DrDaq

Lean Python trading assistant. Telegram-driven. Perplexity-backed. Neon-stored. Kovner + Hougaard + ICT voice.

Agent mode

—

Kill switch

—

Daily trades

—

Last snapshot

—

connecting to status.drdaq.me...

Quick reference

/briefing	Morning read — auto-fires 4:44am PT daily	auto
/prices	Live snapshot — NQ, ES, YM, GC, VIX, CL, DX	anytime
/logtrade	Log trade immediately after entry /logtrade MNQ BUY 3 21450 21480 90 breakout	required
/risk	Drawdown, trade count, kill switch state	daily
/debrief	Post-session review after 4pm ET	EOD
/memory_status	Memory health — candidate / quarantined / promoted	anytime
/clear	Wipe in-memory conversation history without restart	anytime
/help	Full command list inside Telegram	in bot

Design principles

Prefer deterministic systems over probabilistic where possible
Keep orchestration modular — bot.py, brain.py, engine.py each own their domain
Treat memory as a governed asset — quarantine first, promote later
Prefer observability over hidden behavior
Never fabricate prices, event times, or economic data
Never mix NQ/MNQ futures with Nasdaq Composite or cash-index values
Soft-quarantine questionable memory before deletion or promotion

Walk me through it

Five concepts that make DrDaq work

1. Runtime

2. Memory

3. AMD + PO3

4. Briefing

5. Risk

How a message becomes a response

Every Telegram message goes through a strict pipeline. Nothing random — each step gates or enriches the message before the model sees it.

Telegram message

→

Auth gate

→

Rate limiter

→

Kill switch

→

State check

Deterministic?

yes →

Stored snapshot

→

Response

no →

Econ calendar fetch

→

Market + Risk + Strategy

→

Perplexity

→

Validator

→

Response

If you ask "what was NQ at 4pm?" and DrDaq has the snapshot stored, it answers from Neon — no model call, no hallucination risk. The price snapshot schedule exists for exactly this reason.

How memory is governed

DrDaq does not store everything you say. Every potential memory goes through a quarantine-first flow with a duplicate guard, deterministic scoring, TTL, and daily expiry.

Candidate text

→

Duplicate guard

→

Score 0–10

→

TTL assigned

Score < min

→

quarantined

Score ≥ min

→

temporary_memory

→

promote → persistent

Scoring: high-signal category (+3), 40+ char content (+2), validation language (+2), high confidence (+2), trusted source (+1). Max 10. TTL: high score + high confidence → 90 days. Score ≥ 5 → 30 days. Otherwise → 7 days. Duplicate guard skips identical content in same category within last 7 days. Daily expiry runs every session loop tick.

/memory_status shows counts by status. /purge_today preview shows what would be quarantined. /clear wipes the in-memory conversation history without touching Neon.

AMD phases and Power of 3

Every timeframe — daily, weekly, monthly — moves through three phases. DrDaq tracks your read in Neon and injects it into every briefing and analysis.

A — Accumulation. Smart money building position. Ranging, false breaks, liquidity hunts. The quiet phase before direction.
M — Manipulation. The stop hunt. Sharp move into liquidity engineered to fill orders against retail.
D — Distribution. The actual directional move. Smart money offloading. Position with this, not against it.

/setphase daily M

→

amd_tracking Neon

→

injected into all briefings

Set phase before the open. Update if it shifts mid-session. London M into NY D is the most common sequence. /po3 shows the current read across all timeframes.

The briefing pipeline

The 4:44am PT briefing is the most important automated event. Here is exactly what happens:

Scheduler fires at 4:44am PT — checks is_am_briefing_time()
Captures briefing_am price snapshot from Yahoo Finance
Fetches verified economic calendar from Finnhub + FRED APIs
Assembles: live prices + daily levels + framework + fractal + PO3 + strategy rules
Structured retrieval runs with verified calendar context already injected
Perplexity Sonar generates briefing in Kovner/Hougaard/ICT voice
Validator checks response — symbol boundaries, price sanity, timestamp
Written to Neon briefings table with full market context snapshot
Sent to all authorized Telegram IDs with DailyTouchpoint postscript

Miss the 4:44am briefing? /briefing triggers it on demand. PM briefing fires at NY PM kill zone open — same pipeline, different prompt focus.

The risk system

Two-trade daily limits and weekly drawdown guard. Code-enforced gates, not suggestions. Kill switch persists through restarts.

Daily trade limit. 2 trades per day max. /risk shows count. Prevents overtrading on bad days.
Weekly drawdown guard. Tracks P&L against high-water mark. Escalates risk messaging when threshold exceeded.
Kill switch. /kill_on — complete silence, no response, disables all conversation. Persists through restarts. Emergency only.
State OFF. /state OFF — returns "Session closed" message. Data collection and snapshots continue. Use at end of day.
Agent modes. OFF — ANALYSIS_ONLY — PAPER — LIVE. All labels until IBKR is wired.

/logtrade + PNL

→

record_trade()

→

weekly_ledger

→

/risk + /weekly

Trade logs export as CSV for California tax/CPA review via /export. Records only, not tax advice. Kill switch does NOT currently gate IBKR execution — must be wired before LIVE.

Daily workflow

What to run, when, and why — in order

Pre-market — before 6:30am PT

4:44am PT

/briefing (auto)

Auto-fires. Macro backdrop, session structure, key levels, verified economic calendar. Read before forming any bias.

pre-market

/prices

Check where Globex opened. Note NQ, ES, GC before anything else.

pre-market

/setlevels MNQ [high] [low]

Feed pre-market high and low once established. DrDaq references these all session.

example: /setlevels MNQ 21500 21350

backfill: /setlevels MNQ 21500 21350 2026-05-14

pre-market

/setphase daily [A/M/D]

Confirm AMD read before the open. Feeds every response until updated.

example: /setphase daily A London building inventory below 21350

pre-market

/ready

Mark yourself ready. Signals DrDaq you are in active session mode.

Session — 9:30am to 4:00pm ET

on entry

/logtrade immediately

Log at entry, not end of day. Add exit and PNL when trade closes. Catalyst feeds memory.

example: /logtrade MNQ BUY 3 21450 21480 90 NY open sweep above VWAP

if shifts

/setphase daily M (or D)

Update if AMD phase changes. London M into NY D is the most common sequence.

any question

Just talk to DrDaq

Conversational messages route through Perplexity with your full market context and verified calendar automatically injected.

before acting

/factcheck [claim]

Verify any economic event time or news. Uses verified Finnhub + FRED data, not Perplexity guessing.

example: /factcheck CPI release is 8:30am ET today

End of session — after 4:00pm ET

4:00pm ET

ny_close (auto)

Captures futures marks for NQ, MNQ, ES, MES, YM, MYM and promotes them into daily close levels in Neon.

EOD

/debrief

Post-session review. DrDaq pulls your logged trades and gives honest feedback on execution and discipline.

EOD

/pnl + /risk

Confirm P&L and check drawdown posture going into tomorrow.

EOD

/state ANALYSIS_ONLY

Close trading session. DrDaq still talks, data still collects. Use this not /state OFF.

Weekly — Monday pre-market

Sunday 6pm ET

/setweekopen MNQ [price]

Record Globex weekly open. Auto-fires at London pre Monday but verify it captured correctly.

Monday AM

/setphase weekly [A/M/D]

Set weekly AMD bias before Monday open. Feeds all briefings for the week.

Monday AM

/weekly

Review last week — P&L, drawdown, high-water mark, discipline, lessons.

Log a trade

Fill in the fields — command builds live — copy to Telegram

Trade details

Symbol

Side

Qty

Entry price

Exit price

P&L ($)

Catalyst

Be specific — this feeds memory and debrief context

/logtrade MNQ BUY 1 —

Syntax

/logtrade SYMBOL SIDE QTY ENTRY [EXIT] [PNL] [CATALYST]
Required: SYMBOL SIDE QTY ENTRY — Optional: EXIT PNL CATALYST

entry only: /logtrade MNQ BUY 3 21450
full close: /logtrade MNQ BUY 3 21450 21480 90 NY open sweep
sell side: /logtrade MNQ SELL 2 21500 21460 80 liquidity sweep

Set levels

Pre-market structure, session reference levels, opens and closes

Pre-market high / low

Set once highs and lows are established — usually 6:00–9:30am ET. DrDaq references these all session. Add an optional date to backfill past sessions.

Symbol

Pre-market high

Pre-market low

Date (optional — for backfill)

Leave blank for today. Enter YYYY-MM-DD to backfill a past session.

/setlevels MNQ — fill fields

Weekly / monthly open

Weekly open = Sunday 6pm ET Globex open. Monthly open = first of month price. Both auto-fire — use this to verify or correct.

Type

Symbol

Price

/setweekopen MNQ —

Manual daily close

ny_close auto-captures futures marks at 4pm ET. Use this only if the snapshot missed.

Symbol

4pm ET close price

/setclose MNQ —

AMD phase

Confirm your read — stored in Neon and injected into every briefing

Set phase

Timeframe

Phase

Notes — be specific

Stored in Neon. Specificity compounds over time in briefings.

/setphase daily A

Phase reference

A — accumulation

Smart money building. Ranging, false breaks, liquidity hunts.

M — manipulation

Stop hunt. Sharp move into liquidity engineered against retail.

D — distribution

The actual move. Smart money offloading. Position with this.

Common sequence

London A or M, NY D

Check current

/po3

Add strategy rule

Rules stored in Neon and seeded into DrDaq's context on every restart

New rule

Strategy rules shape how DrDaq reasons about setups, risk, and discipline. Injected into every briefing and analysis. Write them as principles you have internalized — specific and actionable.

Rule type

Rule text

First person, specific, actionable. The way you would tell yourself.

/addrule risk — fill in rule text

Examples by type

risk

never hold through a major news event

entry

only enter on confirmed liquidity sweep with displacement

discipline

two losses in a row means stop for the day

session

avoid trading the first 5 minutes of NY open

position_sizing

max 3 contracts MNQ until 60 days consistently profitable

All commands

31 commands across 5 categories

Market data

/prices	Live snapshot — NQ, MNQ, ES, MES, YM, MYM, GC, VIX, CL, DX	anytime
/briefing	AM briefing on demand. Auto-fires 4:44am PT. Verified economic calendar injected. no arguments	auto
/debrief	Post-session review — trades, risk, discipline, market read	EOD
/weekly	Weekly summary — P&L, drawdown, high-water mark, trade count	Monday
/bias	Directional bias read across timeframes	anytime
/po3	Power of 3 — current AMD phase read all timeframes	anytime
/fractal	Decade fractal context — long-cycle AMD positioning	anytime

Trade management

/logtrade	Log a trade to Neon and local ledger SYMBOL SIDE QTY ENTRY [EXIT] [PNL] [CATALYST] example: /logtrade MNQ BUY 3 21450 21480 90 NY open sweep	required
/pnl	Today's realized P&L summary	EOD
/risk	Risk state — drawdown %, daily trade count, kill switch	daily
/export	Export trade log CSV for CPA/tax review	monthly

Market structure

/setlevels	Set pre-market high/low — session reference levels SYMBOL HIGH LOW [YYYY-MM-DD] today: /setlevels MNQ 21500 21350 backfill: /setlevels MNQ 21500 21350 2026-05-14	pre-mkt
/setclose	Manual daily close — only if auto-snapshot missed SYMBOL PRICE example: /setclose MNQ 21460	EOD
/setmonthopen	Record monthly open — auto-fires 1st of month SYMBOL PRICE [YYYY-MM] example: /setmonthopen MNQ 21100	auto
/setweekopen	Record weekly open — auto-fires Monday london_pre SYMBOL PRICE example: /setweekopen MNQ 21400	auto
/setphase	Confirm AMD phase — stored in Neon, injected into all briefings TIMEFRAME PHASE [notes] — timeframe: daily weekly monthly — phase: A M D example: /setphase daily M London swept sell-side 21340	pre-mkt

Strategy and memory

/strategy	View current strategy rules in DrDaq's context	anytime
/addrule	Add a strategy rule to Neon TYPE rule text example: /addrule risk never hold through a major news event	anytime
/memory_status	Memory health — candidate / quarantined / promoted / expired counts and oldest entry	anytime
/purge_today	Preview or confirm soft-quarantine of today's memory preview \| confirm	careful
/clear	Wipe in-memory conversation history for current session — does not touch Neon no arguments	anytime
/factcheck	Verify a claim via Finnhub + FRED + Perplexity structured retrieval claim text example: /factcheck Fed meeting is June 11 at 2pm ET	anytime

System

/status	Bot status — mode, kill switch, authorized users, model	anytime
/state	Set agent mode OFF \| ANALYSIS_ONLY \| PAPER \| LIVE	careful
/kill_on	Emergency panic button — complete silence, disables conversation, persists through restarts	emergency
/kill_off	Re-enable after kill switch — check /status before trading	restore
/ready	Mark yourself ready for session	pre-mkt
/collect	Toggle data collection mode — listen and record only, no analysis	anytime
/help	Full command list inside Telegram	in bot

Data dictionary

Every Neon table — columns, types, what's stored, example rows. Tap to expand.

trades

Every logged trade — permanent record for P&L, tax, and debrief

↓

columntypedescription

ts_pt

TEXT

Timestamp in Pacific Time ISO format

symbol

TEXT

MNQ, MES, MYM, NQ, ES, YM

side

TEXT

BUY or SELL

qty

REAL

Number of contracts

entry_price

REAL

Entry price

exit_price

REAL

Exit price — NULL if still open when logged

realized_pnl

REAL

Realized P&L in dollars — NULL if not provided

catalyst

TEXT

What triggered entry — feeds memory and debrief

example: symbol="MNQ" side="BUY" qty=3 entry=21450 exit=21480 pnl=90 catalyst="NY open sweep above VWAP"

price_snapshots

Instrument prices at each of 9+ session labels per day

↓

columntypedescription

session_label

TEXT

london_pre, pre_market, ny_open, ny_am_close, lunch, ny_pm_open, ny_close, globex_reopen, overnight, briefing_am, briefing_pm, on_demand

symbol

TEXT

NQ, MNQ, ES, MES, YM, MYM, GC, VIX, CL, DX

price

REAL

Last price from Yahoo Finance at snapshot time

day_high

REAL

Day high at time of snapshot

day_low

REAL

Day low at time of snapshot

captured_at

TIMESTAMPTZ

When the snapshot was taken UTC

example: session_label="ny_close" symbol="MNQ" price=21460 day_high=21510 day_low=21320

daily_levels

Pre-market highs/lows and 4pm close prices per date

↓

columntypedescription

instrument

TEXT

Instrument

trade_date

DATE

The trading date — YYYY-MM-DD

session_high

REAL

Pre-market high — set via /setlevels

session_low

REAL

Pre-market low — set via /setlevels

closing_price

REAL

4pm ET futures close — auto or /setclose

true_day_open

REAL

2:30am ET true day open — auto at london_pre

example: instrument="MNQ" trade_date="2026-05-15" session_high=21500 session_low=21350 closing_price=21460

temporary_memory

Candidate observations — scored, TTL-governed, quarantine-first

↓

columntypedescription

owner_id

BIGINT

Telegram user ID — owner memory isolated from guests