An LLM (Qwen2.5-7B local) digests every other panel into a structured forecast: bias (bullish/bearish/neutral), confidence, key drivers, 24-72h thesis, top-3 scenarios with probabilities, target/invalidation levels. Forecast use: the AI's trade_bias + confidence is the highest-level synthesis of every input — a high-confidence flip is the single most actionable line on the page.

Tracks every past AI forecast vs. realized 24h/72h price moves. Hit-rate > 60% = analyst is well-tuned to current regime; < 40% = regime change, distrust new forecasts. Forecast use: read this before acting on the AI Brief — a fresh BULLISH call from a 35% hit-rate analyst is far weaker than the same call from a 70% one.

Append-only feed of short narrative shifts (e.g., "funding turned negative as price climbed — squeeze setup"). Forecast use: captures the why of regime turns that single-metric panels can't express; a clustering of narrator entries on one theme is itself a signal.

Weighted -100…+100 score blending funding, OI delta, taker flow, book imbalance, RSI/MACD, sweeps, news, AI bias. ≥+70 strong-bull, ≤-70 strong-bear, ±30 noise. Forecast use: the closest thing to a single "what now" number — directional bias for the next few hours; combine with confluence count to gauge robustness.

Counts how many of ~10 independent indicators currently agree. 8/10 agreement = high-conviction setup; 5/10 = mixed (mean-revert regime). Forecast use: filters the composite score — a +60 with 9/10 confluence is far more durable than +60 with 4/10. Mixed confluence favors range-trading strategies.

24h/7d/30d trends of price, composite score, book imbalance, equity. Forecast use: spots divergence — e.g., price flat but composite score steadily rising = energy building for breakout; price rising but composite falling = exhaustion.

Reference price + bid-ask gap. Widening spread = market makers pulling = imminent volatility. Tight spread = orderly market. Forecast use: spread expansion on flat price often precedes a sharp move within minutes.

Sum of bid depth ÷ ask depth in top of book. >1.5 = buy-side stacked, supports rallies. <0.67 = sell-side wall. Forecast use: persistent skew across many minutes is a leading indicator; flickering ratios are noise from layering bots.

Size-weighted mid (mid biased by which side has more depth). Microprice > mid = market wants higher. Microprice < mid = market wants lower. Forecast use: short-horizon (seconds-to-minutes) directional pressure indicator that beats raw mid-price.

Cumulative USD liquidity at each %-from-mid bucket. Asymmetry shows where stops/walls live. Forecast use: identifies where price will get absorbed (thick zone) vs slip through (thin zone).

Price ranges with abnormally thin book depth. Forecast use: price travels fast through voids — once a void edge is breached, the move usually completes the void's full distance. Critical for take-profit placement.

Aggressive market-orders that ate ≥3 levels of resting liquidity. Buy-side sweep = size buyer urgent; sell-side = size seller urgent. Forecast use: a $1M+ sweep without follow-through often marks a local high/low (liquidity grab); with follow-through, marks a breakout.

Hidden orders detected when a level keeps refilling after being eaten. Forecast use: location of an iceberg = institution's defended price; price tends to bounce off icebergs and break only with surprise volume.

VWAP-clustered walls ≥1.6% from mid, unified across spot+futures. Confidence score grades how durable each wall is (size × age × reaction history). Forecast use: primary swing-trade entries/targets — high-conf supports are buy zones; high-conf resistances are short or take-profit zones.

Running sum of (taker buys - taker sells). Rising CVD = aggressive buyers, falling = aggressive sellers. Forecast use: CVD/price divergence is the killer signal — price up but CVD flat = rally on no buying = suspect; price down but CVD rising = absorption = bullish reversal setup.

Bid vs ask USD within 1% of mid. Forecast use: a 70/30 split with sweeps stacking the heavier side often precedes a directional break of the 1% band.

Cost longs pay shorts (perpetual swaps). Very positive (>0.02%/8h) = longs over-crowded, squeeze risk down. Very negative = shorts crowded, squeeze risk up. Forecast use: contrarian positioning gauge — extreme funding usually marks local exhaustion within 24-48h.

Total outstanding futures contracts. OI rising + price rising = new longs. OI rising + price falling = new shorts. OI falling = position closing (rally/sell-off losing fuel). Forecast use: trend continuation needs OI growth; OI roll-off mid-move warns of reversal.

Retail trader account ratio. Counter-indicator: when retail is >3:1 long, smart money tends to fade. Forecast use: extreme retail positioning = soon-to-be-liquidated side; price magnetizes toward stops.

Real-time stream of liquidated leveraged positions. Long liqs cascade = price flushes lower; short liqs cascade = price squeezes higher. Forecast use: a $5M+ cascade in 5m typically marks the end of the immediate move — capitulation. Use cascade clusters as reversal-watch zones.

Funding spread across Binance/Bybit/OKX. Divergence (one venue much more positive) = arb-driven flow coming. Forecast use: the venue with extreme funding is the one likely to mean-revert first.

Futures price - spot price. Premium > 0.3% = euphoria/leverage demand. Discount = fear/forced unwinds. Forecast use: sustained premium during rally = trend has fuel; flipping to discount = trend exhausted.

Net aggressor side over rolling windows. Persistent taker-buy = trend up has follow-through; taker-sell into rallies = distribution. Forecast use: confirms whether a price move has real flow behind it or is order-book mechanics only.

Visualizes price levels where leveraged longs/shorts get liquidated. Big clusters = magnets for price (market-makers hunt them). Forecast use: price often travels toward the largest nearby liq cluster before reversing — primary stop-hunt prediction.

Implied volatility on at-the-money near-dated options. Rising IV = market expects bigger moves; falling IV = complacency. Forecast use: IV crush after a news event marks the move's end; IV spike from a low base often precedes the move.

OTM put IV - OTM call IV. Positive (put-skewed) = downside fear, traders paying up for crash protection. Negative (call-skewed) = upside chase. Forecast use: extreme positive skew often marks bottoms (max fear); extreme negative skew marks tops.

Total put open interest ÷ call OI. PCR >1 = more hedging/bearish positioning; PCR <0.7 = call-heavy/bullish chase. Forecast use: contrarian — PCR extremes mark sentiment-driven turns.

IV across expiries. Backwardation (front > back) = near-term stress. Contango (back > front) = orderly market. Forecast use: backwardation signals a known catalyst window; the IV peak expiry hints at when traders expect the move.

Strikes with the heaviest open interest. Acts like magnets near expiry (max-pain dynamics). Forecast use: in the final 5-7 days before monthly expiry, price tends to gravitate to the largest-OI strike — useful for setting tactical targets.

Pending transaction count + sat/vB fee. Fee spikes = network demand surge (often retail FOMO entry/exit). Forecast use: coincident fee spikes with price spikes = blow-off character; fee collapse with price flat = waning interest.

Network security + miner economics. Falling hashrate during low price = miner capitulation = often a bottom signal. Forecast use: hashrate trend confirms cycle position — sustained hashrate ATH during pullbacks = strong network conviction.

Composite of mempool depth + fee market. Forecast use: sustained high congestion = active settlement demand = bullish flow context; ghost-town mempool = waning use case attention.

15m/1h/4h/1d candles with overlays: green/red bias ribbon (AI), S/R lines, paper-trade ▲▼ markers, VWAP. Forecast use: primary visual — confirm every other signal against the actual price structure (HH/HL = uptrend, LH/LL = downtrend).

Volume-weighted average price. Above VWAP = buyers in control intraday; below = sellers in control. Forecast use: institutional execution benchmark — VWAP is where size accumulates; first-touch reactions are high-probability mean-revert entries.

Momentum oscillator. >70 overbought, <30 oversold. RSI/price divergence > absolute level. Forecast use: use divergence to anticipate reversal; absolute levels alone are noisy in trends.

Trend + momentum hybrid (12/26 EMA difference vs signal). Bullish cross = trend turn up; histogram expansion = momentum accelerating. Forecast use: confirms regime — best for filtering false breakouts on smaller timeframes.

20-SMA ± 2 standard deviations. Squeeze (bands tight) = volatility coiling; expansion = breakout in progress. Forecast use: a band-walk (price riding the upper/lower band) signals strong trend; sharp band-rip with no follow-through = exhaustion.

Stack order of EMAs (e.g., 9 > 21 > 50 > 200 = strong uptrend). Forecast use: instant trend-strength gauge — full alignment = trade with trend; tangled EMAs = stay out / range-trade.

Composite 0-100 sentiment gauge. >75 Extreme Greed (contrarian sell), <25 Extreme Fear (contrarian buy). Forecast use: mean-reversion signal at extremes; in mid-range it's just noise.

BTC market cap ÷ total crypto cap. Rising = capital rotating into BTC (often defensive); falling = alts season risk-on. Forecast use: defines the macro crypto regime — BTC outperformance vs alts hints at risk appetite.

Curated news, scored LOW/MEDIUM/HIGH relevance. HIGH items move price. Forecast use: a HIGH item within the last 10 min explains/anticipates the next move; cluster of HIGH items = thematic catalyst regime.

Background news from major sources (CoinDesk, Bloomberg, etc.). Forecast use: context for narrative shifts the AI Narrator picks up; spot themes (regulation, ETF flows) before they reach Telegram noise.

FOMC, CPI, NFP, etc. (next 7 days, High/Medium impact). Forecast use: price often compresses ahead of high-impact prints (vol crush) and explodes on release. Avoid initiating directional trades within 1h of a High event unless that's the thesis.

Fires on: composite ±70 cross, AI bias flip, $1M+ sweep, HIGH-relevance news (<10m), $5M+ liq cascade in 5m. Forecast use: the "look at the dashboard right now" filter — every alert has historically preceded a tradable move. Stack of alerts in <5min = high-conviction window.

Auto-executes the AI's recommended bias against live data. Tracks open positions, closed P&L, equity curve. Forecast use: a flat/falling equity curve over weeks = the system is mis-tuned to the current regime — distrust signals; rising curve = current regime suits the model, lean in.