Load market data: CSV, Bitget, OANDA, and samples

Every data source in QUIMERIA-HYPERION is normalized by backend/data_connectors.py into a single canonical OHLCV format before entering the SMK pipeline. This means you can switch between a CSV file exported from MetaTrader and a live Bitget REST feed without changing any pipeline configuration.

Normalized OHLCV format

All loaders return a JSON array of bar objects. Timestamps are Unix seconds (UTC). Volume falls back to 100 if the source does not provide it.

[
  {
    "time":   1705276800,
    "open":   1.10512,
    "high":   1.10601,
    "low":    1.10488,
    "close":  1.10557,
    "volume": 1245.0
  }
]

The pipeline calls _normalize_bars() on every dataset and deduplicates bars by timestamp before loading. Bars are always sorted ascending by time.

Data sources overview

Source	Format	Endpoint
MT4/MT5 CSV export	Dot-separated date with optional TICKVOL/SPREAD columns	`POST /api/load/csv`
TradingView ISO export	ISO 8601 timestamp (`2024-01-15T00:00:00+00:00`)	`POST /api/load/csv`
TradingView US export	US date format (`01/15/2024 00:00`)	`POST /api/load/csv`
Dukascopy UTC format	`DD.MM.YYYY HH:MM:SS.mmm UTC`	`POST /api/load/csv`
Bitget REST API	Spot candle endpoint, auto-paged	`POST /api/load/bitget`
OANDA REST API	Granularity-based candle fetch	`POST /api/load/oanda`
Built-in sample	Synthetic 500-bar AMD cycle	`POST /api/load/sample`

CSV uploads

The /api/load/csv endpoint accepts any of the four supported CSV dialects. The format detector reads the header line and first data row to pick the correct parser automatically.

MT4 / MT5 format

MT4 exports use a dot-separated date (2024.01.15) and separate DATE and TIME columns. MT5 adds TICKVOL, VOL, and SPREAD columns — all are accepted and the volume field is mapped to tickvol or vol whichever is present.

<DATE>,<TIME>,<OPEN>,<HIGH>,<LOW>,<CLOSE>,<VOL>
2024.01.15,00:00,1.10512,1.10601,1.10488,1.10557,1245

DATE,TIME,OPEN,HIGH,LOW,CLOSE,TICKVOL,VOL,SPREAD
2024.01.15,00:00:00,1.10512,1.10601,1.10488,1.10557,1245,0,2

TradingView ISO format

TradingView’s ISO export uses a single time column with a full ISO 8601 timestamp including timezone offset.

time,open,high,low,close,volume
2024-01-15T00:00:00+00:00,1.10512,1.10601,1.10488,1.10557,1245

TradingView US format

TradingView’s US date export uses DATE and TIME as separate columns with MM/DD/YYYY date notation.

DATE,TIME,OPEN,HIGH,LOW,CLOSE,VOLUME
01/15/2024,00:00,1.10512,1.10601,1.10488,1.10557,1245

Dukascopy UTC format

Dukascopy exports combine date, time, and milliseconds into a single UTC column.

UTC,OPEN,HIGH,LOW,CLOSE,VOLUME
19.03.2026 12:00:00.000 UTC,1.10512,1.10601,1.10488,1.10557,1245

Uploading a CSV file

curl
Python

curl -X POST http://localhost:8000/api/load/csv \
  -F "file=@/path/to/EURUSD_H1.csv"

import httpx

with open("EURUSD_H1.csv", "rb") as f:
    r = httpx.post(
        "http://localhost:8000/api/load/csv",
        files={"file": ("EURUSD_H1.csv", f, "text/csv")},
    )
print(r.json())
# {"bars": 2340, "symbol": "EURUSD", "status": "ok"}

Column name aliases

The normalizer accepts multiple column name variants for each OHLCV field:

Field	Accepted column names
`time`	`time`, `timestamp`, `UTC`, `datetime`, `date`, `Datetime`, `Date`
`open`	`open`, `Open`, `o`, `OPEN`
`high`	`high`, `High`, `h`, `HIGH`
`low`	`low`, `Low`, `l`, `LOW`
`close`	`close`, `Close`, `c`, `CLOSE`
`volume`	`volume`, `Volume`, `tickvol`, `TickVol`, `vol`, `VOL`, `v`

Bitget REST API

The /api/load/bitget endpoint fetches historical candles from the Bitget v2 spot market API. The granularity maps to Bitget’s internal interval strings automatically.

Interval	Bitget string
`1m`	`1min`
`3m`	`3min`
`5m`	`5min`
`15m`	`15min`
`1h`	`1H`

curl
Python

curl -X POST http://localhost:8000/api/load/bitget \
  -H "Content-Type: application/json" \
  -d '{
    "symbol": "EURUSDT",
    "granularity": "1h",
    "limit": 500
  }'

import httpx

r = httpx.post(
    "http://localhost:8000/api/load/bitget",
    json={
        "symbol":      "EURUSDT",
        "granularity": "1h",
        "limit":       500,
    },
)
data = r.json()
print(f"Loaded {data['bars']} bars")

Public Bitget candle endpoints do not require an API key. Set BITGET_API_KEY in your .env file if you need access to private endpoints or higher rate limits.

OANDA REST API

The /api/load/oanda endpoint fetches candles from the OANDA v20 REST API using your account credentials. The granularity parameter maps directly to OANDA’s granularity strings (e.g., H1, M15, D).

OANDA requires a live or practice account API token. Never commit your token to version control — store it as OANDA_API_KEY in your .env file.

curl
Python

curl -X POST http://localhost:8000/api/load/oanda \
  -H "Content-Type: application/json" \
  -d '{
    "instrument": "EUR_USD",
    "granularity": "H1",
    "count": 500,
    "api_key": "YOUR_OANDA_TOKEN",
    "account_type": "practice"
  }'

import httpx, os

r = httpx.post(
    "http://localhost:8000/api/load/oanda",
    json={
        "instrument":   "EUR_USD",
        "granularity":  "H1",
        "count":        500,
        "api_key":      os.environ["OANDA_API_KEY"],
        "account_type": "practice",   # or "live"
    },
)
print(r.json())

Built-in sample data

The /api/load/sample endpoint generates a synthetic 500-bar AMD (Accumulate → Manipulate → Distribute → Retrace) price cycle. It requires no credentials and is useful for verifying your pipeline configuration before connecting a live feed.

curl
Python

curl -X POST http://localhost:8000/api/load/sample

import httpx

r = httpx.post("http://localhost:8000/api/load/sample")
data = r.json()
print(f"Sample loaded: {data['bars']} bars")
# Sample loaded: 500 bars

Use the sample data source when first deploying QUIMERIA-HYPERION to confirm that all 18 SMK detector modules load correctly and the dashboard renders the sensor panel before you connect real market data.

Verifying a loaded dataset

After any successful load, call GET /api/status to confirm the bar count, symbol, and pipeline state:

curl http://localhost:8000/api/status

{
  "status": "ready",
  "bars_loaded": 500,
  "symbol": "EURUSD",
  "pipeline": "initialized",
  "modules_ok": 18,
  "import_errors": 0
}

Getting Started

Architecture

Core Modules

Integrations

Operations

Load market data: CSV, Bitget, OANDA, and samples

Normalized OHLCV format

Data sources overview

CSV uploads

MT4 / MT5 format

TradingView ISO format

TradingView US format

Dukascopy UTC format

Uploading a CSV file

Column name aliases

Bitget REST API

OANDA REST API

Built-in sample data

Verifying a loaded dataset

Build docs developers (and LLMs) love

Getting Started

Architecture

Core Modules

Integrations

Operations

Documentation Index

​Normalized OHLCV format

​Data sources overview

​CSV uploads

​MT4 / MT5 format

​TradingView ISO format

​TradingView US format

​Dukascopy UTC format

​Uploading a CSV file

​Column name aliases

​Bitget REST API

​OANDA REST API

​Built-in sample data

​Verifying a loaded dataset

Build docs developers (and LLMs) love

Normalized OHLCV format

Data sources overview

CSV uploads

MT4 / MT5 format

TradingView ISO format

TradingView US format

Dukascopy UTC format

Uploading a CSV file

Column name aliases

Bitget REST API

OANDA REST API

Built-in sample data

Verifying a loaded dataset