest": "sha256:355e036064fa9b3a96ce0cdbb69abe54f5c7ce0b6aa2c7d5f8ec8580b011c20e",
+      "mediaType": "application/vnd.ollama.image.license",
+      "size": 10124
+    },
+    {
+      "digest": "sha256:12e88b2a8727339b5a4a8b3e2d0d637ac1c61085b1072e77865f0c25d6e468eb",
+      "mediaType": "application/vnd.ollama.image.params",
+      "size": 28
+    }
+  ],
+  "model": "nemotron-3-nano:4b",
+  "observed_at": "2026-07-15T10:45:00Z",
+  "registry_manifest_sha256": "sha256:6cc467f054393a55e98a74098abde0c762ffb6d1d8cd64becf30458f38886197",
+  "registry_manifest_url": "https://registry.ollama.ai/v2/library/nemotron-3-nano/manifests/4b",
+  "schema": "szl.ollama-base-model-manifest/v1",
+  "tag_is_mutable": true,
+  "upstream_huggingface": {
+    "bf16": "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16",
+    "gguf": "nvidia/NVIDIA-Nemotron-3-Nano-4B-GGUF",
+    "license": "nvidia-nemotron-open-model-license",
+    "license_url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16/blob/main/LICENSE"
+  }
+}

Modelfile.nemo → Modelfile

@@ -5,8 +5,10 @@
 # The base model's own chat template is inherited unchanged. Benchmarks: none
 # measured yet — treat quality as UNKNOWN until measured on SZL hardware.
 #
+# Upstream registry manifest observed 2026-07-15: sha256:6cc467f054393a55e98a74098abde0c762ffb6d1d8cd64becf30458f38886197
+# The tag is mutable; verify BASE_MODEL_MANIFEST.json after pulling.
 # Create on the tower after `ollama pull nemotron-3-nano:4b`:
-#   ollama create szl-nemo -f Modelfile.nemo
+#   ollama create szl-nemo -f Modelfile
 
 FROM nemotron-3-nano:4b
 

README.md

@@ -1,26 +1,38 @@
 ---
 license: apache-2.0
+language:
+  - en
 pipeline_tag: text-generation
 tags:
   - nemotron
   - ollama
+  - recipe-only
+  - no-weights
   - sovereign-ai
   - governed-ai
   - szl-holdings
 ---
 
-# SZL-Nemo — doctrine-wrapped NVIDIA Nemotron 3 Nano, served on SZL metal
+# SZL-Nemo - Ollama prompt recipe for NVIDIA Nemotron 3 Nano 4B
 
 **Recipe tier — honest labels first:**
 
 | Claim | Status |
 | --- | --- |
 | What this repo contains | An Ollama `Modelfile` recipe + doctrine system prompt. **No weights are republished here.** |
-| Whose weights | NVIDIA's open **Nemotron 3 Nano 4B** (`nemotron-3-nano:4b` via Ollama — 2.8 GB, 256K context, REPORTED from ollama.com 2026-07-11). |
+| Whose weights | NVIDIA Nemotron 3 Nano 4B via Ollama tag `nemotron-3-nano:4b`; upstream weights are not stored here. |
 | Did SZL fine-tune them | **No.** SZL-Nemo is a SYSTEM-prompt wrapper, not an SZL fine-tune. It says so if you ask it. |
 | Benchmarks | **None measured** on SZL hardware yet — quality is UNKNOWN until measured. |
 | Serving status | **Prepared · wired · not yet serving** — Alloy's sovereign fleet has a live third slot (`tower·nemo`, model `szl-nemo`); it serves once the tower pulls and creates the model (tower was offline at authoring time, MEASURED 530). |
 
+## Base artifact lock
+
+`BASE_MODEL_MANIFEST.json` records the Ollama `4b` registry-manifest SHA-256
+`6cc467f054393a55e98a74098abde0c762ffb6d1d8cd64becf30458f38886197`,
+the config digest, all layer digests/sizes, the observation time, the official
+Hugging Face upstream IDs, and the NVIDIA license link. The tag is mutable; a
+deployment is reproducible only when it verifies or deliberately updates that manifest.
+
 ## Why it exists
 
 The LangChain × NVIDIA **NemoClaw Deep Agents blueprint** (July 2026) pairs an
@@ -37,8 +49,8 @@ On a machine running Ollama:
 
 ```bash
 ollama pull nemotron-3-nano:4b
-curl -L -o Modelfile.nemo https://huggingface.co/SZLHOLDINGS/szl-nemo/raw/main/Modelfile.nemo
-ollama create szl-nemo -f Modelfile.nemo
+curl -L -o Modelfile https://huggingface.co/SZLHOLDINGS/szl-nemo/raw/main/Modelfile
+ollama create szl-nemo -f Modelfile
 ollama run szl-nemo "Who are you, and did SZL train your weights?"
 ```
 
@@ -48,7 +60,8 @@ Full one-command-per-step tower runbook: [`szl-forge/RUNBOOK-NEMO.md`](https://g
 
 SZL-Nemo answers under SZL's honesty doctrine: claims are labeled MEASURED,
 REPORTED, or UNKNOWN, and an honest UNKNOWN stands rather than an invented
-answer. The base model's own license applies to the weights (see
-ollama.com/library/nemotron-3-nano); this recipe is Apache-2.0.
+answer. This repository's Modelfile and prompt text are Apache-2.0. The upstream weights
+are not redistributed here and remain under the
+[NVIDIA Nemotron Open Model License](https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16/blob/main/LICENSE).
 
 Built and maintained by [SZL Holdings](https://a-11-oy.com).