artemium/tvsignals-to-tg
1
0
Fork 0
mirror of https://github.com/artemium428/tvsignals-to-tg.git synced 2026-09-15 17:16:21 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Replace TradingView polling with a local LTF/FVG scanner that posts Telegram cards and Heryon webhooks.

Browse source 
Per-strategy Heryon accounts, reversal captions with previous-trade path on charts, and Telegram replies chained by ticker.

Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
Artemii Peretiachenko 2026-08-30 20:37:16 +02:00
parent c2571f4558
commit cdbda8fea3
24 changed files with 2125 additions and 196 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

19
.env.example
View file

@ -1,7 +1,26 @@
TELEGRAM_BOT_TOKEN=123456:ABC-DEF
TELEGRAM_CHAT_ID=-1001234567890
# Default forum topic when webhook URL has no /{thread_id} segment
# and watchlist strategy telegram_thread_id is null
TELEGRAM_MESSAGE_THREAD_ID=1
WEBHOOK_SECRET=change-me-to-a-long-random-string
HOST=0.0.0.0
PORT=8000
# Local scanner (replaces TradingView alerts)
SCANNER_ENABLED=true
SCANNER_POLL_SECONDS=20
WATCHLIST_PATH=watchlist.yaml
SCANNER_STATE_PATH=data/scanner.db
RISK_USD=40
# Heryon / Geryon — same JSON as Pine alertcondition (leave URL empty to skip)
GERYON_WEBHOOK_URL=
GERYON_ORDER_TYPE=bbo
# Per-strategy account/secret (fallback: GERYON_ACCOUNT_ID / GERYON_SECRET)
GERYON_ACCOUNT_ID_LTF=
GERYON_SECRET_LTF=
GERYON_ACCOUNT_ID_FVG=
GERYON_SECRET_FVG=
GERYON_ACCOUNT_ID=admin1
GERYON_SECRET=

2
.gitignore vendored
View file

@ -8,3 +8,5 @@ __pycache__/
.pytest_cache/
.mypy_cache/
.DS_Store
data/*.db
data/*.db-*

3
Dockerfile
View file

@ -12,6 +12,9 @@ COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt
COPY app ./app
COPY watchlist.yaml .
RUN mkdir -p /app/data
ENV PYTHONUNBUFFERED=1
ENV MPLBACKEND=Agg

159
README.md
View file

@ -1,10 +1,12 @@
# TradingView → Telegram setup service
# Signals → Telegram + Heryon
Accepts TradingView webhook alerts, renders a Binance Futures candlestick setup chart, and posts photo + caption into a Telegram forum topic.
Runs the 15m LTF and 6h FVG indicators on a Binance Futures watchlist, posts setup charts to a Telegram forum topic, and forwards the same Heryon webhook JSON that TradingView `alertcondition` used to send.
Inbound `POST /h/...` is still accepted for manual/debug payloads. Those requests are **not** forwarded to Heryon (avoids doubles if TradingView is still on).
## Quick start (Docker / VPS)
1. Copy env and fill Telegram values:
1. Copy env and fill Telegram + Heryon values:
```bash
cp .env.example .env
@ -14,33 +16,33 @@ cp .env.example .env
TELEGRAM_BOT_TOKEN=...
TELEGRAM_CHAT_ID=-100...
TELEGRAM_MESSAGE_THREAD_ID=...
HOST=0.0.0.0
PORT=8000
WEBHOOK_SECRET=...
GERYON_WEBHOOK_URL=http://geryon:PORT/path # empty = Telegram only
GERYON_ORDER_TYPE=bbo
GERYON_ACCOUNT_ID_LTF=...
GERYON_SECRET_LTF=...
GERYON_ACCOUNT_ID_FVG=...
GERYON_SECRET_FVG=...
```
2. Build and run:
2. Edit [`watchlist.yaml`](watchlist.yaml): symbols and optional `telegram_thread_id` per strategy (15m vs 6h topics).
3. Build and run:
```bash
mkdir -p data
docker compose up -d --build
```
3. Health check:
4. Health check:
```bash
curl http://127.0.0.1:8000/health
```
4. Put HTTPS in front (nginx/Caddy) and point TradingView webhook to:
On first start the scanner **primes** the last closed bar per symbol and does not replay history. The next closed 15m / 6h bar can fire a signal.
`https://your-domain/h/<WEBHOOK_SECRET>`
or, for a specific forum topic:
`https://your-domain/h/<WEBHOOK_SECRET>/<TELEGRAM_MESSAGE_THREAD_ID>`
Same alert message body; different URLs → different topics (e.g. one alert per timeframe). Without a thread segment, the env `TELEGRAM_MESSAGE_THREAD_ID` is used.
Bot must be added to the group/forum and allowed to post in the target topic.
After it is stable, turn off the TradingView alerts that used to hit this service and Heryon.
## Local run (without Docker)
@ -49,18 +51,53 @@ python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env # fill values
mkdir -p data
uvicorn app.main:app --host 0.0.0.0 --port 8000 --reload
```
## TradingView alert JSON
## Scanner
The webhook accepts a **single-line or pretty-printed JSON** body, including TradingView’s common `Content-Type: text/plain`.
| Strategy | Timeframe | Risk | Heryon TPs |
|---|---|---|---|
| `ltf` (`428th_v16_LTF`) | 15m + 6h filters | SL = 16h low/high; TP 1.75 / 2 / 4 R | `tp1/2/3_fix` 30% / 30% / 40%, `sl_to_bk=tp1` |
| `fvg` (`428th_v16`) | 6h | live SL 2R from ATR×1 stick; TP1 2R | `tp1_fix` 30%, empty tp2/tp3 (leftover until reverse) |
Static example (alert **Webhook message** body) — primary setup (`signal_sequence: 1`):
- Poll interval: `SCANNER_POLL_SECONDS` (default 20). Only **closed** candles are evaluated.
- State: `data/scanner.db` (last bar + Heryon nonces). Restart does not re-send the primed bar.
- Same-side lock matches live Pine (`open_side` on LTF, `use_lock_until_sl` on FVG).
- Tickers in Telegram / Heryon: `{SYMBOL}.P` (e.g. `BTCUSDT.P`).
- Size: `round(RISK_USD / abs(entry−sl)/entry)` with `RISK_USD=40`.
Set `SCANNER_ENABLED=false` to run as a webhook-only Telegram bridge.
### Heryon payload (LTF)
```json
{
"ticker": "{{ticker}}",
"action": "buy",
"ticker": "BTCUSDT.P",
"account_id": "admin1",
"order_type": "bbo",
"position_size_usd": "5000",
"stop_loss": "...",
"tp1": "...", "tp1_fix": "30%",
"tp2": "...", "tp2_fix": "30%",
"tp3": "...", "tp3_fix": "40%",
"sl_to_bk": "tp1",
"nonce": "BTCUSDT.P-limit-long-1721736000",
"secret": "..."
}
```
FVG omits `tp2` / `tp3` and their `_fix` fields. Empty string keys are not sent.
## Manual / debug webhook
`POST /h/<WEBHOOK_SECRET>` or `POST /h/<WEBHOOK_SECRET>/<thread_id>` still accepts the Telegram JSON body (including TradingView `text/plain`). Telegram only — no Heryon forward.
```json
{
"ticker": "BTCUSDT.P",
"action": "long",
"entry_price": "65034.7",
"current_price": "65034.7",
@ -74,84 +111,36 @@ Static example (alert **Webhook message** body) — primary setup (`signal_seque
}
```
Control update (`signal_sequence` > 1) — Pine freezes `entry_price` / SL / TPs / `signal_time` from seq 1 and sends live `current_price`:
`take_profit_2_price` / `take_profit_3_price` are optional (omit or `""` for TP1-only FVG-style setups).
```json
{
"ticker": "{{ticker}}",
"action": "short",
"entry_price": "65034.7",
"current_price": "64000.1",
"stop_loss_price": "65896.8",
"take_profit_1_price": "63904.9",
"take_profit_2_price": "62781.6",
"take_profit_3_price": "61635.3",
"visual_timeframe": "15",
"signal_sequence": 2,
"signal_time": 1721736000
}
```
### PineScript `alert()` (recommended)
Build the JSON inside `alert()`. A continuous one-line string is fine.
In the TradingView alert dialog:
- Webhook URL: `https://your-domain/h/<WEBHOOK_SECRET>` or `https://your-domain/h/<WEBHOOK_SECRET>/<thread_id>` (per-topic; same message body)
- Message: only `{{alert_message}}` (do not paste a second JSON next to it)
On **seq == 1**: store `entry_price = close`, freeze SL/TPs, and `signal_time = time / 1000` (bar open, unix seconds). On **seq > 1**: keep those frozen fields; only refresh `current_price` (= live `close`). Example shape:
```pinescript
alert('{"ticker":"' + syminfo.ticker + '","action":"long","entry_price":"' + str.tostring(long_entry_price, format.mintick) + '","current_price":"' + str.tostring(close, format.mintick) + '","stop_loss_price":"' + str.tostring(long_sl_price, format.mintick) + '","take_profit_1_price":"' + str.tostring(long_tp1_price, format.mintick) + '","take_profit_2_price":"' + str.tostring(long_tp2_price, format.mintick) + '","take_profit_3_price":"' + str.tostring(long_tp3_price, format.mintick) + '","visual_timeframe":"' + timeframe.period + '","signal_sequence":' + str.tostring(buyCount) + ',"signal_time":' + str.tostring(signal_buy_time) + '}', alert.freq_once_per_bar_close)
```
(Same idea for shorts with `short_entry_price` / `sellCount` / `signal_sell_time`.) Caption emoji/labels are built by the service from `action` + `signal_sequence` — do not put them in the webhook JSON.
Field notes:
Caption emoji/labels are built from `action` + `signal_sequence`.
| Field | Description |
|---|---|
| `ticker` | any common TV form (`BTCUSDT.P`, `BTCUSDT`, `BINANCE:ETHUSDT`, `BTC/USDT`, …) → normalized to Binance Futures symbol for the chart; caption keeps the original |
| `ticker` | TV form (`BTCUSDT.P`, `BINANCE:ETHUSDT`, …) → Binance Futures for the chart; caption keeps the original |
| `action` | `long` or `short` |
| `entry_price` | trade entry from seq 1 (equals `current_price` on primary signal) |
| `current_price` | live price (`close` at alert time) |
| `*_price` | strings with your display precision |
| `visual_timeframe` | `1`, `3`, `5`, `15`, `30`, `60`, `120`, `240`, `D`, `W` (also `15m`, `1h`, …) |
| `signal_sequence` | `1` = primary setup; `>1` = control update of that trade |
| `signal_time` | unix seconds of the seq-1 bar open (UTC); chart draws Entry/SL/TP zones from that candle |
| `*_price` | strings; TP2/TP3 optional |
| `visual_timeframe` | `15`, `360`, `15m`, `6h`, … |
| `signal_sequence` | `1` = setup; `>1` = control update |
| `signal_time` | unix seconds of the seq-1 bar open (UTC) |
## Caption format
**seq `1` (setup):**
**seq `1` (setup):** `💚 Buy` / `💔 Sell` — Price / SL (risk %) / TP1 and TP2/TP3 when present.
- long: `BTCUSDT.P 💚 Buy`
- short: `BTCUSDT.P 💔 Sell`
- body: `Price` / `SL (risk %)` / `TP1–3`
**seq `>1` (control):** `🌱 Buy Seq: N` / `🥀 Sell Seq: N` — Entry / live Price / `Current profit`.
**seq `>1` (control):**
- long: `BTCUSDT.P 🌱 Buy Seq: N`
- short: `BTCUSDT.P 🥀 Sell Seq: N`
- body: `Entry price` (from seq 1) / live `Price` / `Current profit: +1.6% (RR 1:1.2)`
- no new SL/TP lines in the caption
Prices are shown with `$` and thousand spaces (`65034.7` → `$65 034.7`).
For seq 1, SL includes distance from entry: `SL: $63 904.9 (-2.37%)` (risk %, negative for both long and short).
For seq >1, profit % is signed vs entry; RR is `|price−entry| / |entry−SL|` with the same sign as profit.
Prices: `$` and thousand spaces (`65034.7` → `$65 034.7`).
## Behavior
1. Validate secret + payload, then immediately respond `200` (`{"ok": true, "accepted": true}`) so TradingView does not time out
2. In the background: fetch ~90 klines from Binance USDT-M Futures (public, no API key)
3. Render PNG: candles + Entry / SL / TP1–3 from the payload, starting at the `signal_time` candle (seq `>1` reuses frozen seq-1 levels/time and also marks live `Price`)
4. `sendPhoto` to `TELEGRAM_CHAT_ID` topic from URL path or env `TELEGRAM_MESSAGE_THREAD_ID`
5. If chart/klines fail → text-only `sendMessage` fallback
6. If Telegram fails → logged only (HTTP already returned `200`)
1. Scanner (if enabled): closed bar → indicator → Telegram chart + Heryon JSON
2. Inbound webhook: validate secret, `200` immediately, background Telegram only
3. Chart: ~90 USDT-M klines; on failure, text-only Telegram
4. Telegram or Heryon failure is logged (webhook HTTP already returned `200`)
## Endpoints
- `GET /health` → `{"status":"ok"}`
- `POST /h/<WEBHOOK_SECRET>` → signal payload above; topic from env `TELEGRAM_MESSAGE_THREAD_ID` (wrong/missing secret → `404`)
- `POST /h/<WEBHOOK_SECRET>/<thread_id>` → same payload; topic from path (`thread_id` must be `>= 1`)
- `POST /h/<WEBHOOK_SECRET>` → debug payload; topic from `TELEGRAM_MESSAGE_THREAD_ID`
- `POST /h/<WEBHOOK_SECRET>/<thread_id>` → same; topic from path (`thread_id` >= 1)

117
app/binance.py
View file

@ -8,7 +8,25 @@ import pandas as pd
logger = logging.getLogger(__name__)
BINANCE_FUTURES_KLINES_URL = "https://fapi.binance.com/fapi/v1/klines"
BINANCE_EXCHANGE_INFO_URL = "https://fapi.binance.com/fapi/v1/exchangeInfo"
DEFAULT_LIMIT = 90
# 6h charts use 1/3 of the default window so price action looks closer.
CHART_LIMIT_BY_INTERVAL: dict[str, int] = {
"15m": 135,
"6h": 75,
}
def chart_kline_limit(interval: str) -> int:
return CHART_LIMIT_BY_INTERVAL.get(interval, DEFAULT_LIMIT)
def chart_right_pad(interval: str) -> int:
"""Empty candles to the right; scale with the visible window."""
if interval == "6h":
return 12
return 15
# TradingView-style timeframe → Binance Futures interval
TIMEFRAME_MAP: dict[str, str] = {
@ -95,6 +113,45 @@ def to_binance_interval(visual_timeframe: str) -> str:
return interval
_INTERVAL_DELTA: dict[str, pd.Timedelta] = {
"1m": pd.Timedelta(minutes=1),
"3m": pd.Timedelta(minutes=3),
"5m": pd.Timedelta(minutes=5),
"15m": pd.Timedelta(minutes=15),
"30m": pd.Timedelta(minutes=30),
"1h": pd.Timedelta(hours=1),
"2h": pd.Timedelta(hours=2),
"4h": pd.Timedelta(hours=4),
"6h": pd.Timedelta(hours=6),
"8h": pd.Timedelta(hours=8),
"12h": pd.Timedelta(hours=12),
"1d": pd.Timedelta(days=1),
"1w": pd.Timedelta(weeks=1),
}
def interval_timedelta(interval: str) -> pd.Timedelta:
key = interval if interval in _INTERVAL_DELTA else to_binance_interval(interval)
delta = _INTERVAL_DELTA.get(key)
if delta is None:
raise ValueError(f"Unsupported interval: {interval!r}")
return delta
def drop_forming_candles(df: pd.DataFrame, interval: str) -> pd.DataFrame:
"""Drop the in-progress candle (open + interval > now)."""
if df.empty:
return df
delta = interval_timedelta(interval)
now = pd.Timestamp.now(tz="UTC")
idx = df.index
if idx.tz is None:
idx = idx.tz_localize("UTC")
else:
idx = idx.tz_convert("UTC")
return df.loc[idx + delta <= now]
async def fetch_klines(
symbol: str,
interval: str,
@ -102,12 +159,16 @@ async def fetch_klines(
limit: int = DEFAULT_LIMIT,
end_ms: int | None = None,
timeout: float = 15.0,
closed_only: bool = False,
client: httpx.AsyncClient | None = None,
) -> pd.DataFrame:
"""Fetch OHLCV klines from Binance USDT-M Futures.
If ``end_ms`` is set, returns candles ending at/before that UTC epoch millis
(useful for historical / as-of charts).
"""
if limit > 1500:
raise ValueError("Binance klines limit is 1500")
params: dict[str, str | int] = {
"symbol": symbol,
"interval": interval,
@ -115,10 +176,16 @@ async def fetch_klines(
}
if end_ms is not None:
params["endTime"] = end_ms
async with httpx.AsyncClient(timeout=timeout) as client:
response = await client.get(BINANCE_FUTURES_KLINES_URL, params=params)
http = client or httpx.AsyncClient(timeout=timeout)
own_client = client is None
try:
response = await http.get(BINANCE_FUTURES_KLINES_URL, params=params)
response.raise_for_status()
raw = response.json()
finally:
if own_client:
await http.aclose()
if not raw:
raise ValueError(f"Empty klines for {symbol} {interval}")
@ -146,6 +213,52 @@ async def fetch_klines(
df = df.set_index("Date")[["open", "high", "low", "close", "volume"]]
df.columns = ["Open", "High", "Low", "Close", "Volume"]
df = df.dropna()
if closed_only:
df = drop_forming_candles(df, interval)
return df
if df.empty:
raise ValueError(f"No valid OHLCV rows for {symbol} {interval}")
return df
_TICK_SIZE: dict[str, float] = {}
async def load_tick_sizes(
client: httpx.AsyncClient | None = None,
*,
timeout: float = 20.0,
) -> None:
"""Cache Binance USDT-M PRICE_FILTER.tickSize per symbol (Pine mintick)."""
http = client or httpx.AsyncClient(timeout=timeout)
own = client is None
try:
response = await http.get(BINANCE_EXCHANGE_INFO_URL)
response.raise_for_status()
payload = response.json()
finally:
if own:
await http.aclose()
ticks: dict[str, float] = {}
for item in payload.get("symbols") or []:
name = str(item.get("symbol") or "")
if not name:
continue
for filt in item.get("filters") or []:
if filt.get("filterType") == "PRICE_FILTER":
raw = filt.get("tickSize")
if raw is None:
continue
tick = float(raw)
if tick > 0:
ticks[name] = tick
break
if ticks:
_TICK_SIZE.update(ticks)
logger.info("Loaded tick sizes for %s symbols", len(ticks))
def get_tick_size(symbol: str) -> float:
key = to_binance_symbol(symbol)
return _TICK_SIZE.get(key, 0.01)

156
app/chart.py
View file

@ -125,6 +125,91 @@ def _position_start_x(df: pd.DataFrame, signal_time: int | None) -> int:
return min(pos, last_x)
def _draw_previous_trade(
ax,
df: pd.DataFrame,
*,
prev_entry: float,
prev_side: str,
prev_signal_time: int,
exit_price: float,
exit_x: int,
) -> None:
"""Entry → exit of the trade that this reversal closes."""
prev_x = _position_start_x(df, prev_signal_time)
if prev_x >= exit_x:
return
profitable = (
exit_price >= prev_entry if prev_side == "long" else exit_price <= prev_entry
)
color = COLORS["up"] if profitable else COLORS["down"]
y_low = min(prev_entry, exit_price)
y_high = max(prev_entry, exit_price)
height = y_high - y_low
if height <= 0:
height = abs(prev_entry) * 1e-6 or 1e-8
ax.add_patch(
Rectangle(
(prev_x, y_low),
exit_x - prev_x,
height,
facecolor=color,
edgecolor="none",
alpha=0.04,
zorder=1,
)
)
closes = df["Close"].iloc[prev_x : exit_x + 1].astype(float)
xs = list(range(prev_x, prev_x + len(closes)))
ax.plot(
xs,
closes.to_numpy(),
color=color,
linewidth=0.9,
linestyle=(0, (3, 4)),
alpha=0.28,
zorder=5,
solid_capstyle="round",
)
ax.plot(
[prev_x, exit_x],
[prev_entry, exit_price],
color=color,
linewidth=1.05,
linestyle=(0, (4, 5)),
alpha=0.45,
zorder=6,
)
ax.scatter(
[prev_x],
[prev_entry],
s=22,
c=color,
marker="o",
zorder=7,
edgecolors="#ffffff",
linewidths=0.7,
)
ax.annotate(
"Prev",
xy=(prev_x, prev_entry),
xytext=(6, 8),
textcoords="offset points",
va="bottom",
ha="left",
fontsize=7,
color=color,
zorder=8,
bbox={
"boxstyle": "round,pad=0.2",
"facecolor": COLORS["label_bg"],
"edgecolor": color,
"linewidth": 0.6,
"alpha": 0.88,
},
)
def render_setup_chart(
df: pd.DataFrame,
*,
@ -133,18 +218,23 @@ def render_setup_chart(
entry: str,
stop_loss: str,
tp1: str,
tp2: str,
tp3: str,
timeframe: str,
tp2: str | None = None,
tp3: str | None = None,
current_price: str | None = None,
signal_time: int | None = None,
right_pad: int | None = None,
prev_entry: str | None = None,
prev_side: str | None = None,
prev_signal_time: int | None = None,
) -> bytes:
entry_p = _parse_price(entry)
sl_p = _parse_price(stop_loss)
tp1_p = _parse_price(tp1)
tp2_p = _parse_price(tp2)
tp3_p = _parse_price(tp3)
tp2_p = _parse_price(tp2) if tp2 else None
tp3_p = _parse_price(tp3) if tp3 else None
current_p = _parse_price(current_price) if current_price is not None else None
prev_entry_p = _parse_price(prev_entry) if prev_entry else None
is_long = action == "long"
reward_rgb = COLORS["reward_fill"]
@ -154,13 +244,20 @@ def render_setup_chart(
# Position tool starts at seq==1 candle; "now" is the last real candle
now_x = len(df) - 1
entry_x = _position_start_x(df, signal_time)
plot_df = _pad_right(df, RIGHT_PAD_CANDLES)
plot_df = _pad_right(df, RIGHT_PAD_CANDLES if right_pad is None else right_pad)
level_prices = [entry_p, sl_p, tp1_p, tp2_p, tp3_p]
level_prices = [entry_p, sl_p, tp1_p]
if tp2_p is not None:
level_prices.append(tp2_p)
if tp3_p is not None:
level_prices.append(tp3_p)
if current_p is not None:
level_prices.append(current_p)
if prev_entry_p is not None:
level_prices.append(prev_entry_p)
y_min = min(float(df["Low"].min()), *level_prices)
y_max = max(float(df["High"].max()), *level_prices)
hold_remainder = tp2_p is None and tp3_p is None
price_pad = (y_max - y_min) * 0.06 or y_max * 0.002
vol_scaled, vol_colors = _volume_overlay(plot_df, y_low=y_min, y_high=y_max)
@ -227,18 +324,22 @@ def render_setup_chart(
# Room for volume bars under candles
vol_floor = float(vol_scaled.dropna().min()) if vol_scaled.notna().any() else y_min
ax.set_ylim(min(y_min - price_pad, vol_floor) - price_pad * 0.3, y_max + price_pad)
ylim_lo, ylim_hi = ax.get_ylim()
remainder_end = (ylim_hi if is_long else ylim_lo) if hold_remainder else None
x_right = ax.get_xlim()[1]
zone_width = x_right - entry_x
# Levels + zones start at the entry (last real) candle, not full chart width
level_specs = [
level_specs: list[tuple[float, str, str, float]] = [
(entry_p, COLORS["entry"], "--", 1.2),
(sl_p, COLORS["sl"], "-", 1.2),
(tp1_p, COLORS["tp1"], ":", 1.0),
(tp2_p, COLORS["tp2"], ":", 1.0),
(tp3_p, COLORS["tp3"], ":", 1.0),
]
if tp2_p is not None:
level_specs.append((tp2_p, COLORS["tp2"], ":", 1.0))
if tp3_p is not None:
level_specs.append((tp3_p, COLORS["tp3"], ":", 1.0))
for price, color, ls, lw in level_specs:
ax.hlines(
price,
@ -264,12 +365,17 @@ def render_setup_chart(
)
)
# Three reward bands with decreasing opacity toward farther TPs
reward_bands = (
reward_bands: list[tuple[float, float, float]] = [
(entry_p, tp1_p, REWARD_ALPHAS[0]),
(tp1_p, tp2_p, REWARD_ALPHAS[1]),
(tp2_p, tp3_p, REWARD_ALPHAS[2]),
)
]
if tp2_p is not None:
reward_bands.append((tp1_p, tp2_p, REWARD_ALPHAS[1]))
if tp2_p is not None and tp3_p is not None:
reward_bands.append((tp2_p, tp3_p, REWARD_ALPHAS[2]))
elif tp3_p is not None:
reward_bands.append((tp1_p, tp3_p, REWARD_ALPHAS[2]))
elif remainder_end is not None:
reward_bands.append((tp1_p, remainder_end, REWARD_ALPHAS[1]))
for price_a, price_b, alpha in reward_bands:
band_low = min(price_a, price_b)
band_high = max(price_a, price_b)
@ -295,6 +401,22 @@ def render_setup_chart(
linewidths=0.7,
)
if (
prev_entry_p is not None
and prev_side is not None
and prev_signal_time is not None
and prev_signal_time > 0
):
_draw_previous_trade(
ax,
df,
prev_entry=prev_entry_p,
prev_side=prev_side,
prev_signal_time=prev_signal_time,
exit_price=entry_p,
exit_x=entry_x,
)
if current_p is not None:
ax.hlines(
current_p,
@ -321,9 +443,11 @@ def render_setup_chart(
(entry_p, f"Entry {entry}", COLORS["entry"]),
(sl_p, f"SL {stop_loss}", COLORS["sl"]),
(tp1_p, f"TP1 {tp1}", COLORS["tp1"]),
(tp2_p, f"TP2 {tp2}", COLORS["tp2"]),
(tp3_p, f"TP3 {tp3}", COLORS["tp3"]),
]
if tp2_p is not None and tp2 is not None:
labels.append((tp2_p, f"TP2 {tp2}", COLORS["tp2"]))
if tp3_p is not None and tp3 is not None:
labels.append((tp3_p, f"TP3 {tp3}", COLORS["tp3"]))
if current_p is not None and current_price is not None:
labels.append((current_p, f"Price {current_price}", COLORS["price"]))
for price, text, color in labels:

25
app/config.py
View file

@ -17,6 +17,31 @@ class Settings(BaseSettings):
host: str = "0.0.0.0"
port: int = 8000
scanner_enabled: bool = True
scanner_poll_seconds: float = 20.0
watchlist_path: str = "watchlist.yaml"
scanner_state_path: str = "data/scanner.db"
risk_usd: float = 40.0
geryon_webhook_url: str = ""
geryon_account_id: str = "admin1"
geryon_account_id_ltf: str = ""
geryon_account_id_fvg: str = ""
geryon_order_type: str = "bbo"
geryon_secret: str = ""
geryon_secret_ltf: str = ""
geryon_secret_fvg: str = ""
def geryon_account_id_for(self, strategy_id: str) -> str:
if strategy_id == "ltf":
return self.geryon_account_id_ltf or self.geryon_account_id
return self.geryon_account_id_fvg or self.geryon_account_id
def geryon_secret_for(self, strategy_id: str) -> str:
if strategy_id == "ltf":
return self.geryon_secret_ltf or self.geryon_secret
return self.geryon_secret_fvg or self.geryon_secret
@lru_cache
def get_settings() -> Settings:

45
app/formatter.py
View file

@ -101,13 +101,18 @@ def format_caption(signal: SignalPayload) -> str:
signal.stop_loss_price,
is_long=is_long,
)
return (
text = (
f"<b>{signal.ticker}</b> {label}\n"
f"\n"
f"Entry price: {entry}\n"
f"Price: {price}\n"
f"Current profit: {profit}"
)
if signal.is_reversal and signal.realized_pnl_pct is not None:
text += (
f"\n\n<i>reversal, realized PnL {signal.realized_pnl_pct:+.2f}%</i>"
)
return text
price = format_price(signal.entry_price)
sl = format_price(signal.stop_loss_price)
@ -115,16 +120,30 @@ def format_caption(signal: SignalPayload) -> str:
signal.entry_price, signal.stop_loss_price, is_long=is_long
)
tp1 = format_price(signal.take_profit_1_price)
tp2 = format_price(signal.take_profit_2_price)
tp3 = format_price(signal.take_profit_3_price)
return (
f"<b>{signal.ticker}</b> {label}\n"
f"\n"
f"Price: {price}\n"
f"SL: {sl} {sl_pct}\n"
f"\n"
f"TP1: {tp1}\n"
f"TP2: {tp2}\n"
f"TP3: {tp3}"
hold_remainder = not signal.take_profit_2_price and not signal.take_profit_3_price
lines = [
f"<b>{signal.ticker}</b> {label}\n",
f"Price: {price}",
f"SL: {sl} {sl_pct}",
"",
f"TP1+BE: {tp1}" if hold_remainder else f"TP1: {tp1}",
]
if signal.take_profit_2_price:
lines.append(f"TP2: {format_price(signal.take_profit_2_price)}")
if signal.take_profit_3_price:
lines.append(f"TP3: {format_price(signal.take_profit_3_price)}")
if hold_remainder:
lines.extend(
[
"",
"Fix 30%, hold remainder until reversal signal.",
]
)
if signal.is_reversal and signal.realized_pnl_pct is not None:
lines.extend(
[
"",
f"<i>reversal, realized PnL {signal.realized_pnl_pct:+.2f}%</i>",
]
)
return "\n".join(lines)

81
app/heryon.py Normal file
View file

@ -0,0 +1,81 @@
from __future__ import annotations
import logging
from typing import Any
import httpx
from app.binance import get_tick_size
from app.config import Settings
from app.indicators.common import IndicatorSignal, calc_size, format_px, round_to_mintick
logger = logging.getLogger(__name__)
LTF_FIX = {"tp1_fix": "30%", "tp2_fix": "30%", "tp3_fix": "40%"}
FVG_FIX = {"tp1_fix": "30%"}
def to_tv_perp_ticker(symbol: str) -> str:
base = symbol.strip().upper()
if base.endswith(".P"):
return base
return f"{base}.P"
def heryon_nonce(ticker: str, side: str, bar_open_ts: int) -> str:
limit_side = "long" if side == "long" else "short"
return f"{ticker}-limit-{limit_side}-{bar_open_ts}"
def build_heryon_payload(
signal: IndicatorSignal,
*,
symbol: str,
settings: Settings,
) -> dict[str, Any]:
ticker = to_tv_perp_ticker(symbol)
tick = get_tick_size(symbol)
action = "buy" if signal.side == "long" else "sell"
fixes = LTF_FIX if signal.strategy_id == "ltf" else FVG_FIX
entry = round_to_mintick(signal.entry, tick)
sl = round_to_mintick(signal.sl, tick)
size = calc_size(entry, sl, float(settings.risk_usd))
size_str = "" if size != size else str(int(round(size)))
payload: dict[str, Any] = {
"action": action,
"ticker": ticker,
"account_id": settings.geryon_account_id_for(signal.strategy_id),
"order_type": settings.geryon_order_type,
"position_size_usd": size_str,
"stop_loss": format_px(signal.sl, tick),
"tp1": format_px(signal.tp1, tick),
"tp1_fix": fixes.get("tp1_fix", ""),
"sl_to_bk": "tp1",
"nonce": heryon_nonce(ticker, signal.side, signal.bar_open_ts),
"secret": settings.geryon_secret_for(signal.strategy_id),
}
if signal.strategy_id == "ltf":
payload["tp2"] = format_px(signal.tp2, tick)
payload["tp2_fix"] = fixes.get("tp2_fix", "")
payload["tp3"] = format_px(signal.tp3, tick)
payload["tp3_fix"] = fixes.get("tp3_fix", "")
return {key: value for key, value in payload.items() if value != ""}
async def send_heryon(settings: Settings, payload: dict[str, Any]) -> None:
url = (settings.geryon_webhook_url or "").strip()
if not url:
logger.info("Heryon skipped (GERYON_WEBHOOK_URL empty): %s", payload.get("nonce"))
return
async with httpx.AsyncClient(timeout=20.0) as client:
response = await client.post(url, json=payload)
if response.status_code >= 400:
raise RuntimeError(
f"Heryon webhook {response.status_code}: {response.text[:500]}"
)
logger.info(
"Heryon accepted nonce=%s ticker=%s action=%s",
payload.get("nonce"),
payload.get("ticker"),
payload.get("action"),
)

11
app/indicators/__init__.py Normal file
View file

@ -0,0 +1,11 @@
from app.indicators.common import IndicatorSignal, calc_size, format_px
from app.indicators.fvg import evaluate_fvg
from app.indicators.ltf import evaluate_ltf
__all__ = [
"IndicatorSignal",
"calc_size",
"evaluate_fvg",
"evaluate_ltf",
"format_px",
]

236
app/indicators/common.py Normal file
View file

@ -0,0 +1,236 @@
from __future__ import annotations
from dataclasses import dataclass
from decimal import Decimal, ROUND_HALF_UP
from typing import Literal
import numpy as np
import pandas as pd
StrategyId = Literal["ltf", "fvg"]
Side = Literal["long", "short"]
@dataclass(frozen=True)
class IndicatorSignal:
strategy_id: StrategyId
side: Side
entry: float
close: float
sl: float
tp1: float
tp2: float | None
tp3: float | None
size_usd: float
bar_open_ts: int
visual_timeframe: str
def ohlc_frame(df: pd.DataFrame) -> pd.DataFrame:
"""Normalize OHLC frame to lowercase columns and UTC DatetimeIndex."""
out = df.copy()
out.columns = [str(c).lower() for c in out.columns]
required = {"open", "high", "low", "close"}
missing = required - set(out.columns)
if missing:
raise ValueError(f"OHLC frame missing columns: {sorted(missing)}")
if not isinstance(out.index, pd.DatetimeIndex):
raise ValueError("OHLC frame must be indexed by timestamp")
if out.index.tz is None:
out.index = out.index.tz_localize("UTC")
else:
out.index = out.index.tz_convert("UTC")
cols = ["open", "high", "low", "close"]
if "volume" in out.columns:
cols.append("volume")
return out[cols]
def epoch_ms(index: pd.DatetimeIndex) -> np.ndarray:
utc = index.tz_convert("UTC") if index.tz is not None else index.tz_localize("UTC")
return utc.asi8.astype(np.float64) / 1_000_000.0
def alma(
series: pd.Series, length: int = 500, offset: float = 0.85, sigma: float = 5.0
) -> pd.Series:
"""Arnaud Legoux Moving Average (Pine ta.alma compatible)."""
if length < 1:
raise ValueError("ALMA length must be >= 1")
m = offset * (length - 1)
s = length / sigma
idx = np.arange(length, dtype=float)
weights = np.exp(-((idx - m) ** 2) / (2 * s * s))
weights /= weights.sum()
values = series.to_numpy(dtype=float)
out = np.full(len(values), np.nan, dtype=float)
if len(values) < length:
return pd.Series(out, index=series.index)
valid = np.convolve(values, weights[::-1], mode="valid")
out[length - 1 :] = valid
nan_in_window = np.convolve(np.isnan(values).astype(float), np.ones(length), mode="valid") > 0
out[length - 1 :][nan_in_window] = np.nan
return pd.Series(out, index=series.index)
def rma(series: pd.Series, length: int) -> pd.Series:
"""Wilder's RMA (Pine ta.rma)."""
return series.ewm(alpha=1 / length, adjust=False, min_periods=length).mean()
def ha_rsi(
open_: pd.Series,
high: pd.Series,
low: pd.Series,
close: pd.Series,
length: int = 14,
) -> pd.Series:
"""RSI on Heikin-Ashi close (Pine f_ha_rsi)."""
ha_close = (open_ + high + low + close) / 4.0
delta = ha_close.diff()
up = rma(delta.clip(lower=0), length)
down = rma((-delta).clip(lower=0), length)
rs = up / down.replace(0, np.nan)
rsi = 100 - (100 / (1 + rs))
rsi = rsi.where(down != 0, 100.0)
rsi = rsi.where(up != 0, 0.0)
both_zero = (up == 0) & (down == 0)
return rsi.where(~both_zero, 100.0)
def true_range(high: pd.Series, low: pd.Series, close: pd.Series) -> pd.Series:
prev_close = close.shift(1)
return pd.concat(
[high - low, (high - prev_close).abs(), (low - prev_close).abs()],
axis=1,
).max(axis=1)
def atr(high: pd.Series, low: pd.Series, close: pd.Series, length: int) -> pd.Series:
"""Pine ta.atr(length)."""
return rma(true_range(high, low, close), length)
def compute_fractals(high: np.ndarray, low: np.ndarray, n: int = 5) -> tuple[np.ndarray, np.ndarray]:
"""Pine upFractal / downFractal flags on the confirmation bar (pivot at i - n)."""
size = len(high)
up = np.zeros(size, dtype=bool)
down = np.zeros(size, dtype=bool)
for c in range(n, size):
p = c - n
hp = high[p]
lp = low[p]
if np.isnan(hp) or np.isnan(lp):
continue
up_prefix1 = (p + 1 < size) and (high[p + 1] <= hp)
up_prefix2 = up_prefix1 and (p + 2 < size) and (high[p + 2] <= hp)
up_prefix3 = up_prefix2 and (p + 3 < size) and (high[p + 3] <= hp)
up_prefix4 = up_prefix3 and (p + 4 < size) and (high[p + 4] <= hp)
down_prefix1 = (p + 1 < size) and (low[p + 1] >= lp)
down_prefix2 = down_prefix1 and (p + 2 < size) and (low[p + 2] >= lp)
down_prefix3 = down_prefix2 and (p + 3 < size) and (low[p + 3] >= lp)
down_prefix4 = down_prefix3 and (p + 4 < size) and (low[p + 4] >= lp)
upflag_down = True
upflag0 = True
upflag1 = True
upflag2 = True
upflag3 = True
upflag4 = True
for i in range(1, n + 1):
if p - i < 0 or not (high[p - i] < hp):
upflag_down = False
if p + i >= size or not (high[p + i] < hp):
upflag0 = False
if p + i + 1 >= size or not (high[p + i + 1] < hp):
upflag1 = False
if p + i + 2 >= size or not (high[p + i + 2] < hp):
upflag2 = False
if p + i + 3 >= size or not (high[p + i + 3] < hp):
upflag3 = False
if p + i + 4 >= size or not (high[p + i + 4] < hp):
upflag4 = False
upflag1 = upflag1 and up_prefix1
upflag2 = upflag2 and up_prefix2
upflag3 = upflag3 and up_prefix3
upflag4 = upflag4 and up_prefix4
up[c] = upflag_down and (upflag0 or upflag1 or upflag2 or upflag3 or upflag4)
downflag_down = True
downflag0 = True
downflag1 = True
downflag2 = True
downflag3 = True
downflag4 = True
for i in range(1, n + 1):
if p - i < 0 or not (low[p - i] > lp):
downflag_down = False
if p + i >= size or not (low[p + i] > lp):
downflag0 = False
if p + i + 1 >= size or not (low[p + i + 1] > lp):
downflag1 = False
if p + i + 2 >= size or not (low[p + i + 2] > lp):
downflag2 = False
if p + i + 3 >= size or not (low[p + i + 3] > lp):
downflag3 = False
if p + i + 4 >= size or not (low[p + i + 4] > lp):
downflag4 = False
downflag1 = downflag1 and down_prefix1
downflag2 = downflag2 and down_prefix2
downflag3 = downflag3 and down_prefix3
downflag4 = downflag4 and down_prefix4
down[c] = downflag_down and (downflag0 or downflag1 or downflag2 or downflag3 or downflag4)
return up, down
def tick_decimals(tick: float) -> int:
exponent = Decimal(str(tick)).normalize().as_tuple().exponent
if isinstance(exponent, int):
return max(0, -exponent)
return 0
def round_to_mintick(price: float, tick: float) -> float:
"""Pine math.round_to_mintick (half-up to exchange tick)."""
step = Decimal(str(tick))
if step <= 0:
return price
quantized = (Decimal(str(price)) / step).quantize(Decimal("1"), rounding=ROUND_HALF_UP)
return float(quantized * step)
def realized_pnl_pct(prev_side: str, prev_entry: float, exit_price: float) -> float:
"""Signed percent from previous entry to exit; long/short from the closed side."""
if prev_entry == 0:
raise ValueError("entry price is zero")
if prev_side == "long":
return (exit_price - prev_entry) / prev_entry * 100
return (prev_entry - exit_price) / prev_entry * 100
def calc_size(entry: float, sl: float, risk_usd: float) -> float:
"""Pine calc_size: round(risk / abs(entry-sl)/entry)."""
if entry == 0 or np.isnan(entry) or np.isnan(sl):
return float("nan")
stop_pct = abs(entry - sl) / entry
if stop_pct <= 0:
return float("nan")
return float(round(risk_usd / stop_pct))
def format_px(value: float | None, tick: float | None = None) -> str:
if value is None:
return ""
number = float(value)
if np.isnan(number):
return ""
if tick is not None and tick > 0:
rounded = round_to_mintick(number, tick)
return f"{rounded:.{tick_decimals(tick)}f}"
if abs(number) >= 1000:
text = f"{number:.4f}"
elif abs(number) >= 1:
text = f"{number:.6f}"
else:
text = f"{number:.8f}"
return text.rstrip("0").rstrip(".")

364
app/indicators/fvg.py Normal file
View file

@ -0,0 +1,364 @@
from __future__ import annotations
from dataclasses import dataclass
import numpy as np
import pandas as pd
from app.indicators.common import (
IndicatorSignal,
atr,
calc_size,
ohlc_frame,
)
@dataclass(frozen=True)
class FvgParams:
engulf_cover: float = 0.85
breakout_n: int = 10
max_bars: int = 300
use_lock_until_sl: bool = True
use_energy: bool = False
ss_fast_len: int = 20
ss_slow_len: int = 50
energy_k: float = 0.15
atr_len: int = 14
sl_atr_n: float = 1.0
sl_live_r: float = 2.0
tp1_rr: float = 2.0
tp2_rr: float = 0.0
tp3_rr: float = 0.0
risk_usd: float = 40.0
visual_timeframe: str = "360"
def _calc_levels(
entry: float,
ind_sl: float,
is_long: bool,
sl_live_r: float,
tp1_rr: float,
tp2_rr: float,
tp3_rr: float,
) -> tuple[float, float | None, float | None, float | None]:
r = abs(entry - ind_sl)
live_sl = entry - r * sl_live_r if is_long else entry + r * sl_live_r
def _tp(rr: float) -> float | None:
if rr <= 0:
return None
return entry + r * rr if is_long else entry - r * rr
return live_sl, _tp(tp1_rr), _tp(tp2_rr), _tp(tp3_rr)
def evaluate_fvg(df_ohlc: pd.DataFrame, params: FvgParams | None = None) -> pd.DataFrame:
"""FVG + engulf/breakout matching live Pine (lock until SL, same-dir supersede)."""
p = params or FvgParams()
df = ohlc_frame(df_ohlc)
n = len(df)
if n == 0:
return df
open_ = df["open"].to_numpy(dtype=float)
high = df["high"].to_numpy(dtype=float)
low = df["low"].to_numpy(dtype=float)
close = df["close"].to_numpy(dtype=float)
atr_s = atr(df["high"], df["low"], df["close"], p.atr_len)
atr_ok = np.nan_to_num(atr_s.to_numpy(dtype=float), nan=0.0)
long_energy_ok = np.ones(n, dtype=bool)
short_energy_ok = np.ones(n, dtype=bool)
prior_high_n = (
df["high"].shift(1).rolling(p.breakout_n, min_periods=p.breakout_n).max().to_numpy(dtype=float)
)
prior_low_n = (
df["low"].shift(1).rolling(p.breakout_n, min_periods=p.breakout_n).min().to_numpy(dtype=float)
)
prev_body = np.empty(n, dtype=float)
prev_body[0] = np.nan
prev_body[1:] = np.abs(close[:-1] - open_[:-1])
prev_bearish = np.zeros(n, dtype=bool)
prev_bullish = np.zeros(n, dtype=bool)
prev_bearish[1:] = close[:-1] < open_[:-1]
prev_bullish[1:] = close[:-1] > open_[:-1]
bull_min_cover = np.empty(n, dtype=float)
bear_min_cover = np.empty(n, dtype=float)
bull_min_cover[0] = np.nan
bear_min_cover[0] = np.nan
bull_min_cover[1:] = close[:-1] + prev_body[1:] * p.engulf_cover
bear_min_cover[1:] = close[:-1] - prev_body[1:] * p.engulf_cover
bull_engulf = (close > open_) & prev_bearish & (close >= bull_min_cover)
bear_engulf = (close < open_) & prev_bullish & (close <= bear_min_cover)
bull_bo = close > prior_high_n
bear_bo = close < prior_low_n
sl_long_px = close - atr_ok * p.sl_atr_n
sl_short_px = close + atr_ok * p.sl_atr_n
accepted_long = np.zeros(n, dtype=bool)
accepted_short = np.zeros(n, dtype=bool)
entry_px = np.full(n, np.nan)
long_sl_out = np.full(n, np.nan)
short_sl_out = np.full(n, np.nan)
long_tp1 = np.full(n, np.nan)
long_tp2 = np.full(n, np.nan)
long_tp3 = np.full(n, np.nan)
short_tp1 = np.full(n, np.nan)
short_tp2 = np.full(n, np.nan)
short_tp3 = np.full(n, np.nan)
bots: list[float] = []
tops: list[float] = []
dirs: list[int] = []
created: list[int] = []
mid_lo: list[float] = []
mid_hi: list[float] = []
valid: list[bool] = []
long_locked = False
short_locked = False
lock_long_sl = np.nan
lock_short_sl = np.nan
lock_long_bar = -1
lock_short_bar = -1
pos_dir = 0
pos_sl = np.nan
pos_bar = -1
last_long_sl = np.nan
last_long_tp1 = np.nan
last_long_tp2 = np.nan
last_long_tp3 = np.nan
last_short_sl = np.nan
last_short_tp1 = np.nan
last_short_tp2 = np.nan
last_short_tp3 = np.nan
last_entry = np.nan
def _remove(idx: int) -> None:
del bots[idx], tops[idx], dirs[idx], created[idx], mid_lo[idx], mid_hi[idx], valid[idx]
def _add(direction: int, bot: float, top: float, mlo: float, mhi: float, bar: int) -> None:
for k in range(len(dirs)):
if dirs[k] == direction:
valid[k] = False
bots.append(bot)
tops.append(top)
dirs.append(direction)
created.append(bar)
mid_lo.append(mlo)
mid_hi.append(mhi)
valid.append(True)
for i in range(n):
j = len(bots) - 1
while j >= 0:
if i - created[j] > p.max_bars:
_remove(j)
j -= 1
for k in range(len(bots)):
if not valid[k]:
continue
broken = close[i] < mid_lo[k] if dirs[k] == 1 else close[i] > mid_hi[k]
if broken:
valid[k] = False
new_bull = i >= 2 and low[i] > high[i - 2]
new_bear = i >= 2 and high[i] < low[i - 2]
if new_bull:
_add(1, float(high[i - 2]), float(low[i]), float(low[i - 1]), float(high[i - 1]), i)
if new_bear:
_add(-1, float(high[i]), float(low[i - 2]), float(low[i - 1]), float(high[i - 1]), i)
if (new_bull or new_bear) and bots:
last = len(bots) - 1
broken_new = close[i] < mid_lo[last] if dirs[last] == 1 else close[i] > mid_hi[last]
if broken_new:
valid[last] = False
sl_long_hit = pos_dir == 1 and i > pos_bar and low[i] <= pos_sl
sl_short_hit = pos_dir == -1 and i > pos_bar and high[i] >= pos_sl
if sl_long_hit or sl_short_hit:
pos_dir = 0
pos_sl = np.nan
pos_bar = -1
if p.use_lock_until_sl:
if sl_long_hit or (long_locked and i > lock_long_bar and low[i] <= lock_long_sl):
long_locked = False
lock_long_sl = np.nan
lock_long_bar = -1
if sl_short_hit or (short_locked and i > lock_short_bar and high[i] >= lock_short_sl):
short_locked = False
lock_short_sl = np.nan
lock_short_bar = -1
cur_idx = -1
for k in range(len(bots) - 1, -1, -1):
if valid[k]:
cur_idx = k
break
sig_long = False
sig_short = False
if cur_idx >= 0:
cur_dir = dirs[cur_idx]
cur_bot = bots[cur_idx]
cur_top = tops[cur_idx]
opp_ok = True
prev = cur_idx - 1
while prev >= 0:
if dirs[prev] != cur_dir:
opp_ok = not valid[prev]
break
prev -= 1
if opp_ok:
long_free = (not p.use_lock_until_sl) or (not long_locked)
short_free = (not p.use_lock_until_sl) or (not short_locked)
long_ok = cur_dir == 1 and close[i] > cur_bot
short_ok = cur_dir == -1 and close[i] < cur_top
long_pat = long_ok and long_free and (bool(bull_engulf[i]) or bool(bull_bo[i]))
short_pat = short_ok and short_free and (bool(bear_engulf[i]) or bool(bear_bo[i]))
sig_long = long_pat and bool(long_energy_ok[i])
sig_short = short_pat and bool(short_energy_ok[i])
if sig_long:
entry = float(close[i])
ind_sl = float(sl_long_px[i])
if np.isnan(ind_sl) or ind_sl >= entry:
sig_long = False
else:
live_sl, t1, t2, t3 = _calc_levels(
entry, ind_sl, True, p.sl_live_r, p.tp1_rr, p.tp2_rr, p.tp3_rr
)
last_long_sl, last_long_tp1, last_long_tp2, last_long_tp3 = (
live_sl, t1, t2 if t2 is not None else np.nan, t3 if t3 is not None else np.nan
)
last_short_sl = last_short_tp1 = last_short_tp2 = last_short_tp3 = np.nan
last_entry = entry
pos_dir = 1
pos_sl = live_sl
pos_bar = i
if p.use_lock_until_sl:
long_locked = True
lock_long_sl = live_sl
lock_long_bar = i
short_locked = False
lock_short_sl = np.nan
lock_short_bar = -1
if sig_short:
entry = float(close[i])
ind_sl = float(sl_short_px[i])
if np.isnan(ind_sl) or ind_sl <= entry:
sig_short = False
else:
live_sl, t1, t2, t3 = _calc_levels(
entry, ind_sl, False, p.sl_live_r, p.tp1_rr, p.tp2_rr, p.tp3_rr
)
last_short_sl, last_short_tp1, last_short_tp2, last_short_tp3 = (
live_sl, t1, t2 if t2 is not None else np.nan, t3 if t3 is not None else np.nan
)
last_long_sl = last_long_tp1 = last_long_tp2 = last_long_tp3 = np.nan
last_entry = entry
pos_dir = -1
pos_sl = live_sl
pos_bar = i
if p.use_lock_until_sl:
short_locked = True
lock_short_sl = live_sl
lock_short_bar = i
long_locked = False
lock_long_sl = np.nan
lock_long_bar = -1
accepted_long[i] = sig_long
accepted_short[i] = sig_short
entry_px[i] = last_entry
long_sl_out[i] = last_long_sl
short_sl_out[i] = last_short_sl
long_tp1[i] = last_long_tp1
long_tp2[i] = last_long_tp2
long_tp3[i] = last_long_tp3
short_tp1[i] = last_short_tp1
short_tp2[i] = last_short_tp2
short_tp3[i] = last_short_tp3
df["atr"] = atr_s
df["buy_sig"] = accepted_long
df["sell_sig"] = accepted_short
df["entry_px"] = entry_px
df["long_sl"] = long_sl_out
df["short_sl"] = short_sl_out
df["long_tp1"] = long_tp1
df["long_tp2"] = long_tp2
df["long_tp3"] = long_tp3
df["short_tp1"] = short_tp1
df["short_tp2"] = short_tp2
df["short_tp3"] = short_tp3
return df
def last_fvg_signal(df: pd.DataFrame, params: FvgParams | None = None) -> IndicatorSignal | None:
if df.empty:
return None
p = params or FvgParams()
last = df.iloc[-1]
ts = int(df.index[-1].timestamp())
close = float(last["close"])
def _opt(value: object) -> float | None:
if value is None or (isinstance(value, float) and np.isnan(value)):
return None
number = float(value)
return None if np.isnan(number) else number
if bool(last["buy_sig"]):
sl = float(last["long_sl"])
tp1 = float(last["long_tp1"])
if np.isnan(sl) or np.isnan(tp1) or np.isnan(close):
return None
size = calc_size(close, sl, p.risk_usd)
if np.isnan(size):
return None
return IndicatorSignal(
strategy_id="fvg",
side="long",
entry=close,
close=close,
sl=sl,
tp1=tp1,
tp2=_opt(last["long_tp2"]),
tp3=_opt(last["long_tp3"]),
size_usd=size,
bar_open_ts=ts,
visual_timeframe=p.visual_timeframe,
)
if bool(last["sell_sig"]):
sl = float(last["short_sl"])
tp1 = float(last["short_tp1"])
if np.isnan(sl) or np.isnan(tp1) or np.isnan(close):
return None
size = calc_size(close, sl, p.risk_usd)
if np.isnan(size):
return None
return IndicatorSignal(
strategy_id="fvg",
side="short",
entry=close,
close=close,
sl=sl,
tp1=tp1,
tp2=_opt(last["short_tp2"]),
tp3=_opt(last["short_tp3"]),
size_usd=size,
bar_open_ts=ts,
visual_timeframe=p.visual_timeframe,
)
return None

384
app/indicators/ltf.py Normal file
View file

@ -0,0 +1,384 @@
from __future__ import annotations
from dataclasses import dataclass
import numpy as np
import pandas as pd
from app.indicators.common import (
IndicatorSignal,
alma,
calc_size,
compute_fractals,
epoch_ms,
ha_rsi,
ohlc_frame,
)
@dataclass(frozen=True)
class LtfParams:
fractal_n: int = 5
alma_length: int = 500
engulf_threshold: float = 0.9
rsi_6h_overbought: float = 75.0
rsi_6h_oversold: float = 20.0
tp1_rr: float = 1.75
tp2_rr: float = 2.0
tp3_rr: float = 4.0
risk_usd: float = 40.0
cooldown_ms: int = 360 * 60 * 1000
engulf_reset_ms: int = 12 * 60 * 60 * 1000
tf_minutes: int = 15
visual_timeframe: str = "15"
def _merge_asof(df_15m: pd.DataFrame, df_6h: pd.DataFrame, columns: list[str]) -> pd.DataFrame:
left = pd.DataFrame({"ts": df_15m.index})
right = pd.DataFrame({"ts": df_6h.index})
for col in columns:
right[col] = df_6h[col].to_numpy()
left = left.sort_values("ts")
right = right.sort_values("ts")
merged = pd.merge_asof(left, right, on="ts", direction="backward")
merged = merged.set_index("ts")
return merged.reindex(df_15m.index)
def _compute_6h(
df_6h: pd.DataFrame,
*,
engulf_threshold: float,
forming: pd.Series | None,
) -> pd.DataFrame:
df = ohlc_frame(df_6h)
df["ha_rsi"] = ha_rsi(df["open"], df["high"], df["low"], df["close"], 14)
df["h6_close"] = df["close"]
df["h6_prev_close"] = df["close"].shift(1)
df["h6_open"] = df["open"]
df["h6_prev_open"] = df["open"].shift(1)
body2 = (df["h6_prev_open"] - df["h6_prev_close"]).abs()
bull_min = df["h6_prev_close"] + body2 * engulf_threshold
bear_min = df["h6_prev_close"] - body2 * engulf_threshold
bull_engulf = (
(df["h6_close"] > df["h6_open"])
& (df["h6_close"] >= bull_min)
& (df["h6_prev_close"] < df["h6_prev_open"])
)
bear_engulf = (
(df["h6_close"] < df["h6_open"])
& (df["h6_close"] <= bear_min)
& (df["h6_prev_close"] > df["h6_prev_open"])
)
n = len(df)
is_forming = np.zeros(n, dtype=bool)
if forming is not None:
is_forming = forming.reindex(df.index).fillna(False).to_numpy(dtype=bool)
state = np.empty(n, dtype=object)
state[:] = ""
engulf_time = np.full(n, np.nan)
cur_state = ""
cur_time = np.nan
ts_ms = epoch_ms(df.index)
bull_a = bull_engulf.fillna(False).to_numpy(dtype=bool)
bear_a = bear_engulf.fillna(False).to_numpy(dtype=bool)
for i in range(n):
ts = ts_ms[i]
if not np.isnan(cur_time) and (ts - cur_time) > 12 * 60 * 60 * 1000:
cur_state = ""
cur_time = np.nan
if is_forming[i]:
state[i] = cur_state
engulf_time[i] = cur_time
continue
if bull_a[i]:
cur_state = "bull"
cur_time = ts
elif bear_a[i]:
cur_state = "bear"
cur_time = ts
else:
cur_state = ""
cur_time = np.nan
state[i] = cur_state
engulf_time[i] = cur_time
df["engulf_state"] = state
df["engulf_time"] = engulf_time
return df
def attach_forming_6h(df_6h_closed: pd.DataFrame, df_15m: pd.DataFrame) -> pd.DataFrame:
"""Rebuild the in-progress 6h candle from closed 15m bars (Pine request.security)."""
closed = ohlc_frame(df_6h_closed)
ltf = ohlc_frame(df_15m)
if ltf.empty:
closed["_forming"] = False
return closed
period = ltf.index[-1].floor("6h")
window = ltf.loc[ltf.index >= period]
if window.empty:
closed["_forming"] = False
return closed
# 6h just closed and is already in the closed frame.
if not closed.empty and closed.index[-1] == period:
closed["_forming"] = False
return closed
row = pd.DataFrame(
{
"open": [float(window["open"].iloc[0])],
"high": [float(window["high"].max())],
"low": [float(window["low"].min())],
"close": [float(window["close"].iloc[-1])],
},
index=pd.DatetimeIndex([period], tz="UTC"),
)
if "volume" in window.columns:
row["volume"] = float(window["volume"].sum())
row["_forming"] = True
base = closed[closed.index < period].copy()
base["_forming"] = False
return pd.concat([base, row])
def evaluate_ltf(
df_15m: pd.DataFrame,
df_6h: pd.DataFrame,
params: LtfParams | None = None,
*,
forming_6h: pd.Series | None = None,
) -> pd.DataFrame:
"""Compute LTF columns. Last row is the latest closed 15m bar."""
p = params or LtfParams()
df = ohlc_frame(df_15m)
forming_flag = forming_6h
if forming_flag is None and "_forming" in df_6h.columns:
forming_flag = df_6h["_forming"].astype(bool)
h6 = _compute_6h(df_6h, engulf_threshold=p.engulf_threshold, forming=forming_flag)
if forming_flag is not None:
closed_mask = ~forming_flag.reindex(h6.index).fillna(False)
h6_closed = h6.loc[closed_mask]
else:
h6_closed = h6
n = int(p.fractal_n)
high = df["high"].to_numpy(dtype=float)
low = df["low"].to_numpy(dtype=float)
up_frac, down_frac = compute_fractals(high, low, n=n)
df["up_fractal"] = up_frac
df["down_fractal"] = down_frac
up_level = np.full(len(df), np.nan)
down_level = np.full(len(df), np.nan)
last_up = np.nan
last_down = np.nan
for i in range(len(df)):
if up_frac[i]:
last_up = high[i - n]
if down_frac[i]:
last_down = low[i - n]
up_level[i] = last_up
down_level[i] = last_down
df["up_fractal_level"] = up_level
df["down_fractal_level"] = down_level
prev_up = df["up_fractal_level"].shift(1)
prev_down = df["down_fractal_level"].shift(1)
df["buy_crossover"] = (
(df["close"] > prev_up) & (df["close"].shift(1) <= prev_up.shift(1)) & prev_up.notna()
)
df["sell_crossover"] = (
(df["close"] < prev_down) & (df["close"].shift(1) >= prev_down.shift(1)) & prev_down.notna()
)
bars_16h = max(1, int(round(1440 / p.tf_minutes)))
high_shift = df["high"].shift(1)
low_shift = df["low"].shift(1)
df["high_16h"] = high_shift.rolling(bars_16h, min_periods=bars_16h).max()
df["low_16h"] = low_shift.rolling(bars_16h, min_periods=bars_16h).min()
df["bull_breakout"] = df["high"] > df["high_16h"]
df["bear_breakout"] = df["low"] < df["low_16h"]
df["alma"] = alma(df["close"], length=p.alma_length, offset=0.85, sigma=5.0)
rsi_m = _merge_asof(df, h6, ["ha_rsi"])
closed_src = h6_closed if not h6_closed.empty else h6
closed_m = _merge_asof(
df, closed_src, ["engulf_state", "engulf_time", "h6_close", "h6_prev_close"]
)
df["rsi_6h"] = rsi_m["ha_rsi"]
df["h6_engulf_state"] = closed_m["engulf_state"].fillna("").astype(str)
df["h6_closed_close"] = closed_m["h6_close"]
df["h6_prev_close"] = closed_m["h6_prev_close"]
engulf_time = closed_m["engulf_time"].to_numpy(dtype=float)
ts_ms = epoch_ms(df.index)
engulf_state = df["h6_engulf_state"].to_numpy()
for i in range(len(df)):
et = engulf_time[i]
if not np.isnan(et) and (ts_ms[i] - et) > p.engulf_reset_ms:
engulf_state[i] = ""
df["h6_engulf_state"] = engulf_state
bull_arr = (df["bull_breakout"] | (df["h6_engulf_state"] == "bull")).to_numpy(dtype=bool)
bear_arr = (df["bear_breakout"] | (df["h6_engulf_state"] == "bear")).to_numpy(dtype=bool)
last_eng = np.empty(len(df), dtype=object)
last_eng[:] = ""
cur = ""
for i in range(len(df)):
b = bull_arr[i]
s = bear_arr[i]
if b and not s:
cur = "bull"
elif s and not b:
cur = "bear"
last_eng[i] = cur
df["last_eng"] = last_eng
alma_up = (df["alma"] > df["alma"].shift(1)) & (df["alma"].shift(1) > df["alma"].shift(2))
alma_down = (df["alma"] < df["alma"].shift(1)) & (df["alma"].shift(1) < df["alma"].shift(2))
alma_6h_long = (df["h6_closed_close"] > df["alma"]) & (
df["h6_prev_close"] > df["alma"].shift(1)
)
alma_6h_short = (df["h6_closed_close"] < df["alma"]) & (
df["h6_prev_close"] < df["alma"].shift(1)
)
buy_6h_ok = df["rsi_6h"] < p.rsi_6h_overbought
sell_6h_ok = df["rsi_6h"] > p.rsi_6h_oversold
buy_sig = np.zeros(len(df), dtype=bool)
sell_sig = np.zeros(len(df), dtype=bool)
last_buy_time = np.nan
last_sell_time = np.nan
open_side = ""
long_sl_px = np.nan
short_sl_px = np.nan
close_a = df["close"].to_numpy(dtype=float)
low_a = df["low"].to_numpy(dtype=float)
high_a = df["high"].to_numpy(dtype=float)
buy_x = df["buy_crossover"].fillna(False).to_numpy(dtype=bool)
sell_x = df["sell_crossover"].fillna(False).to_numpy(dtype=bool)
alma_up_a = alma_up.fillna(False).to_numpy(dtype=bool)
alma_down_a = alma_down.fillna(False).to_numpy(dtype=bool)
alma_6h_long_a = alma_6h_long.fillna(False).to_numpy(dtype=bool)
alma_6h_short_a = alma_6h_short.fillna(False).to_numpy(dtype=bool)
buy_6h_a = buy_6h_ok.fillna(False).to_numpy(dtype=bool)
sell_6h_a = sell_6h_ok.fillna(False).to_numpy(dtype=bool)
long_sl_arr = df["low_16h"].to_numpy(dtype=float)
short_sl_arr = df["high_16h"].to_numpy(dtype=float)
for i in range(len(df)):
if not np.isnan(long_sl_px) and (close_a[i] < long_sl_px or low_a[i] <= long_sl_px):
long_sl_px = np.nan
if open_side == "buy":
open_side = ""
if not np.isnan(short_sl_px) and (close_a[i] > short_sl_px or high_a[i] >= short_sl_px):
short_sl_px = np.nan
if open_side == "sell":
open_side = ""
ts = ts_ms[i]
buy_cd = np.isnan(last_buy_time) or (ts - last_buy_time >= p.cooldown_ms)
sell_cd = np.isnan(last_sell_time) or (ts - last_sell_time >= p.cooldown_ms)
is_buy = (
buy_x[i]
and open_side != "buy"
and last_eng[i] == "bull"
and (alma_up_a[i] or alma_6h_long_a[i])
and buy_cd
and buy_6h_a[i]
)
is_sell = (
sell_x[i]
and open_side != "sell"
and last_eng[i] == "bear"
and (alma_down_a[i] or alma_6h_short_a[i])
and sell_cd
and sell_6h_a[i]
)
if is_buy:
buy_sig[i] = True
last_buy_time = ts
open_side = "buy"
long_sl_px = long_sl_arr[i]
short_sl_px = np.nan
if is_sell:
sell_sig[i] = True
last_sell_time = ts
open_side = "sell"
short_sl_px = short_sl_arr[i]
long_sl_px = np.nan
df["buy_sig"] = buy_sig
df["sell_sig"] = sell_sig
df["long_sl"] = df["low_16h"]
df["short_sl"] = df["high_16h"]
df["long_tp1"] = df["close"] + (df["close"] - df["long_sl"]) * p.tp1_rr
df["long_tp2"] = df["close"] + (df["close"] - df["long_sl"]) * p.tp2_rr
df["long_tp3"] = df["close"] + (df["close"] - df["long_sl"]) * p.tp3_rr
df["short_tp1"] = df["close"] - (df["short_sl"] - df["close"]) * p.tp1_rr
df["short_tp2"] = df["close"] - (df["short_sl"] - df["close"]) * p.tp2_rr
df["short_tp3"] = df["close"] - (df["short_sl"] - df["close"]) * p.tp3_rr
return df
def last_ltf_signal(df: pd.DataFrame, params: LtfParams | None = None) -> IndicatorSignal | None:
if df.empty:
return None
p = params or LtfParams()
last = df.iloc[-1]
ts = int(df.index[-1].timestamp())
close = float(last["close"])
if bool(last["buy_sig"]):
sl = float(last["long_sl"])
if np.isnan(sl) or np.isnan(close):
return None
size = calc_size(close, sl, p.risk_usd)
if np.isnan(size):
return None
return IndicatorSignal(
strategy_id="ltf",
side="long",
entry=close,
close=close,
sl=sl,
tp1=float(last["long_tp1"]),
tp2=float(last["long_tp2"]),
tp3=float(last["long_tp3"]),
size_usd=size,
bar_open_ts=ts,
visual_timeframe=p.visual_timeframe,
)
if bool(last["sell_sig"]):
sl = float(last["short_sl"])
if np.isnan(sl) or np.isnan(close):
return None
size = calc_size(close, sl, p.risk_usd)
if np.isnan(size):
return None
return IndicatorSignal(
strategy_id="ltf",
side="short",
entry=close,
close=close,
sl=sl,
tp1=float(last["short_tp1"]),
tp2=float(last["short_tp2"]),
tp3=float(last["short_tp3"]),
size_usd=size,
bar_open_ts=ts,
visual_timeframe=p.visual_timeframe,
)
return None

103
app/main.py
View file

@ -4,18 +4,16 @@ import asyncio
import json
import logging
import secrets
from contextlib import asynccontextmanager
from typing import Any
from fastapi import BackgroundTasks, FastAPI, HTTPException, Request
from fastapi.responses import JSONResponse
from pydantic import ValidationError
from app.binance import fetch_klines, to_binance_interval, to_binance_symbol
from app.chart import render_setup_chart
from app.config import Settings, get_settings
from app.formatter import format_caption
from app.models import SignalPayload
from app.telegram import TelegramError, send_message, send_photo
from app.pipeline import deliver_telegram
logging.basicConfig(
level=logging.INFO,
@ -23,12 +21,37 @@ logging.basicConfig(
)
logger = logging.getLogger(__name__)
@asynccontextmanager
async def lifespan(_app: FastAPI):
settings = get_settings()
task: asyncio.Task[None] | None = None
if settings.scanner_enabled:
from app.scanner import run_scanner
task = asyncio.create_task(run_scanner(settings))
logger.info(
"Scanner enabled watchlist=%s poll=%ss",
settings.watchlist_path,
settings.scanner_poll_seconds,
)
yield
if task is not None:
task.cancel()
try:
await task
except asyncio.CancelledError:
pass
logger.info("Scanner stopped")
app = FastAPI(
title="TV Signals → Telegram",
version="1.0.0",
docs_url=None,
redoc_url=None,
openapi_url=None,
lifespan=lifespan,
)
@ -53,76 +76,6 @@ def _parse_signal_body(raw: bytes) -> SignalPayload:
raise HTTPException(status_code=422, detail=json.loads(exc.json())) from exc
async def _deliver_signal(
settings: Settings,
signal: SignalPayload,
message_thread_id: int,
) -> None:
"""Fetch chart + post to Telegram after the webhook HTTP response is sent."""
caption = format_caption(signal)
photo: bytes | None = None
chart_error: str | None = None
try:
symbol = to_binance_symbol(signal.ticker)
interval = to_binance_interval(signal.visual_timeframe)
df = await fetch_klines(symbol, interval)
photo = await asyncio.to_thread(
render_setup_chart,
df,
ticker=signal.ticker,
action=signal.action.value,
entry=signal.entry_price,
stop_loss=signal.stop_loss_price,
tp1=signal.take_profit_1_price,
tp2=signal.take_profit_2_price,
tp3=signal.take_profit_3_price,
timeframe=signal.visual_timeframe,
current_price=(
signal.current_price if signal.signal_sequence > 1 else None
),
signal_time=signal.signal_time,
)
except Exception as exc: # noqa: BLE001 — fallback to text-only post
chart_error = str(exc)
logger.exception("Chart generation failed, falling back to text-only: %s", exc)
try:
if photo is not None:
await send_photo(
settings,
photo=photo,
caption=caption,
message_thread_id=message_thread_id,
)
logger.info(
"Delivered photo: %s %s seq=%s thread=%s",
signal.ticker,
signal.action.value,
signal.signal_sequence,
message_thread_id,
)
return
await send_message(
settings,
text=caption,
message_thread_id=message_thread_id,
)
logger.info(
"Delivered text: %s %s seq=%s thread=%s chart_error=%s",
signal.ticker,
signal.action.value,
signal.signal_sequence,
message_thread_id,
chart_error,
)
except TelegramError as exc:
logger.exception("Telegram delivery failed: %s", exc)
except Exception as exc: # noqa: BLE001
logger.exception("Unexpected delivery error: %s", exc)
@app.api_route("/health", methods=["GET", "HEAD"])
async def health() -> dict[str, str]:
return {"status": "ok"}
@ -153,7 +106,7 @@ async def _handle_webhook(
)
background_tasks.add_task(
_deliver_signal,
deliver_telegram,
settings,
signal,
message_thread_id,

25
app/models.py
View file

@ -1,4 +1,7 @@
from __future__ import annotations
from enum import Enum
from typing import Optional
from pydantic import BaseModel, Field, field_validator
@ -15,11 +18,16 @@ class SignalPayload(BaseModel):
current_price: str
stop_loss_price: str
take_profit_1_price: str
take_profit_2_price: str
take_profit_3_price: str
take_profit_2_price: Optional[str] = None
take_profit_3_price: Optional[str] = None
visual_timeframe: str
signal_sequence: int = Field(ge=1)
signal_time: int = Field(gt=0) # unix seconds of seq==1 bar open (UTC)
is_reversal: bool = False
realized_pnl_pct: Optional[float] = None
prev_side: Optional[str] = None
prev_entry_price: Optional[str] = None
prev_signal_time: Optional[int] = None
@field_validator("action", mode="before")
@classmethod
@ -34,8 +42,6 @@ class SignalPayload(BaseModel):
"current_price",
"stop_loss_price",
"take_profit_1_price",
"take_profit_2_price",
"take_profit_3_price",
"visual_timeframe",
mode="before",
)
@ -44,3 +50,14 @@ class SignalPayload(BaseModel):
if isinstance(value, str):
return value.strip()
return value
@field_validator("take_profit_2_price", "take_profit_3_price", mode="before")
@classmethod
def empty_optional_price(cls, value: object) -> object:
if value is None:
return None
if isinstance(value, str) and not value.strip():
return None
if isinstance(value, str):
return value.strip()
return value

181
app/pipeline.py Normal file
View file

@ -0,0 +1,181 @@
from __future__ import annotations
import asyncio
import logging
from app.binance import (
chart_kline_limit,
chart_right_pad,
fetch_klines,
get_tick_size,
interval_timedelta,
to_binance_interval,
to_binance_symbol,
)
from app.chart import render_setup_chart
from app.config import Settings
from app.formatter import format_caption
from app.heryon import build_heryon_payload, send_heryon, to_tv_perp_ticker
from app.indicators.common import IndicatorSignal, format_px, realized_pnl_pct
from app.models import SignalPayload
from app.state import ScannerStore
from app.telegram import TelegramError, send_message, send_photo, telegram_message_id
logger = logging.getLogger(__name__)
def to_telegram_payload(signal: IndicatorSignal, ticker: str) -> SignalPayload:
tick = get_tick_size(ticker)
return SignalPayload(
ticker=ticker,
action=signal.side,
entry_price=format_px(signal.entry, tick),
current_price=format_px(signal.close, tick),
stop_loss_price=format_px(signal.sl, tick),
take_profit_1_price=format_px(signal.tp1, tick),
take_profit_2_price=format_px(signal.tp2, tick) or None,
take_profit_3_price=format_px(signal.tp3, tick) or None,
visual_timeframe=signal.visual_timeframe,
signal_sequence=1,
signal_time=signal.bar_open_ts,
)
async def deliver_telegram(
settings: Settings,
signal: SignalPayload,
message_thread_id: int,
*,
store: ScannerStore | None = None,
) -> None:
"""Fetch chart + post to Telegram (used by inbound webhook and scanner)."""
caption = format_caption(signal)
photo: bytes | None = None
chart_error: str | None = None
owns_store = store is None
db = store or ScannerStore(settings.scanner_state_path)
reply_id = db.get_last_tg_message(message_thread_id, signal.ticker)
try:
symbol = to_binance_symbol(signal.ticker)
interval = to_binance_interval(signal.visual_timeframe)
limit = chart_kline_limit(interval)
if signal.prev_signal_time and signal.prev_signal_time > 0:
bar_sec = interval_timedelta(interval).total_seconds()
if bar_sec > 0:
span = int((signal.signal_time - signal.prev_signal_time) / bar_sec) + 24
limit = min(1500, max(limit, span))
df = await fetch_klines(symbol, interval, limit=limit)
# LTF 15m: chart shows TP1 + remainder green (like FVG); caption still has TP2/TP3.
chart_tp2 = None if interval == "15m" else signal.take_profit_2_price
chart_tp3 = None if interval == "15m" else signal.take_profit_3_price
photo = await asyncio.to_thread(
render_setup_chart,
df,
ticker=signal.ticker,
action=signal.action.value,
entry=signal.entry_price,
stop_loss=signal.stop_loss_price,
tp1=signal.take_profit_1_price,
tp2=chart_tp2,
tp3=chart_tp3,
timeframe=signal.visual_timeframe,
current_price=(
signal.current_price if signal.signal_sequence > 1 else None
),
signal_time=signal.signal_time,
right_pad=chart_right_pad(interval),
prev_entry=signal.prev_entry_price,
prev_side=signal.prev_side,
prev_signal_time=signal.prev_signal_time,
)
except Exception as exc: # noqa: BLE001 — fallback to text-only post
chart_error = str(exc)
logger.exception("Chart generation failed, falling back to text-only: %s", exc)
try:
if photo is not None:
response = await send_photo(
settings,
photo=photo,
caption=caption,
message_thread_id=message_thread_id,
reply_to_message_id=reply_id,
)
else:
response = await send_message(
settings,
text=caption,
message_thread_id=message_thread_id,
reply_to_message_id=reply_id,
)
message_id = telegram_message_id(response)
if message_id is not None:
db.set_last_tg_message(message_thread_id, signal.ticker, message_id)
logger.info(
"Delivered %s: %s %s seq=%s thread=%s chart_error=%s",
"photo" if photo is not None else "text",
signal.ticker,
signal.action.value,
signal.signal_sequence,
message_thread_id,
chart_error,
)
except TelegramError as exc:
logger.exception("Telegram delivery failed: %s", exc)
except Exception as exc: # noqa: BLE001
logger.exception("Unexpected delivery error: %s", exc)
finally:
if owns_store:
db.close()
async def deliver_generated(
settings: Settings,
signal: IndicatorSignal,
symbol: str,
message_thread_id: int,
*,
store: ScannerStore | None = None,
) -> None:
ticker = to_tv_perp_ticker(symbol)
telegram_payload = to_telegram_payload(signal, ticker)
if store is not None:
previous = store.get_last_open(signal.strategy_id, symbol)
if previous is not None:
prev_side, prev_entry, prev_ts = previous
if prev_side != signal.side:
tick = get_tick_size(ticker)
telegram_payload = telegram_payload.model_copy(
update={
"is_reversal": True,
"realized_pnl_pct": realized_pnl_pct(
prev_side, prev_entry, signal.entry
),
"prev_side": prev_side,
"prev_entry_price": format_px(prev_entry, tick),
"prev_signal_time": prev_ts or None,
}
)
heryon_payload = build_heryon_payload(signal, symbol=symbol, settings=settings)
await deliver_telegram(
settings,
telegram_payload,
message_thread_id,
store=store,
)
if store is not None:
store.set_last_open(
signal.strategy_id,
symbol,
signal.side,
signal.entry,
signal.bar_open_ts,
)
try:
await send_heryon(settings, heryon_payload)
except Exception as exc: # noqa: BLE001
logger.exception(
"Heryon send failed nonce=%s: %s", heryon_payload.get("nonce"), exc
)

164
app/scanner.py Normal file
View file

@ -0,0 +1,164 @@
from __future__ import annotations
import asyncio
import logging
import httpx
from app.binance import fetch_klines, load_tick_sizes, to_binance_symbol
from app.config import Settings
from app.heryon import heryon_nonce, to_tv_perp_ticker
from app.indicators.fvg import FvgParams, evaluate_fvg, last_fvg_signal
from app.indicators.ltf import LtfParams, attach_forming_6h, evaluate_ltf, last_ltf_signal
from app.pipeline import deliver_generated
from app.state import ScannerStore
from app.watchlist import Watchlist, load_watchlist
logger = logging.getLogger(__name__)
LTF_15M_BARS = 600
LTF_6H_BARS = 250
FVG_6H_BARS = 400
async def run_scanner(settings: Settings) -> None:
store = ScannerStore(settings.scanner_state_path)
try:
await load_tick_sizes()
while True:
try:
watchlist = load_watchlist(settings.watchlist_path)
await scan_once(settings, watchlist, store)
except asyncio.CancelledError:
raise
except Exception:
logger.exception("Scanner cycle failed")
await asyncio.sleep(settings.scanner_poll_seconds)
finally:
store.close()
async def scan_once(
settings: Settings,
watchlist: Watchlist,
store: ScannerStore,
) -> None:
ltf_cfg = watchlist.strategies.get("ltf")
fvg_cfg = watchlist.strategies.get("fvg")
async with httpx.AsyncClient(timeout=20.0) as client:
for symbol in watchlist.symbols:
pair = to_binance_symbol(symbol)
if ltf_cfg is not None and ltf_cfg.enabled:
try:
await _scan_ltf(settings, store, client, pair, ltf_cfg.telegram_thread_id)
except Exception:
logger.exception("LTF scan failed for %s", pair)
if fvg_cfg is not None and fvg_cfg.enabled:
try:
await _scan_fvg(settings, store, client, pair, fvg_cfg.telegram_thread_id)
except Exception:
logger.exception("FVG scan failed for %s", pair)
def _thread(settings: Settings, override: int | None) -> int:
if override is not None and override >= 1:
return override
return settings.telegram_message_thread_id
async def _scan_ltf(
settings: Settings,
store: ScannerStore,
client: httpx.AsyncClient,
symbol: str,
thread_id: int | None,
) -> None:
df_15m = await fetch_klines(
symbol, "15m", limit=LTF_15M_BARS, closed_only=True, client=client
)
if df_15m.empty:
return
bar_ts = int(df_15m.index[-1].timestamp())
last = store.get_last_bar("ltf", symbol)
if last is None:
store.set_last_bar("ltf", symbol, bar_ts)
logger.info("LTF primed %s at %s (skip history)", symbol, bar_ts)
return
if bar_ts <= last:
return
df_6h_closed = await fetch_klines(
symbol, "6h", limit=LTF_6H_BARS, closed_only=True, client=client
)
if df_6h_closed.empty:
logger.warning("LTF %s: empty 6h klines", symbol)
return
df_6h = attach_forming_6h(df_6h_closed, df_15m)
params = LtfParams(risk_usd=settings.risk_usd)
analyzed = await asyncio.to_thread(evaluate_ltf, df_15m, df_6h, params)
signal = last_ltf_signal(analyzed, params)
await _emit_if_new(settings, store, "ltf", symbol, bar_ts, signal, thread_id)
async def _scan_fvg(
settings: Settings,
store: ScannerStore,
client: httpx.AsyncClient,
symbol: str,
thread_id: int | None,
) -> None:
df_6h = await fetch_klines(
symbol, "6h", limit=FVG_6H_BARS, closed_only=True, client=client
)
if df_6h.empty:
return
bar_ts = int(df_6h.index[-1].timestamp())
last = store.get_last_bar("fvg", symbol)
if last is None:
store.set_last_bar("fvg", symbol, bar_ts)
logger.info("FVG primed %s at %s (skip history)", symbol, bar_ts)
return
if bar_ts <= last:
return
params = FvgParams(risk_usd=settings.risk_usd)
analyzed = await asyncio.to_thread(evaluate_fvg, df_6h, params)
signal = last_fvg_signal(analyzed, params)
await _emit_if_new(settings, store, "fvg", symbol, bar_ts, signal, thread_id)
async def _emit_if_new(
settings: Settings,
store: ScannerStore,
strategy: str,
symbol: str,
bar_ts: int,
signal,
thread_id: int | None,
) -> None:
if signal is None:
store.set_last_bar(strategy, symbol, bar_ts)
logger.debug("%s %s new bar %s — no signal", strategy, symbol, bar_ts)
return
ticker = to_tv_perp_ticker(symbol)
nonce = heryon_nonce(ticker, signal.side, signal.bar_open_ts)
if store.nonce_sent(nonce):
store.set_last_bar(strategy, symbol, bar_ts)
logger.info("Skip duplicate nonce %s", nonce)
return
logger.info(
"Signal %s %s %s bar=%s sl=%s tp1=%s",
strategy,
symbol,
signal.side,
bar_ts,
signal.sl,
signal.tp1,
)
await deliver_generated(
settings, signal, symbol, _thread(settings, thread_id), store=store
)
store.mark_nonce(nonce)
store.set_last_bar(strategy, symbol, bar_ts)

145
app/state.py Normal file
View file

@ -0,0 +1,145 @@
from __future__ import annotations
import sqlite3
import time
from pathlib import Path
class ScannerStore:
"""Persist last processed bar per (strategy, symbol) and sent Heryon nonces."""
def __init__(self, path: str | Path) -> None:
self._path = Path(path)
self._path.parent.mkdir(parents=True, exist_ok=True)
self._conn = sqlite3.connect(self._path, check_same_thread=False)
self._conn.execute("PRAGMA journal_mode=WAL")
self._conn.execute(
"""
CREATE TABLE IF NOT EXISTS processed (
strategy TEXT NOT NULL,
symbol TEXT NOT NULL,
bar_open_ts INTEGER NOT NULL,
PRIMARY KEY (strategy, symbol)
)
"""
)
self._conn.execute(
"""
CREATE TABLE IF NOT EXISTS sent_nonces (
nonce TEXT PRIMARY KEY,
sent_at INTEGER NOT NULL
)
"""
)
self._conn.execute(
"""
CREATE TABLE IF NOT EXISTS last_open (
strategy TEXT NOT NULL,
symbol TEXT NOT NULL,
side TEXT NOT NULL,
entry REAL NOT NULL,
bar_open_ts INTEGER NOT NULL DEFAULT 0,
PRIMARY KEY (strategy, symbol)
)
"""
)
self._conn.execute(
"""
CREATE TABLE IF NOT EXISTS last_tg_message (
thread_id INTEGER NOT NULL,
ticker TEXT NOT NULL,
message_id INTEGER NOT NULL,
PRIMARY KEY (thread_id, ticker)
)
"""
)
last_open_cols = {
row[1] for row in self._conn.execute("PRAGMA table_info(last_open)")
}
if "bar_open_ts" not in last_open_cols:
self._conn.execute(
"ALTER TABLE last_open ADD COLUMN bar_open_ts INTEGER NOT NULL DEFAULT 0"
)
self._conn.commit()
def get_last_bar(self, strategy: str, symbol: str) -> int | None:
row = self._conn.execute(
"SELECT bar_open_ts FROM processed WHERE strategy = ? AND symbol = ?",
(strategy, symbol),
).fetchone()
return int(row[0]) if row else None
def set_last_bar(self, strategy: str, symbol: str, bar_open_ts: int) -> None:
self._conn.execute(
"""
INSERT INTO processed (strategy, symbol, bar_open_ts)
VALUES (?, ?, ?)
ON CONFLICT(strategy, symbol) DO UPDATE SET bar_open_ts = excluded.bar_open_ts
""",
(strategy, symbol, bar_open_ts),
)
self._conn.commit()
def nonce_sent(self, nonce: str) -> bool:
row = self._conn.execute(
"SELECT 1 FROM sent_nonces WHERE nonce = ?", (nonce,)
).fetchone()
return row is not None
def mark_nonce(self, nonce: str) -> None:
self._conn.execute(
"INSERT OR IGNORE INTO sent_nonces (nonce, sent_at) VALUES (?, ?)",
(nonce, int(time.time())),
)
self._conn.commit()
def get_last_open(self, strategy: str, symbol: str) -> tuple[str, float, int] | None:
row = self._conn.execute(
"SELECT side, entry, bar_open_ts FROM last_open WHERE strategy = ? AND symbol = ?",
(strategy, symbol),
).fetchone()
if row is None:
return None
return str(row[0]), float(row[1]), int(row[2] or 0)
def set_last_open(
self,
strategy: str,
symbol: str,
side: str,
entry: float,
bar_open_ts: int,
) -> None:
self._conn.execute(
"""
INSERT INTO last_open (strategy, symbol, side, entry, bar_open_ts)
VALUES (?, ?, ?, ?, ?)
ON CONFLICT(strategy, symbol) DO UPDATE SET
side = excluded.side,
entry = excluded.entry,
bar_open_ts = excluded.bar_open_ts
""",
(strategy, symbol, side, entry, bar_open_ts),
)
self._conn.commit()
def get_last_tg_message(self, thread_id: int, ticker: str) -> int | None:
row = self._conn.execute(
"SELECT message_id FROM last_tg_message WHERE thread_id = ? AND ticker = ?",
(thread_id, ticker.strip().upper()),
).fetchone()
return int(row[0]) if row else None
def set_last_tg_message(self, thread_id: int, ticker: str, message_id: int) -> None:
self._conn.execute(
"""
INSERT INTO last_tg_message (thread_id, ticker, message_id)
VALUES (?, ?, ?)
ON CONFLICT(thread_id, ticker) DO UPDATE SET message_id = excluded.message_id
""",
(thread_id, ticker.strip().upper(), message_id),
)
self._conn.commit()
def close(self) -> None:
self._conn.close()

21
app/telegram.py
View file

@ -15,6 +15,23 @@ class TelegramError(Exception):
pass
def telegram_message_id(payload: dict) -> int | None:
result = payload.get("result")
if not isinstance(result, dict):
return None
raw = result.get("message_id")
return int(raw) if raw is not None else None
def _reply_fields(reply_to_message_id: int | None) -> dict[str, str]:
if reply_to_message_id is None:
return {}
return {
"reply_to_message_id": str(reply_to_message_id),
"allow_sending_without_reply": "true",
}
async def send_photo(
settings: Settings,
*,
@ -22,6 +39,7 @@ async def send_photo(
caption: str,
message_thread_id: int,
filename: str = "setup.png",
reply_to_message_id: int | None = None,
) -> dict:
url = f"{TELEGRAM_API}/bot{settings.telegram_bot_token}/sendPhoto"
data = {
@ -29,6 +47,7 @@ async def send_photo(
"message_thread_id": str(message_thread_id),
"caption": caption,
"parse_mode": "HTML",
**_reply_fields(reply_to_message_id),
}
files = {"photo": (filename, photo, "image/png")}
async with httpx.AsyncClient(timeout=30.0) as client:
@ -44,6 +63,7 @@ async def send_message(
*,
text: str,
message_thread_id: int,
reply_to_message_id: int | None = None,
) -> dict:
url = f"{TELEGRAM_API}/bot{settings.telegram_bot_token}/sendMessage"
data = {
@ -52,6 +72,7 @@ async def send_message(
"text": text,
"parse_mode": "HTML",
"disable_web_page_preview": True,
**_reply_fields(reply_to_message_id),
}
async with httpx.AsyncClient(timeout=30.0) as client:
response = await client.post(url, data=data)

54
app/watchlist.py Normal file
View file

@ -0,0 +1,54 @@
from __future__ import annotations
from dataclasses import dataclass
from pathlib import Path
from typing import Any
import yaml
@dataclass(frozen=True)
class StrategyWatch:
id: str
enabled: bool
timeframe: str
telegram_thread_id: int | None
@dataclass(frozen=True)
class Watchlist:
symbols: list[str]
strategies: dict[str, StrategyWatch]
def load_watchlist(path: str | Path) -> Watchlist:
raw_path = Path(path)
if not raw_path.is_file():
raise FileNotFoundError(f"Watchlist not found: {raw_path}")
data: Any = yaml.safe_load(raw_path.read_text(encoding="utf-8")) or {}
symbols_raw = data.get("symbols") or []
symbols = [str(s).strip().upper() for s in symbols_raw if str(s).strip()]
if not symbols:
raise ValueError("Watchlist has no symbols")
strategies: dict[str, StrategyWatch] = {}
for key, cfg in (data.get("strategies") or {}).items():
if not isinstance(cfg, dict):
continue
strategies[str(key)] = StrategyWatch(
id=str(key),
enabled=bool(cfg.get("enabled", True)),
timeframe=str(cfg.get("timeframe", "")).strip(),
telegram_thread_id=_optional_int(cfg.get("telegram_thread_id")),
)
if "ltf" not in strategies:
strategies["ltf"] = StrategyWatch("ltf", True, "15m", None)
if "fvg" not in strategies:
strategies["fvg"] = StrategyWatch("fvg", True, "6h", None)
return Watchlist(symbols=symbols, strategies=strategies)
def _optional_int(value: object) -> int | None:
if value is None or value == "":
return None
return int(value)

0
data/.gitkeep Normal file
View file

3
docker-compose.yml
View file

@ -9,6 +9,9 @@ services:
HOST: "0.0.0.0"
PORT: "8000"
MPLBACKEND: Agg
volumes:
- ./watchlist.yaml:/app/watchlist.yaml:ro
- ./data:/app/data
restart: unless-stopped
networks:
default:

2
requirements.txt
View file

@ -4,6 +4,8 @@ pydantic>=2.9.0
pydantic-settings>=2.6.0
httpx>=0.27.0
pandas>=2.2.0
numpy>=2.0.0
pyyaml>=6.0.2
mplfinance>=0.12.10b0
matplotlib>=3.9.0
python-multipart>=0.0.12

19
watchlist.yaml Normal file
View file

@ -0,0 +1,19 @@
# Binance USDT-M perpetual symbols (no .P suffix). Edit without rebuilding if the
# file is volume-mounted (see docker-compose.yml).
symbols:
- BTCUSDT
- ETHUSDT
- SOLUSDT
- ZECUSDT
strategies:
ltf:
enabled: true
timeframe: 15m
# 15m forum topic id (integer from the Telegram topic URL)
telegram_thread_id: 51247
fvg:
enabled: true
timeframe: 6h
# 6h forum topic id — replace with the other thread
telegram_thread_id: 61476