docs: rewrite README to match current CLI + workflow

Reflects everything that shipped after the initial teammate pass:
calibrate region-select flow, startup delays, auto-focus, canary startup
check, --start-at/--stop-at, atm debug, samples/ auto-save, fires/
annotated screenshots, Task Scheduler setup, current troubleshooting.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Claude Agent
2026-04-16 13:06:12 +00:00
parent ec86f52f1f
commit 34fde8328c

325
README.md
View File

@@ -1,179 +1,216 @@
# ATM — Automated Trading Monitor
Personal tool for the **M2D strategy** on TradeStation DIA/GLD charts with US30/XAUUSD execution on TradeLocker. The bot watches the colored dot strip produced by the *M2D MAPS* custom indicator and sends a Telegram/Discord notification (with chart screenshot + SL/TP levels) when a BUY or SELL trigger fires — so you execute the trade manually in TradeLocker instead of watching two screens.
Personal Faza-1 tool for the **M2D strategy**. Watches the M2D MAPS colored-dot strip on a TradeStation chart, runs a phased state machine (ARMED→PRIMED→FIRE), pushes Discord + Telegram alerts with an annotated screenshot on BUY/SELL. You execute the trade manually in TradeLocker.
**Current phase: Faza 1 — notification-only.** No auto-execution until prop firm TOS has been audited (see `docs/phase2-prop-firm-audit.md`).
No auto-execution. Faza 2 (auto-execute) is blocked on prop-firm TOS auditsee `docs/phase2-prop-firm-audit.md`.
---
## Project structure
## Project layout
```
atm/
├── pyproject.toml
├── configs/ # calibration configs (YYYY-MM-DD-HHMM.toml + current.txt)
├── logs/ # audit JSONL + dead-letter queue
├── samples/ # screenshots for dry-run validation
└── src/atm/
── config.py # Config dataclass + loader
├── detector.py # screenshot → color → state machine
├── state_machine.py
├── vision.py # color matching helpers
├── levels.py # SL/TP pixel-to-price
├── calibrate.py # Tkinter calibration wizard
├── labeler.py # Tkinter sample labeler
├── journal.py # trade journal (JSONL)
├── report.py # weekly performance report
├── audit.py # structured audit log
├── canary.py # layout drift detection
├── dryrun.py # replay saved screenshots
├── notifier/
│ ├── discord.py
── telegram.py
── main.py # unified CLI
├── configs/ # calibration outputs + current.txt marker
├── logs/
│ ├── YYYY-MM-DD.jsonl # per-cycle audit log, rotates at local midnight
│ ├── dead_letter.jsonl # alerts that failed after retries
│ ├── fires/ # annotated screenshots, one per BUY/SELL trigger
── calibrate_capture_*.png / debug_*.png # gitignored debug artifacts
├── samples/ # full frames saved automatically on each colour change
├── src/atm/ # package
├── config.py # frozen dataclass + TOML loader
├── vision.py # ROI crop, phash, pixelprice, Hough, connected-components
├── state_machine.py # 5-state phased FSM, per-direction lockout
├── detector.py # capture → crop → find rightmost dot → classify → debounce
├── canary.py # layout phash drift watchdog with pause-file gating
├── levels.py # Phase-B SL/TP line extraction
├── notifier/ # FanoutNotifier + Discord webhook + Telegram bot
├── audit.py # line-buffered JSONL, daily rotation
├── calibrate.py # Tk wizard (region-select + click-sample)
├── labeler.py # Tk UI → labels.json
│ ├── dryrun.py # replay corpus, precision/recall gate
── journal.py # trade entries
── report.py # weekly R-multiple PnL
│ └── main.py # unified CLI
├── tests/ # 105 pytest cases
└── TODOS.md # P1/P2/P3 backlog, Faza 2 items
```
---
## Setup
## Install
### Prerequisites
- Python 3.11+
- Windows 10/11 (required for live capture; dry-run works on any OS)
- TradeStation running with a 3-minute DIA or GLD chart open
### Install
Python 3.11+ required. Clone, then:
```bash
# Clone (internal repo)
git clone git@gitea.romfast.ro:romfast/atm.git
cd atm
# Create venv and install
python -m venv .venv
.venv\Scripts\activate # Windows
pip install -e .
# Windows-only extras (screen capture, window detection)
pip install -e ".[windows]"
pip install -e ".[windows]" # Windows: live capture + window focus
pip install -e . # Linux / macOS: dev / dryrun only (no live)
atm --help
```
### Environment variables (notifiers)
`[windows]` pulls `mss`, `pygetwindow`, `pywin32`.
Create a `.env` file or set these in your shell before running:
---
## Calibration
One-time per chart layout. Run on the machine that will do live capture.
```powershell
atm calibrate # 3s default countdown; use --delay 10 if you want more time
```
Flow:
1. Dialog: substring of the chart window title (e.g. `TradeStation` or `DIA`). Stored in config for later auto-focus.
2. **"Ready?" message** → click OK → 3s countdown in terminal. Alt-tab TradeStation to the foreground and minimize anything covering it.
3. Full-desktop screenshot is captured and shown in a scaled Tk window.
4. **Drag a rectangle** over the chart (include the M2D MAPS strip). Enter = confirm. Esc = cancel.
5. Step-by-step clicks on the selected region:
- M2D MAPS strip: top-left + bottom-right corners
- One click on each of: turquoise, yellow, dark_green, dark_red, light_green, light_red, gray dot + chart background (8 total — "Skip" if a colour isn't currently visible)
- Chart area: top-left + bottom-right (for Phase-B line detection)
- Two known price levels on the y-axis (pixel y → enter price)
- Canary region: top-left + bottom-right on a stable UI element (axis label, title bar)
6. **Save** → writes `configs/YYYY-MM-DD-HHMM.toml` + marker `configs/current.txt`. Pulls Discord/Telegram creds from env (`ATM_DISCORD_URL`, `ATM_TG_TOKEN`, `ATM_TG_CHAT`) if set; otherwise `REPLACE_ME` placeholders — edit the TOML manually.
What gets written:
- `chart_window_region = {x, y, w, h}` — virtual-desktop absolute rectangle. Runtime capture crops the same box, so the window must stay in that position.
- `dot_roi`, `chart_roi`, `canary.roi` — coords relative to the selected region.
- Per-colour RGB (sampled via saturation-snap within 15px of the click, mean of 5x5 around the snapped centre).
- `y_axis` linear-interp pair.
- `canary.baseline_phash` of the canary ROI.
Sampling tips:
- Click colours that are **actually present** in the current dot history. If a colour isn't visible, skip it — `atm dryrun` will tell you if the skipped value doesn't match real dots.
- Default tolerance is 60 for dot colours, 25 for background. Tighten via TOML after dryrun if misclassifications creep in.
---
## Smoke-test after calibration
```powershell
atm debug --delay 5
```
Captures one frame. Saves `logs/debug_full_<ts>.png`, `logs/debug_dot_roi_<ts>.png`, `logs/debug_annotated_<ts>.png`. Prints:
```
ATM_DISCORD_WEBHOOK=https://discord.com/api/webhooks/...
ATM_TELEGRAM_TOKEN=123456789:AABBcc...
ATM_TELEGRAM_CHAT_ID=-100123456789
window_found: True
dot_found: True
rgb: (114, 114, 114)
classified: gray distance=24 confidence=0.79
accepted: True color=gray
```
Open the annotated PNG: yellow rectangle = `dot_roi`, red circle = detected dot. The circle should land on the ACTUAL rightmost colored dot in the M2D MAPS strip. If not:
- Circle mid-strip → wrong window under the capture region (bring TradeStation to front).
- Circle on a non-dot UI element → `dot_roi` boundaries capture too much; recalibrate narrower.
- `color=None` + `UNKNOWN` → tolerances too tight OR sampled RGBs don't match real dots; recalibrate clicking on actual dots.
---
## Live run
```powershell
# Today's session 16:3023:00 Romania local
atm run --start-at 16:30 --stop-at 23:00
# Indefinite
atm run
# Fixed duration (hours)
atm run --duration 2
# Linux / headless smoke (reads samples/*.png in a loop)
atm run --capture-stub --duration 0.05
```
Startup sequence:
1. Wall-clock wait until `--start-at` (if set).
2. `pygetwindow.activate()` on the first window matching `cfg.window_title` — brings TradeStation to the foreground automatically (restores if minimised).
3. 5s countdown (`--startup-delay`).
4. Capture first frame + canary check. Status (`drift=X/Y` or `capture_failed`) is included in the startup ping.
5. **"ATM started" ping** on Discord + Telegram.
6. Main loop: every `loop_interval_s` (default 5s) — capture → canary → detect → state machine → maybe notify → maybe Phase-B.
7. At `--stop-at` (or `--duration`): **"ATM stopped" ping**, then exit.
Per-cycle behaviour:
- Canary drift → auto-pause (logs `paused`, skips detection). Clear by running `atm run` again with the pause-file removed.
- Detector reports UNKNOWN → stays in current state (logged as `noise`).
- Colour change → full frame saved to `samples/YYYYMMDD_HHMMSS_<color>.png` (for corpus).
- FIRE (BUY/SELL, not locked) → annotated PNG saved to `logs/fires/`, attached to the alert, `LevelsExtractor` armed.
- Phase-B complete → "Levels SL=… TP1=… TP2=…" push.
- Heartbeat every `heartbeat_min` minutes.
Keep PowerShell minimized during the session so it doesn't cover TradeStation.
---
## After the session
```powershell
atm label samples # Tk UI — label each saved frame with true dot colour
atm dryrun samples # replay through detector + FSM; exits 0 if precision=100%, recall>=95%
```
If the gate fails, tune per-colour `tolerance` in `configs/<current>.toml`, or recalibrate colour samples that didn't match. Re-run `atm dryrun` until it passes. Only then do you trust live signals.
Trade record-keeping:
```powershell
atm journal # interactive entry after a real trade
atm report --week 2026-16 # weekly win rate + R PnL + slippage
```
---
## Calibration workflow
## DPI / multi-monitor notes
Calibration maps the TradeStation window layout to the config that drives detection. **Redo calibration whenever you resize the chart window, change DPI, or switch monitors.**
### Step-by-step
1. **Open TradeStation** with the 3-minute DIA chart. Maximise or snap to a fixed position. Do not resize it during the session.
2. **Run the calibration wizard:**
```bash
atm calibrate
```
3. **Pick window title** — type the exact string that appears in the TradeStation title bar (e.g. `DIA - 3 Min`). The bot uses this to locate the window via `pygetwindow`.
4. **Mark the dot ROI** — a screenshot of the current window is shown. Click the top-left corner and the bottom-right corner of the M2D MAPS dot strip to define the region of interest.
5. **Sample dot colours** — for each of the 7 dot colours (turquoise, yellow, dark green, dark red, light green, light red, gray), click a representative dot in the screenshot. The wizard records the RGB value and sets an initial tolerance of 30.
6. **Set y-axis calibration points** — click on two price levels that appear as visible horizontal gridlines on the chart, then type the corresponding price for each. This calibrates the pixel-y → price mapping for SL/TP extraction.
7. **Select canary region** — drag a small rectangle over a stable, unchanging part of the chart border (title bar strip works well). This becomes the baseline for canary drift detection.
8. **Save** — the wizard writes `configs/YYYY-MM-DD-HHMM.toml` and updates `configs/current.txt`. Future runs load this config automatically.
### Verify calibration
```bash
atm dryrun --dir samples/
```
Review the dry-run output. Every sample should classify to the expected dot colour. If misclassifications appear, re-run the calibration wizard or adjust tolerances manually in the TOML file.
---
## Per-session operating checklist
Before each trading window (NY open 16:30 RO, NY close 21:00 RO):
- [ ] TradeStation open on 3-minute DIA chart, window not minimised
- [ ] Chart window is in the same position/size as when calibrated
- [ ] TradeLocker open in browser, instrument US30 loaded, position sizing ready
- [ ] Telegram / Discord notification channel visible on mobile or second screen
- [ ] Run `atm canary-check` — confirm no drift alert before starting the bot
- [ ] Start the monitor: `atm run`
- [ ] After the session: `atm journal add` to record trade outcome (or leave `outcome=open` to fill later)
- [ ] At week end: `atm report --week YYYY-WW` to review win rate and PnL in R
---
## DPI and multi-monitor notes
- **High-DPI displays:** Windows DPI scaling can shift pixel coordinates. Set TradeStation to "System (Enhanced)" DPI compatibility mode (right-click EXE → Properties → Compatibility → Change high DPI settings) OR set Python to DPI-unaware via `SetProcessDPIAware()` in the manifest. The calibration wizard and capture code both call `SetProcessDPIAware()` on start.
- **Multiple monitors:** `mss` captures the monitor that contains the target window. The ROI offsets in the config are relative to the window's own top-left, so moving the window between sessions (but not resizing) is usually safe. Moving to a different monitor with a different DPI **requires recalibration**.
- **Virtual desktops / remote desktop:** Screen capture via `mss` does not work through RDP (the window reports on-screen but the pixel data is black). Run the bot locally on the same physical machine as TradeStation.
- Calibration region is virtual-desktop-absolute; runtime capture uses the same rectangle. **Don't move the TradeStation window** after calibrating. Canary will catch drift and pause automatically.
- Changing DPI scaling or moving to a different monitor with different DPI → recalibrate.
- RDP / virtual desktops: `mss` can return black frames over RDP. Run locally on the same physical machine as TradeStation.
---
## Troubleshooting
### Window not found
| Symptom | Likely cause | Fix |
|---|---|---|
| `capture_failed` in startup ping | `chart_window_region` references coords off-screen (different monitor layout) | Recalibrate. |
| Startup canary `drift=X/8` with X >> 8 | Wrong window is in the capture region | Make sure TradeStation is the window at `cfg.chart_window_region`. Relaunch. |
| `WARN: no window contains 'xxx'` at startup | `cfg.window_title` substring matches nothing | Edit `window_title` in TOML to a substring that's unique to TradeStation. |
| No alerts even after trigger ought to fire | Check `logs/YYYY-MM-DD.jsonl` for `event=tick` entries — are colours accepted? Is `trigger` ever set? | If always UNKNOWN → tolerances too tight. If `trigger` but `locked=true` → lockout from prior fire, normal. |
| Discord OK, Telegram silent (or vice versa) | `logs/dead_letter.jsonl` contains failed alerts with error | Fix credentials in TOML, restart. |
| Debug circle on mid-strip instead of right edge | Anti-aliasing bridges dots in the mask | Already fixed via erosion+connected-components — ensure `git pull` is current. |
| Wizard window is tiny / image not visible | Tk geometry default on Windows | Already fixed — `git pull`. Image is scaled to fit screen. |
---
## Windows Task Scheduler (production)
For hands-off daily runs surviving reboots:
1. Task Scheduler → Create Task → name `ATM M2D Monitor`
2. **General**: "Run only when user is logged on", "Run with highest privileges"
3. **Triggers**: New → Daily, Start `16:30`
4. **Actions**: New → Program `C:\path\to\python.exe`, Arguments `-m atm run --stop-at 23:00`, Start in `D:\PROIECTE\atm`
5. **Conditions**: uncheck "Start only if AC power" (if laptop)
6. **Settings**: "If task runs longer than 7 hours → stop"
Click-right → Run to test manually. Manual DST-change check twice a year (Mar / Oct first week).
---
## Quick command reference
```
WindowNotFoundError: No window matching 'DIA - 3 Min'
atm calibrate [--screenshot PATH] [--delay SEC] # Tk wizard
atm debug [--delay SEC] # one-shot capture + detect
atm label SAMPLES_DIR # Tk labeling
atm dryrun SAMPLES_DIR # corpus gate
atm run [--duration H] [--start-at HH:MM] [--stop-at HH:MM] [--startup-delay SEC] [--capture-stub]
atm journal [--file PATH] # interactive trade entry
atm report [--week YYYY-WW] [--file PATH] # weekly summary
```
**Causes and fixes:**
- TradeStation is minimised — restore it to a visible state.
- The window title has changed — re-run `atm calibrate` and provide the exact current title string. Copy-paste from the title bar.
- DPI scaling changed the reported title encoding — confirm the title in `pygetwindow.getWindowsWithTitle("")` output.
### Canary drift alert
```
CanaryDriftAlert: phash distance 12 > threshold 8
```
The chart layout has shifted (e.g., chart scroll, zoom change, indicator redraw). **Do not trade until this is resolved.**
1. Check TradeStation — scroll or zoom may have shifted the DOT ROI out of frame.
2. Reset the chart to the calibrated state (same zoom/scroll as during calibration).
3. If the layout change was intentional, re-run calibration.
4. To suppress a single false alarm without recalibrating, run `atm canary-reset` which re-samples the canary region baseline from the current screenshot.
### Low confidence warnings
```
LowConfidence: cycle 5 — best match 'gray' dist=0.27 (threshold 0.20)
```
The sampled pixel is near the edge of a colour tolerance zone. Usually caused by:
- Screenshot timing during a dot colour transition (rare on 3-min chart).
- Chart re-render artifact (scaling seam at ROI boundary).
If this persists for more than 3 consecutive cycles, an alert is sent automatically. Verify the dot strip is fully visible and not clipped by another window.
### Notification not received
1. Check `logs/audit.jsonl` for the last cycle — look for `"notification_sent": false` and the `reason` field.
2. Verify webhook/token: `atm test-notify`.
3. Check network connectivity from the Windows machine to Discord/Telegram endpoints.
Exit code: `atm dryrun` exits 0 if gate passes, 1 otherwise. Other commands follow standard convention.