Equity Advisor — How It Works

🏦

India Monthly Portfolio

Long-term investing — Nifty 500 screener, AI advisory, automated queue execution

Goal: Pick the top 10 Indian stocks monthly from Nifty 500 using fundamentals + sentiment + momentum. You review the AI advisory, decide which to buy, queue them — and the system executes at 9:15 AM, places a trailing stop-loss GTT on fill, and updates stops daily at 2:30 PM.

🚂

Railway

API server, screener, sentiment services. Always-on cloud.

🖥️

Oracle VPS

Zerodha order execution. Static IP required by SEBI.

📱

Dashboard

Netlify-hosted. Your control panel on mobile.

🔍

Monthly Screener runs

Railway Monthly · Manual

▶

📊Screens entire Nifty 500 universe — ~500 stocks ranked by composite score

⚙️Score = fundamentals (ROE, PE, PEG, revenue growth, D/E) + momentum (1M, 3M, 6M) + sector sentiment bonus/penalty

📐ATR-based stop-loss computed per stock using 14-day historical data from yfinance

🤖Claude LLM generates monthly advisory: action (HOLD / ADD / EXIT), holdings health, top picks with rationale

📤Results posted to: /portfolio/picks/upload · /portfolio/advisory/upload

screener.py → /portfolio/advisory (advisory + top_picks list) screener.py → /portfolio/picks (full bucket with ATR data)

📱

You review Advisory on dashboard

Netlify

▶

📋Monthly Advisory section shows: AI action recommendation, holdings health (HOLD/TRIM/EXIT verdict per stock), top picks ranked by score

🎯Per-pick entry info fetched live from Zerodha: 20-DMA, 50-DMA, optimal entry, 14-day ATR trailing stop

📦Per-pick actions: Q = queue for 9:15 AM · + = add to portfolio immediately · ORDER = send to Zerodha now

Stock Picks section (top of India tab) shows only your actual live holdings. Advisory section shows screener recommendations — separate until you consciously move them.

🕘

9:15 AM — Orders placed automatically

Oracle VPS Cron 9:15 AM IST

▶

📂india_entry.py fetches all queued BUY entries from Railway

💹Gets live open price from Zerodha (kite_executor.py on VPS port 5001)

⚖️Decision logic:

live > max gate (+3%) → ⏭ SKIP (gap-up, R/R broken) live ≤ 20-DMA entry → MARKET order at open live between entry → LIMIT order at 20-DMA and gate

📲Telegram notification sent with order summary

🔄Status updated: queued → order_placed

✅

Fill detected → GTT stop-loss placed

Railway Zerodha Webhook

▶

📡Zerodha sends a postback webhook to Railway /kite/postback instantly when order fills

🏦Stock auto-added to your live portfolio (portfolio_live.json) — appears in Stock Picks section on dashboard

🛡️Trailing stop-loss GTT placed automatically at: fill_price − 1.5×ATR

🔄Status: order_placed → filled · GTT ID stored in queue

If limit order doesn't fill by 3:15 PM: india_cancel.py cancels it on Zerodha and resets status back to queued — retries automatically tomorrow at 9:15 AM until price pulls back to 20-DMA.

🔁

2:30 PM daily — Trailing stop ratcheted up

Oracle VPS Cron 2:30 PM IST

▶

📈trailing_stop.py fetches all filled India entries from Railway

💹Gets today's high from Zerodha for each position

📐New stop = today's high − 1.5×ATR

☝️Stop is only ever moved up — never lowered. If new stop ≤ current stop, no change.

🔔Updates GTT on Zerodha + sends Telegram if stop moved

Exit: When price hits your GTT stop, Zerodha auto-sells. The postback fires → stock removed from live portfolio → trade logged to closed positions.

📊

Performance tracked vs Nifty 50 & 500

Railway

▶

📸Daily P&L snapshots saved to performance_history.json

📉Dashboard chart shows portfolio cumulative return vs Nifty 50 and Nifty 500 benchmarks (1W / 1M / 3M / 1Y)

💰Closed positions tracked with realised P&L, win rate, and date range

📅 Daily Schedule (IST)

9:15 AM

📥 india_entry.py — Place buy orders for queued stocks

3:15 PM

⏸ india_cancel.py — Cancel unfilled LIMITs, re-queue for tomorrow

2:30 PM

📐 trailing_stop.py — Ratchet up GTT stops using day's high

Any time

⚡ /kite/postback — Fill detected, GTT placed, portfolio updated

⚡

Swing Trading

Short-term trades — daily scanner, news sentiment, R/R-based entries with auto GTT targets

Goal: Identify 5–10 day momentum setups daily from Nifty 500. The scanner runs overnight, scores stocks on technical + sentiment signals, and presents candidates. You review and queue — orders execute at 9:15 AM with automated stop-loss and two take-profit GTTs placed on fill.

🌙

Nightly scan — candidates identified

Railway 10:30 PM IST daily

▶

📡swing_news_sentiment.py fetches today's news headlines for each Nifty 500 sector

🤖Claude LLM scores each sector: positive / mild_positive / neutral / cautious / negative

📊Technical scan scores stocks on: RSI, MACD, Bollinger, volume surge, 20-DMA position, momentum

🎯Combines tech score + sentiment score → top candidates with entry, stop, T1 (1.5×R), T2 (2.5×R), R/R ratio

🔀History blending: 40% today's signal + 60% 7-day rolling average → smooths out single-day noise

📤Results posted to Railway: /signals/upload (swing_news_sentiment) + /swing/candidates/upload

SCAN_DAYS = 1 (today's news only, blended with history) ATR_MULT = 2.0 (stop = entry − 2×ATR) T1 = entry + 1.5 × risk T2 = entry + 2.5 × risk

👁️

You review swing candidates on dashboard

Netlify

▶

🃏Swing tab shows candidate cards sorted by score — entry price, stop-loss, T1, T2, conviction level, R/R ratio

🏷️Signal badges on each card: RSI oversold, MACD cross, BB squeeze, volume surge, gap-up, etc.

🌐Category Sentiment section shows LLM's sector verdict (positive/cautious) — feeds into candidate scoring

📦Per-candidate actions: Q = queue for tomorrow's open · Buy Now = enter immediately at LTP

Queuing opens a modal with live Zerodha price, entry basis, ATR computed from 200-day OHLCV, suggested quantity for your risk budget.

🕘

9:15 AM — Swing entry orders placed

Oracle VPS Cron 9:15 AM IST

▶

📂swing_entry.py fetches all queued swing entries from Railway

💹Checks live open price vs optimal_entry and limit_price gate

⚖️Same logic as India: gap-up → skip, at/below entry → MARKET, between entry and gate → LIMIT

📲Telegram alert with entry summary for each stock

🎯

Fill detected → Stop + T1 + T2 GTTs placed

Railway Zerodha Webhook

▶

📡Zerodha postback fires at /kite/postback instantly on fill

📦Stock added to Swing Live Positions panel on dashboard

🛡️Stop-loss GTT placed at scan's computed stop price (ATR-based)

🏆Target 1 GTT placed automatically: entry + 1.5 × risk (50% of position)

🏆Target 2 GTT placed automatically: entry + 2.5 × risk (remaining 50%)

GTT 1 (stop): sell ALL @ stop_loss (protective) GTT 2 (T1): sell 50% @ target1 (partial profit) GTT 3 (T2): sell 50% @ target2 (full exit) All three placed atomically on fill — no manual action needed.

✅ Fully automated: When T1 fires, the postback handler automatically cancels the old stop GTT (full quantity) and places a new stop GTT for the remaining 50% shares at the same price. You get a Telegram confirmation. No manual action needed.

📐

2:30 PM daily — Trailing stop updated

Oracle VPS Cron 2:30 PM IST

▶

📈Same trailing_stop.py handles swing positions alongside India monthly

💹Fetches today's high for each open swing position

☝️If today's high − 1.5×ATR > current stop → update the stop GTT upward

📆Max hold period tracked — alerts if position has been open longer than scan's max_days

🎯 Risk/Reward Structure

Stop

2×ATR

Target 1

1.5×R

Target 2

2.5×R

Min R/R

1.5×

Max Hold

10 days

🧠

Intelligence Layer

News sentiment, earnings signals, LLM analysis — feeds both India Monthly and Swing services

Goal: Provide context-aware market intelligence that adjusts stock scores and surfaces macro themes. Three pipelines run on Railway and feed their outputs into a shared signal store — the screener and swing scanner consume these signals when making decisions.

📰

Swing News Sentiment — daily sector analysis

Railway Service 10:30 PM IST daily

▶

🌐Scrapes today's financial news headlines from NSE/BSE sectors

🤖Claude LLM classifies each sector's news sentiment:
positive / mild_positive / neutral / cautious / negative

🔀History blending: raw signal weighted 40% today + 60% 7-day rolling average

📤Posts to Railway: swing_news_sentiment + swing_sentiment_history

📊Dashboard: "Category Sentiment (LLM Analysed)" section shows per-sector verdict with reasoning

⚙️Consumed by: Swing scanner (sector bonus/penalty) + Monthly screener (sentiment_adj score)

Score adj: positive → +5 mild_positive → +2 neutral → 0 cautious → −2 negative → −5

📈

Monthly Earnings Sentiment — macro + earnings analysis

Railway Service Monthly · before screener

▶

📋Analyses recent earnings seasons: beat/miss rates per sector, guidance tone, revenue trends

🏛️Reads RBI/SEBI/PIB policy signals — rate decisions, liquidity, sector-specific regulations

🤖Claude LLM synthesises into per-sector earnings signal: positive / neutral / cautious / negative

🔗First tries to load swing_sentiment_history from Railway — if available, blends with earnings data for richer context

📤Posts monthly_earnings_sentiment to Railway signals store

⚙️Consumed by: Monthly screener's earnings_adj score adjustment per stock

Cross-service data sharing: swing_news_sentiment POSTs its history to /signals/upload so monthly_earnings_sentiment can read it even though they're separate Railway services with separate volumes.

🏛️

Signal Store — shared API across all services

Railway API

▶

🗄️Railway API's /signals endpoint serves all collected intelligence in one response

📦Contains: swing_news_sentiment · monthly_earnings_sentiment · swing_sentiment_history · policy_signals · llm_synthesis

💾Each signal type persisted to /data/{type}.json on Railway's persistent volume

📱Dashboard reads /signals at load time — powers Category Sentiment, AI Macro Verdict, sector badges

GET /signals → { swing_news_sentiment: { signals: { IT: "positive", ... } }, monthly_earnings_sentiment: { ... }, swing_sentiment_history: [ {date, signals}, ... ], policy_signals: { ... }, llm_synthesis: { ... } }

🔄

How signals flow into stock scores

Screener Logic

▶

1️⃣Base score = fundamentals (ROE, PE, PEG, D/E, revenue growth) + momentum (1M, 3M, 6M)

2️⃣Sector sentiment adj = swing_news_sentiment verdict for the stock's NSE sector (±0 to ±5 pts)

3️⃣Earnings adj = monthly_earnings_sentiment verdict for sector (±0 to ±5 pts)

4️⃣Final score = globally normalised across all 500 stocks → top 10 selected

🚫Stocks in sectors with negative sentiment are excluded entirely from picks (circuit breaker)

final_score = base + sentiment_adj + earnings_adj (normalised 0–100 across Nifty 500 universe) negative sector → stock excluded regardless of fundamentals

📊 Signal Key

positive +5 score bonus · sector included · confidence for new entries

mild_positive +2 score bonus · sector included · moderate confidence

neutral No adjustment · fundamentals drive the decision

cautious −2 score penalty · sector still eligible · watch closely

negative −5 penalty + sector excluded entirely from picks