Correct research claims and current project status
Browse filesDocumentation cleanup dated 2026-10-03: scope historical evidence, correct lineage and dependencies, and remove obsolete release promotions. Model weights, configuration, licenses, access settings, and runtime behavior are unchanged.
README.md
CHANGED
|
@@ -1,116 +1,40 @@
|
|
| 1 |
-
> **Status 2026-09-06 β J.E. Herizon LLC / Qovaryx**
|
| 2 |
-
> Desktop downloads are **paused**. There is no public installer today.
|
| 3 |
-
> Do not treat this card as a live product, a fully offline runtime, or a published win rate.
|
| 4 |
-
> Company file: https://jehorizon.com/llms.txt Β· product: https://qovaryx.jehorizon.com
|
| 5 |
-
> This repo is research/reproducibility only.
|
| 6 |
-
|
| 7 |
---
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
colorTo: purple
|
| 12 |
-
sdk: static
|
| 13 |
-
pinned: false
|
| 14 |
---
|
| 15 |
|
| 16 |
# Qovaryx
|
| 17 |
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
Random-init scratch substrates. Audited compact specialists. Local-first deployment. No borrowed foundation weights, no closed APIs in the inference path. A research project building toward a frontier-grade decision system that runs on hardware a single person can own.
|
| 21 |
-
|
| 22 |
-
---
|
| 23 |
-
|
| 24 |
-
## π 2026-07-20 β Qovaryx 3.0 milestone shipped
|
| 25 |
-
|
| 26 |
-
A 5-year full-pipeline backtest (2021β2026, $500 starting capital, real + synthetic option corpus, all wired safety gates engaged) on the four public risk tiers:
|
| 27 |
-
|
| 28 |
-
| Tier | Total return (5 yrs) | Positive every year? |
|
| 29 |
-
|---|---|---|
|
| 30 |
-
| Conservative | +523% | β
|
|
| 31 |
-
| Standard | +1,793% | β
|
|
| 32 |
-
| Aggressive | +22,785% | β
|
|
| 33 |
-
| Extreme | +47,326% | β
|
|
| 34 |
-
|
| 35 |
-
Every profile positive every year β including 2022 (bear), 2020-style vol shocks in the corpus, and 2025 chop. Aggressive and Extreme require typed operator acknowledgement and enforce mandatory milestone profit-pulling. Universal circuit breakers active on every tier: daily loss halt, correlation cap, drawdown ack, slippage governor. Numbers describe the backtest; live paper begins Q3 2026.
|
| 36 |
-
|
| 37 |
-
**What actually changed in 3.0:**
|
| 38 |
-
|
| 39 |
-
- **Four-tier execution-authority taxonomy on every specialist head.** `REAL_VALIDATED_SCOPED` heads can influence a named consumer strategy. `SYNTH_VALIDATED_SHADOW` runs alongside live decisions, logged, never mutating. `VERIFIED_FORECAST_CONTEXT` heads are feature-input only. `OFFLINE_FORECAST_ADVISORY` is research-only. Nothing overrides an order without a signed headβconsumerβmutation contract. Full write-up: [Head Architecture Advances](https://github.com/thron-j/qovaryx-ai-research/blob/main/research/head_architecture_advances_20260720.md).
|
| 40 |
-
- **A proprietary data moat.** Multi-decade daily equity coverage (thousands of tickers), multi-year real options-chain snapshots for the highly liquid names, app-faithful multi-symbol option-PnL labels, calibrated synthetic bridges with label-bridging validation, intraday minute-bar snapshots, and a privacy-preserving decision-outcome corpus. Full write-up: [The Data Moat](https://github.com/thron-j/qovaryx-ai-research/blob/main/research/the_data_moat_20260720.md).
|
| 41 |
-
- **BYO-AI chat with 7 providers.** OpenAI, Anthropic, Google, xAI, OpenRouter, Ollama (local), and custom OpenAI-compatible. Keys encrypted via OS keychain. Per-context opt-in for what the AI can see. **BYO-AI is for chart analysis and position review β it is not on the order-execution path.** The trending "ChatGPT / Claude will trade for you" wrappers are not consistently profitable when benchmarked out-of-sample; general-purpose LLMs weren't trained to price options or size risk. Different job, different tool.
|
| 42 |
-
- **CPU-only inference. No GPU, no cloud round-trip.** The full trading cluster runs on any x86-64 from 2018 or newer with 8 GB RAM.
|
| 43 |
-
- **13 shipping options strategies** dispatched under the same risk gate: iron condors, credit spreads (both directional variants), PEAD drift (30-DTE and 45-DTE), breakouts, VPA long calls/puts, tactical allocator, VIX-crush credit spread, tail hedges.
|
| 44 |
-
- **Deep cleanup.** 22 previously-dormant safety features wired end-to-end and ~1,750 lines of dead code removed. 80/80 focused test suite. Codex ship audit: APPROVE / HIGH CONFIDENCE.
|
| 45 |
-
|
| 46 |
-
Implementation specifics β exact training recipes, routing heuristics, sizing math, gate thresholds, corpus labeling functions β are intentionally withheld. The framings publish; the recipes do not.
|
| 47 |
-
|
| 48 |
-
*Trading options involves substantial risk of loss. Backtest results are historical simulation, not a forecast. Not financial advice.*
|
| 49 |
-
|
| 50 |
-
---
|
| 51 |
-
|
| 52 |
-
## What lives here
|
| 53 |
|
| 54 |
-
|
| 55 |
|
| 56 |
-
##
|
| 57 |
|
| 58 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 59 |
|
| 60 |
-
|
| 61 |
-
- **[qovaryx-350m-scratch-base](https://huggingface.co/tjarvis91/qovaryx-350m-scratch-base)** β 350M parameter mid-size substrate. Fits on a 12 GB card. The deployment-friendly size.
|
| 62 |
-
- **[qovaryx-50m-scratch-base](https://huggingface.co/tjarvis91/qovaryx-50m-scratch-base)** β 50M parameter proxy substrate. Fits anywhere. Used for pipeline validation before the larger sizes commit to training time.
|
| 63 |
-
- **[qovaryx-3b-scratch-base](https://huggingface.co/tjarvis91/qovaryx-3b-scratch-base)** β 3B parameter scratch-base extension of the same lineage.
|
| 64 |
|
| 65 |
-
|
| 66 |
|
| 67 |
-
|
| 68 |
|
| 69 |
-
The
|
| 70 |
|
| 71 |
-
|
| 72 |
-
- **[vfaix-vpa-options-trader](https://huggingface.co/tjarvis91/vfaix-vpa-options-trader)** β VPA-flavored legacy variant.
|
| 73 |
|
| 74 |
-
##
|
| 75 |
-
|
| 76 |
-
The first **operational** artifact in the lineage β a sovereign-trained compact specialist intent router, audited and deployed live on free Hugging Face CPU infrastructure, serving a real community Discord. Read more in the [public devlog entry](https://github.com/thron-j/qovaryx-ai-research/blob/main/research/sovereign_compact_cognition_deployed.md) on the deployment.
|
| 77 |
-
|
| 78 |
-
You can interact with it directly through the project's Discord community β `/qchat ask <question>` β and the model will route, retrieve, and respond. No closed model anywhere in the loop.
|
| 79 |
-
|
| 80 |
-
---
|
| 81 |
-
|
| 82 |
-
## Discipline
|
| 83 |
-
|
| 84 |
-
This project is built under six explicit commitments:
|
| 85 |
-
|
| 86 |
-
- **Local sovereign AI** β the deployed system runs on hardware the operator owns. No silent dependency on a remote API.
|
| 87 |
-
- **Compact cognition** β small is not the apology. Density per stored bit, routing per active parameter, structure per training row.
|
| 88 |
-
- **Verifier-governed inference** β the wrapper around the model is part of the architecture, not a deployment afterthought.
|
| 89 |
-
- **Adaptive compute** β the model thinks harder on the inputs that deserve it and cheaper on the inputs that do not.
|
| 90 |
-
- **Execution-aware intelligence** β gross alpha is not a result. Numbers must survive realistic costs.
|
| 91 |
-
- **Intelligence per watt** β the only honest unit of progress under a power-, memory-, and bit-constrained budget.
|
| 92 |
-
|
| 93 |
-
These are first-class architectural decisions made at the very first commit. They are not deployment optimizations.
|
| 94 |
-
|
| 95 |
-
---
|
| 96 |
-
|
| 97 |
-
## Where to go next
|
| 98 |
-
|
| 99 |
-
| Surface | Link |
|
| 100 |
-
|---|---|
|
| 101 |
-
| Website | [qovaryx.jehorizon.com](https://qovaryx.jehorizon.com/) |
|
| 102 |
-
| π **Public research devlog** | [github.com/thron-j/qovaryx-ai-research](https://github.com/thron-j/qovaryx-ai-research) |
|
| 103 |
-
| π« **Community Discord** (builders training their own trading/finance models, no signals) | [discord.gg/PtuHZDv5ju](https://discord.gg/PtuHZDv5ju) |
|
| 104 |
-
| π§ͺ **Deployed Q-Chat router** | type `/qchat ask <question>` in the Discord |
|
| 105 |
-
| π€ **Founder profile** | [huggingface.co/tjarvis91](https://huggingface.co/tjarvis91) |
|
| 106 |
-
| π¦ **VFAi-X desktop releases** | [github.com/thron-j/vfai-x-releases](https://github.com/thron-j/vfai-x-releases) |
|
| 107 |
-
| β **Support the next training run** | [ko-fi.com/tjarvis91](https://ko-fi.com/tjarvis91) |
|
| 108 |
-
| π§ **Contact** | thomasjarvis2026@gmail.com |
|
| 109 |
-
|
| 110 |
-
Implementation specifics β exact training recipes, routing heuristics, gate thresholds, curriculum mixing ratios, verifier internals β are intentionally withheld. The framings are publishable; the recipes are not.
|
| 111 |
-
|
| 112 |
-
---
|
| 113 |
|
| 114 |
-
|
|
|
|
|
|
|
|
|
|
| 115 |
|
| 116 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
+
tags:
|
| 3 |
+
- qovaryx
|
| 4 |
+
- research
|
|
|
|
|
|
|
|
|
|
| 5 |
---
|
| 6 |
|
| 7 |
# Qovaryx
|
| 8 |
|
| 9 |
+
Qovaryx is a research project exploring compact models and local inference for focused tasks, including document processing and market-context classification. Work originated in the Bleeding Edge research environment.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
|
| 11 |
+
**Status β October 3, 2026:** Qovaryx is on the back burner. Desktop downloads are paused; there is no current public installer, purchase, or trial. These repositories remain as research references. Published weights are separate from desktop product availability.
|
| 12 |
|
| 13 |
+
## Repository guide
|
| 14 |
|
| 15 |
+
| Family | What the published artifacts represent |
|
| 16 |
+
| --- | --- |
|
| 17 |
+
| [50M](https://huggingface.co/tjarvis91/qovaryx-50m-scratch-base), [350M](https://huggingface.co/tjarvis91/qovaryx-350m-scratch-base), [1B](https://huggingface.co/tjarvis91/qovaryx-1b-scratch-base), [3B](https://huggingface.co/tjarvis91/qovaryx-3b-scratch-base) scratch bases | Random-initialized starting points that require training. Size names are family labels; inspect each configuration and loader. |
|
| 18 |
+
| [Q-RAG](https://huggingface.co/tjarvis91/Q-RAG-50M-Sovereign) and the Q-* specialist cards | Research checkpoints for narrow text tasks. Small internal evaluations have limited scope. |
|
| 19 |
+
| Q-Macro, Q-Pattern, Q-Regime, Q-Volatility, Q-Veto | Experimental market-context classifiers. Classification labels do not establish trading effectiveness. |
|
| 20 |
+
| [VFAi-X](https://huggingface.co/tjarvis91/vfai-x-3.5-9b-options), [VPA / Franken-B](https://huggingface.co/tjarvis91/vfaix-vpa-options-trader) | Legacy Qwen3.5-derived vision-language checkpoints. They use a third-party foundation model. |
|
| 21 |
+
| [Q-Office-Suite](https://huggingface.co/Qovaryx/q-office-suite-runtime) | Historical Windows runtime packaging, separate from model weights and training source. |
|
| 22 |
|
| 23 |
+
## Evidence and maturity
|
|
|
|
|
|
|
|
|
|
| 24 |
|
| 25 |
+
This collection contains distinct architectures and release formats. Scratch-initialized bases and Qwen-derived legacy checkpoints have different provenance; a claim about one family does not apply to all of them.
|
| 26 |
|
| 27 |
+
Earlier cards mixed internal fixtures, simulated trading results, and release plans with statements about shipping products. Those broad claims have been removed. There is no current claim here of verified investment returns, universal hardware compatibility, production readiness, or superiority over other models.
|
| 28 |
|
| 29 |
+
The Q-RAG reports evaluate query/passage classification. They do not establish end-to-end file retrieval, path accuracy, or general retrieval quality. Runtime checks and historical test receipts also do not establish current product availability.
|
| 30 |
|
| 31 |
+
Local inference is a project goal and depends on the particular artifact and integration. Broker access, market data, licensing, and optional external services can require a network connection. Review each repository's dependencies and license before use.
|
|
|
|
| 32 |
|
| 33 |
+
## Project and support
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 34 |
|
| 35 |
+
- [Qovaryx status](https://qovaryx.jehorizon.com/)
|
| 36 |
+
- [Public research notes](https://github.com/thron-j/qovaryx-ai-research)
|
| 37 |
+
- [JE Horizon company information](https://jehorizon.com/llms.txt)
|
| 38 |
+
- Support: support@jehorizon.com
|
| 39 |
|
| 40 |
+
J.E. Herizon LLC, doing business as JE Horizon.
|