@vizzor/cli 0.14.6 → 0.15.1

package/README.md

@@ -50,6 +50,22 @@ ETH at $2,112 | Bullish | Confidence: Medium-High
 
 ---
 
+## What's new in v0.15.1 — Polaris
+
+> [Full release notes →](https://github.com/7ayLabs/vizzor/releases/tag/v0.15.1)
+
+- **Four new free-tier signal sources** — Deribit options term structure + 25-delta skew, Hyperliquid DEX positioning vs CEX funding divergence, Reddit + 4chan retail-sentiment NLP with contrarian top/bottom flags, and a 30-day funding-rate z-score from a new history table.
+- **Closed-loop WR observability** — every signal records its vote + outcome, the auditor classifies each signal as INVERT / DAMPEN / WATCH / KEEP from rolling WR, and a polarity registry applies sign + magnitude overrides BEFORE CF combination. New `/wr-diagnose` + `/signal-override` commands.
+- **Adversarial signal-conflict gate** — when onChain CF and logicRules CF disagree by >0.6 magnitude in opposite directions, the prediction is reframed as RANGE instead of emitting low-conviction directional noise.
+- **Calibration acceleration** — regime-bucketed Platt shards per (family, regime), cross-symbol per-family prior inheritance, confidence-weighted SGD, calibrator bootstrap from a 30-day rolling window.
+- **Production-grade auto-deploy** — merging to `main` auto-SSHes to the VPS, pulls the new image from GHCR, and rolls the docker stack in ~3-5 min. Health-gated, concurrency-protected.
+- **Per-user accuracy correctness** — `user_id` persisted at `logPrediction` insert time closes the long-standing bug where every operator on the same bot saw a shared WR.
+- **Session-scoped DM delivery + orphan re-guard** — old DMs from previous sessions never deliver, mid-session deletes suppress queued DMs.
+- **Env-aware data dir** (`VIZZOR_DATA_DIR`) — dev and prod can now run on the same machine without thrashing shared SQLite state.
+- **Sentry, Prometheus `/metrics`, bearer-auth `/health`** — observability for unattended 24/7 operation.
+
+---
+
 ## Why Vizzor?
 
 Most crypto tools show you what already happened. Vizzor shows you what's likely to happen next.
@@ -72,6 +88,7 @@ That's it. Vizzor launches a guided setup wizard that walks you through:
 - Choosing your AI provider (Claude, GPT-4, Gemini, or local Ollama)
 - Setting up API keys
 - Picking your default chain
+- Optionally installing the ChronoVisor ML engine on first run
 - Learning the first commands
 
 ### Start predicting
@@ -88,7 +105,8 @@ That's it. Vizzor launches a guided setup wizard that walks you through:
 ### Better prediction accuracy
 
 After setup, follow these tips for the best results:
-- Run `/cosmos on` to enable full data pipeline
+- Enable the ChronoVisor engine during setup, or run `vizzor chronovisor setup` later
+- Run `/cosmos on` for stricter tracked-only qualification and fewer low-edge setups
 - For intraday: focus on `BTC`, `ETH`, and `SOL` with `2h`, `3h`, `4h`, `6h` horizons
 - For scalping: focus on `BTC` and `ETH` with exact short horizons like `10m`, `20m`, `45m`
 - For long-range: ask for `7d`, `30d`, `90d`, or `1y` scenario ranges
@@ -155,6 +173,8 @@ docker compose up -d      # Starts CLI + ChronoVisor Engine + PostgreSQL + Web D
 |---------|-------------|
 | **Price predictions** | Bull/bear/likely targets across 9 timeframes (5m to 3 months), batch multi-horizon in a single call |
 | **ChronoVisor engine** | Mathematical prediction system: Bayesian inference, certainty-factor algebra with weak signal dampening, first-order logic rules (17 base + 7 scalping), meta-reasoning, and continuous weight learning with feedback loops |
+| **Triple-system A+ gate** | Predictions promoted to tracked tier only when Vizzor TA + SMC structure + ICT timing all align on direction (Vizzor's mental model: take a trade only when momentum, structure, and timing agree) |
+| **Persistent calibration** | Platt calibrator, FOL rule tracker, and per-symbol volatility registry survive process restarts — tracked WR no longer rebuilds from zero after every deploy |
 | **Alert-hit notifications** | Toast banners when price targets hit, with training feedback (accuracy, streak, weight updates), ESC-to-dismiss, auto-dismiss |
 | **Prediction markets** | Integrates Gamma and Polymarket data for market-implied probabilities |
 | **Polymarket edge scanning** | Compare ChronoVisor predictions vs Polymarket odds, detect mispricings, execute trades via CLOB API |
@@ -165,6 +185,9 @@ docker compose up -d      # Starts CLI + ChronoVisor Engine + PostgreSQL + Web D
 |---------|-------------|
 | **Multi-exchange pricing** | Aggregated prices from 15 exchanges (Binance, Coinbase, Kraken, OKX, Bybit, Gate.io, KuCoin, HTX, MEXC, Bitget, CoinGecko, CryptoCompare, DexScreener, GeckoTerminal, Jupiter) with VWAP/median aggregation and outlier detection |
 | **Microstructure analysis** | Market structure, FVGs, VWAP, volume delta, liquidation map, order book depth, S/R zones, squeeze detection |
+| **CVD / aggressor delta** | Cumulative volume delta from Binance trade stream — short/mid divergence + impulse detection |
+| **Anticipatory signals** | Whale Alert exchange flow, stablecoin mint/burn velocity, token unlock calendar, Coinbase premium spread, SEC EDGAR pre-news filings — fire before price moves, not after |
+| **SMC + ICT layer** | Smart Money Concepts aggregator (BOS, CHoCH, liquidity sweeps, order blocks) + ICT timing (London/NY kill-zones, Asian range sweeps) |
 | **Derivatives positioning** | Funding rates, open interest, long/short ratios from Binance Futures |
 | **Sentiment analysis** | Fear & Greed Index + news sentiment + buy/sell transaction ratios |
 | **Trending discovery** | What's moving right now across DexScreener and CoinGecko |
@@ -188,6 +211,7 @@ docker compose up -d      # Starts CLI + ChronoVisor Engine + PostgreSQL + Web D
 | **Backtesting** | Test strategies against historical data with full metrics (return, win rate, Sharpe, drawdown) |
 | **Wallet management** | AES-256-GCM encrypted wallets with scrypt key derivation |
 | **Notifications & alerts** | Price alerts for all prediction types (directional + sideways breakout), interactive alerts panel with dismiss-on-review |
+| **AI auto-armed alerts** | When the AI composes a directional trade plan (Entry/TP/SL), the `set_trade_plan_alerts` tool automatically arms price-threshold alerts at those levels with confirmation footer in the reply |
 
 ### Platforms
 
@@ -197,7 +221,7 @@ docker compose up -d      # Starts CLI + ChronoVisor Engine + PostgreSQL + Web D
 | **Web Dashboard** | Next.js 15 — AI chat, markets, agents, portfolio, Polymarket, notifications |
 | **REST API** | Fastify + Swagger — 40+ endpoints with API key auth and rate limiting |
 | **Discord Bot** | Full command suite with AI-powered chat via @mention |
-| **Telegram Bot** | Full command suite with AI chat on any message |
+| **Telegram Bot** | Full command suite with AI chat on any message — production-ready with `/health` endpoint, graceful shutdown, automated backups, and 6-hour state persistence (Platt calibrator + rule tracker survive restart). New commands: `/alerts` (list armed alerts), `/whoami` (allowlist onboarding), `/chronovisor minsystems` (D4 gate runtime tuning) |
 
 ---
 
@@ -457,6 +481,12 @@ WS   /ws                              # WebSocket real-time push
 
 The ChronoVisor engine is an optional Python-based prediction accelerator with 16 trained models that enhances Vizzor's predictions with ML. When unavailable, Vizzor falls back to its rule-based heuristics automatically — no setup required to start using Vizzor.
 
+If your tracked setups keep getting skipped, check whether ChronoVisor is actually enabled and healthy:
+
+```bash
+vizzor chronovisor status
+```
+
 ### Quick setup
 
 ```bash
@@ -486,6 +516,18 @@ curl http://localhost:7200/health
 | **Runtime** | Docker (recommended) or Python 3.11+ |
 | **GPU** | Not required — CPU-based by default |
 
+### Production deployment
+
+A `docker-compose.prod.yml` overlay layers production hardening on top of the base compose: `restart: always`, JSON log rotation (10MB × 3 files per service), and the `vizzor-backup` sidecar that runs `scripts/backup.sh` every 6 hours with 14-day retention. The bot exposes a `/health` endpoint (default `:7101`) compatible with UptimeRobot, BetterStack, and Docker healthchecks.
+
+```bash
+# On the VPS, after first build:
+docker compose -f docker-compose.yml -f docker-compose.prod.yml --env-file .env up -d
+curl http://localhost:7101/health      # → {"ok":true,"status":"healthy",...}
+```
+
+See [`scripts/backup.sh`](scripts/backup.sh), [`scripts/restore.sh`](scripts/restore.sh), and [`scripts/deploy-gen-env.py`](scripts/deploy-gen-env.py) for the operator-facing backup, restore, and `.env` bootstrap helpers.
+
 ---
 
 ## Configuration

package/chronovisor-engine/Dockerfile

@@ -0,0 +1,12 @@
+FROM python:3.11-slim
+
+WORKDIR /app
+
+COPY pyproject.toml .
+COPY src/ src/
+
+RUN pip install --no-cache-dir .
+
+EXPOSE 7200
+
+CMD ["uvicorn", "src.server:app", "--host", "0.0.0.0", "--port", "7200"]

package/chronovisor-engine/models/regime_detector/metrics_1773536830.json

@@ -0,0 +1,36 @@
+{
+  "accuracy": 0.22,
+  "per_class": {
+    "trending_bull": {
+      "precision": 0.14285714285714285,
+      "recall": 0.09090909090909091,
+      "f1-score": 0.1111111111111111,
+      "support": 33.0
+    },
+    "trending_bear": {
+      "precision": 0.208955223880597,
+      "recall": 0.45161290322580644,
+      "f1-score": 0.2857142857142857,
+      "support": 31.0
+    },
+    "ranging": {
+      "precision": 0.20689655172413793,
+      "recall": 0.2222222222222222,
+      "f1-score": 0.21428571428571427,
+      "support": 27.0
+    },
+    "volatile": {
+      "precision": 0.2857142857142857,
+      "recall": 0.1875,
+      "f1-score": 0.22641509433962265,
+      "support": 32.0
+    },
+    "capitulation": {
+      "precision": 0.3333333333333333,
+      "recall": 0.14814814814814814,
+      "f1-score": 0.20512820512820512,
+      "support": 27.0
+    }
+  },
+  "test_samples": 150
+}

package/chronovisor-engine/models/rug_detector/metrics_1773536830.json

@@ -0,0 +1,7 @@
+{
+  "accuracy": 0.96,
+  "precision": 0.9285714285714286,
+  "recall": 0.8666666666666667,
+  "f1": 0.896551724137931,
+  "test_samples": 150
+}

package/chronovisor-engine/models/sentiment_nlp/metrics_1773536831.json

@@ -0,0 +1,4 @@
+{
+  "accuracy": 0.37333333333333335,
+  "test_samples": 75
+}

package/chronovisor-engine/models/trend_scorer/metrics_1773536830.json

@@ -0,0 +1,5 @@
+{
+  "r2": 0.9008962744572128,
+  "mae": 9.576466093118714,
+  "test_samples": 150
+}

package/chronovisor-engine/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "vizzor-ml-sidecar"
-version = "0.14.5"
+version = "0.15.0"
 description = "ML sidecar for Vizzor — LSTM, RF, Isolation Forest, Rug Detection, Wallet Classification, Sentiment NLP, Trend Scoring, TA Interpretation, Regime Detection, Strategy Bandit, Project Risk, Portfolio Optimization, Intent Classification"
 requires-python = ">=3.11"
 dependencies = [
@@ -9,7 +9,7 @@ dependencies = [
     "torch>=2.5.0",
     "scikit-learn>=1.6.0",
     "psycopg2-binary>=2.9.10",
-    "numpy>=2.1.0",
+    "numpy>=1.26,<2.0",
     "pandas>=2.2.0",
     "joblib>=1.4.0",
     "transformers>=4.40.0",

package/chronovisor-engine/src/server.py

@@ -664,7 +664,7 @@ async def lifespan(_app: FastAPI):
 
 app = FastAPI(
     title="Vizzor ML Sidecar",
-    version="0.14.5",
+    version="0.15.0",
     lifespan=lifespan,
 )
 
@@ -1509,6 +1509,7 @@ async def catalog_roadmap():
 @app.get("/health")
 async def health():
     return {
+        "version": app.version,
         "models": [
             {
                 "name": "lstm-predictor",