← Chat Archive

<tool_call> {"name": "terminal", "arguments": {"command": "whoami"}} </tool_call>

Hey! How's it going? What can I help you with today?

Found a `BOT_CREATION_CHECKLIST.md` in the minimax folder. Let me read it along with `bot_rules.md`:

Found two very useful bot creation documents: ### 1. `/home/ubuntu/bots/Claude_bot/minimax/BOT_CREATION_CHECKLIST.md` A comprehensive 620-line checklist covering the full bot creation process: | Section | What it covers | |---------|---------------| | 1. Configuration | `momentum_config.py` - symbol, futures vs stock LTP pitfall, lot size, expiry format | | 2. API Calls | `quotes(symbol=)` not `quote(token=)`, LTP fetch example, timeout settings | | 3. Stop Loss Logic | Option value decreases when wrong direction, using futures for LTP vs stock | | 4. Risk Management | Entry retry logic, option LTP validation (MUST validate before orders), exit retry + hedge, position verification via broker | | 5. Database/Candles | Incremental candle updates, filtering in-progress candles | | 6. State Management | Save/load state for bot restart recovery | | 7. Telegram Alerts | Alert types and implementation | | 8. Polling Intervals | LTP (5min), ROC (15min), heartbeat (30min) | | 9. Market Hours | Skip non-trading hours, holiday list | | 10. Systemd Service | Service file template + install/control commands | ### 2. `/home/ubuntu/bots/three_candle_bot/bot_rules.md` A shorter 43-line rules document for that bot: - Symbol construction (futures vs options format) - Max 2 lots position limit - Limit orders only (never market) - Exit failure protocol (2 attempts → hedge → alert) - 15-min candle SL checks, 5-min polling - Market open position check (orphan detection) --- **Summary:** There's a solid checklist already in place at `/home/ubuntu/bots/Claude_bot/minimax/BOT_CREATION_CHECKLIST.md`. Want me to read the rest of it (lines 501-620) or start using it as a template for your new bot?

Let me read the rest of both bots to get the full picture:

Now I have a full picture of both bots. Here's the breakdown of **nuances** I found — things that aren't in the generic checklist but matter a lot in practice: --- ## Key Nuances Across Both Bots ### 1. **Dual-Broker Setup** (three_candle_bot) The 3-candle bot uses **two different brokers simultaneously**: - **Kotak** for live trading (quotes, orders, positions) - **Flattrade** for historical candle data (Kotak has NO history API) - This is a critical infrastructure decision — the checklist implies single-broker but in practice they needed Flattrade for candles since Kotak doesn't provide them. ### 2. **LTP Parsing — The `data` Wrapper Problem** Both bots handle responses differently across brokers: ```python # 3-candle bot uses quotes() and handles 'data' wrapper: def _parse_ltp(self, data): inner = data.get('data', data) # ← Kotak wraps in 'data' for key in ('ltp', 'lp', 'last_price'): v = inner.get(key, 0) ``` But momentum_bot uses `quote(token=)` (wrong API call per the checklist!) and doesn't handle the `data` wrapper — it would **fail on Kotak** but works on Flattrade/Shoonya (no wrapper). ### 3. **In-Progress Candle Filtering** Both bots filter out candles younger than the interval period: ```python # momentum_bot: candles = [c for c in candles if (now_ts - c['timestamp']) >= interval_seconds] # 3-candle_bot: c['ts'] < current_5min # slot-based exclusion ``` This is **critical** — in-progress candles corrupt indicator calculations. ### 4. **3-Candle Bot's Candle Architecture** (unique) The 3-candle bot builds its own candles locally: - Fetches 5-min from Flattrade → stores in SQLite - Builds 15-min OHLC from 3× 5-min candles in DB - Replays 15-min window on restart to restore state - This is a **sophisticated multi-layer local candle system** not covered in the checklist ### 5. **State Reset on New Day** ```python # 3-candle_bot — keeps position but resets counters if s.get("date") != today: s = {**DEFAULT_STATE, "date": today, "position": position} # ← position preserved ``` NRML positions carry overnight; only counters reset. ### 6. **Orphan Position Detection** Both bots check broker for positions not tracked in state: ```python # 3-candle_bot total = self.total_broker_long_lots() if total and total > 0: self.telegram.send(f"⚠️ ORPHAN POSITION | {total} lots at broker") ``` This catches cases where orders filled but bot state is out of sync. ### 7. **optionsymbol API Fallback** ```python # 3-candle_bot resp = self.client.optionsymbol(exchange="NFO", expiry_date=EXPIRY_DATE, offset="ATM", option_type=option_type, underlying=SYMBOL_UNDER) # Falls back to manual if API fails: return f"{SYMBOL_UNDER}{EXPIRY_DATE}{atm_strike(fut_ltp)}{option_type}" ``` ### 8. **Candle Staleness Alerting** 3-candle bot has a dedicated stale candle detector: - Triggers after 20 min gap in 5-min candle feed - One alert per episode (throttled with `_candle_stale_alerted` flag) ### 9. **Exit Price Logic — BUY vs SELL Offsets Are Opposite** ```python # 3-candle_bot — CHECKLIST EVEN EMPHASIZES THIS offset = LIMIT_OFFSET if action == "BUY" else -LIMIT_OFFSET price = round(opt_ltp + offset, 1) # Same formula, opposite meaning! ``` `+5` for BUY (aggressive fill), `-5` for SELL (passive fill). The checklist calls this out explicitly. ### 10. **momentum_bot BUG: Uses `quote(token=)` not `quotes(symbol=)`** Line 666-669 and 954: `self.client.quote(token=..., exchange=...)` — the checklist explicitly warns this is wrong. The 3-candle bot does it correctly with `quotes(symbol=...)`. ### 11. **BOD Setup Pattern** 3-candle bot has elaborate BOD (Beginning of Day) initialization: - Compute/store VIDYA anchors - Seed intraday candles from Flattrade - Replay 15-min window to restore state - Check overnight positions at broker ### 12. **Order Verification — Broker as Source of Truth** Both bots verify fills by **re-querying broker position** after every order: ```python time.sleep(2) qty = self.get_broker_qty(sym) if qty and qty >= LOT_SIZE: # confirmed ``` Never trust the order response — always check broker. --- ## Summary: What's NOT in the Checklist 1. Dual-broker architecture (Kotak for trading, Flattrade for candles) 2. The `data` wrapper parsing for Kotak responses 3. Local multi-layer candle building (5-min → 15-min in SQLite) 4. Candle staleness monitoring & alerting 5. Orphan position detection at every poll 6. `optionsymbol` API fallback chain 7. BOD VIDYA anchor computation pattern 8. `quote(token=)` vs `quotes(symbol=)` — momentum_bot uses the wrong one Want me to update the checklist with these nuances, or start building the new bot based on this understanding?