qriton 0.5.0 → 0.5.1

package/docs/hlm.md

@@ -17,7 +17,7 @@ The editable HLM5 1B models support single-token answers at exact prompts, with
 | [HLM5 3B Base](https://huggingface.co/qriton/hlm5-3b-base) | Pinned sharded-safetensors import/download and FP32 completion | Inference only. Loader coverage does not establish full-model quality or editing support. |
 | [HLM5 certificate package](https://github.com/qriton/hlm5) | Upstream certificate/dose and editable-memory APIs | Separately installed; model weights remain frozen. |
 | `.hlm` programs | Text-addressed `memorize`, `correct`, `forget` through `/hlm`; numerical basin edits through `/energy` | Two explicitly separate adapters. |
-| [Energy Language](https://github.com/qriton/energy-lang) / `qriton-hlm` | Existing HLM3 checkpoint workbench | Degree-3 polynomial Hopfield checkpoints; sampled structural checks and optional text evaluation. |
+| [Energy Language](https://github.com/qriton/energy-lang) / `qriton-hlm` | Existing HLM3 checkpoint workbench (`qriton-hlm` 0.13.0 or later) | Degree-3 polynomial Hopfield checkpoints; Energy Language's decision check, sampled structural checks and optional text evaluation. The editing method is validated on synthetic landscapes only: on trained HLM3 checkpoints no measured edit was local enough to keep, and with an empty protected set the decision check accepted writes that changed up to 14.7% of held-out predictions ([measurements](https://github.com/qriton/energy-lang/blob/master/RELEASE_NOTES.md#trained-weights)). Treat edits to trained weights as experiments and check held-out text. |
 | `@qriton/hopfield-anomaly` | Existing memory recall plus `memory probe` diagnostics | Embedding novelty/composition and plain-vs-Hopfield retrieval comparison. |
 | Qriton's gesture, activity, audio, CIFAR-10 and ModelNet10 models | Pinned downloads and classification of one local input file | Fixed label sets with family-specific runtime setup; no live sensor capture. |
 | [HF certificate explorer](https://huggingface.co/spaces/qriton/hlm5-energy-language) | Reference demonstration | The local integration runs in its own Python process; private prompts are not sent to the Space. |

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "qriton",
-  "version": "0.5.0",
+  "version": "0.5.1",
   "description": "Qriton CLI is a terminal coding agent that runs on infrastructure you control. Local models or any major model API, with a memory record, HLM, and Energy Language.",
   "type": "module",
   "bin": {

package/release/contract.json

@@ -1,5 +1,5 @@
 {
-  "version": "0.5.0",
+  "version": "0.5.1",
   "supportedCodingModel": "local Ollama, plus xAI, Anthropic, OpenAI, and OpenAI-compatible APIs",
   "codingEvidence": "In one repeat of 30 small held-out tasks on a 16 GB laptop, qwen3.5:4b produced 27 passing artifacts; 24 also had successful process exits. Four runs timed out, including three with passing artifacts. openai/gpt-5-mini passed two small editing fixtures. These checks do not establish representative coding quality or superiority over other tools. HLM and Energy Language are separate local workbenches.",
   "localModels": "Supported through Ollama. A fresh installation defaults to qwen3.5:4b; a bare qriton command uses the configured model. Local inference requires sufficient available memory and compute. The light preset is not a memory cap or a guarantee of desktop responsiveness.",