deepalgo-sovereign 0.2.2

Python SDK for the DeepAlgo Sovereign AI Regime API

DeepAlgo Sovereign SDK

Python client for the DeepAlgo Sovereign AI Regime API — institutional-grade regime classification, entry signal quality, session-stratified causal signal quality, and order-flow intelligence across FX, metals, indices, and commodities.

Installation

pip install deepalgo-sovereign

Quick Start

1. Get a free API key

from deepalgo_sovereign import DeepAlgoClient

DeepAlgoClient.enroll("[e-mail]")
# Your API key arrives in your inbox within seconds — no card required.

2. Check a regime

client = DeepAlgoClient(api_key="your_key_here")

regime = client.verify_regime("EUR_USD")
print(regime["is_favorable"])                        # True
print(regime["veto_reason"])                         # None  (or a veto code)
print(regime["metrics"]["regime"])                   # "MEAN_REVERTING"
print(regime["metrics"]["composite_confidence"])     # 0.7843
print(regime["metrics"]["vpin_at_time"])             # 0.31
print(regime["stability_score"])                     # "82.4%"

3. Check session signal quality

sq = client.get_session_signal_quality("EUR_USD")
print(sq["current_session"])     # "Asian"
print(sq["current_quality"])     # "SPURIOUS"
print(sq["current_multiplier"])  # 1.3

# EUR/USD in Asian hours: 99.6% of raw OFI signal is macro noise — not structural.
# USD/JPY in Asian hours: only 2.7% spurious — Tokyo institutional flow is genuine.

# All three sessions at a glance:
for session, data in sq["sessions"].items():
    print(f"{session}: {data['quality']} ({data['ic_reduction_pct']:+.1f}% IC change)")

4. Check entry signal quality

signal = client.get_entry_signal("EUR_USD")
print(signal)
# {
#   "entry_quality": "HIGH",
#   "current_ou_z": 1.42,
#   "ou_z_threshold": 0.75,
#   "recommended_delay_candles": 3,
#   "best_confirmation_signal": "rsi_divergence",
#   "calibrated": True
# }

if signal["entry_quality"] in ("HIGH", "MEDIUM"):
    regime = client.verify_regime("EUR_USD")
    if regime["is_favorable"]:
        # statistically sound entry — proceed
        pass

API Reference

DeepAlgoClient.enroll(email)

Class method. Registers for a free API key — no account creation required. Key is emailed within seconds.

DeepAlgoClient.enroll("[e-mail]")

client.verify_regime(asset, use_tlock=True)

Returns the AI regime verdict for an instrument.

Parameter	Type	Description
asset	str	OANDA instrument name, e.g. "EUR_USD", "XAU_USD"
use_tlock	bool	Enable/disable the third validation layer (default: True)

Response fields:

Field	Type	Description
is_favorable	bool	True if all three locks pass
veto_reason	str | None	Gate that blocked — REGIME_VETO, AI_PROXY_VETO, TRANSFORMER_VETO, ENTROPY_VETO, or None
stability_score	str	Lock 2 neural stability as a percentile, e.g. "82.4%"
divergence	str	Divergence label — "NONE" when regime is clean
metrics.regime	str	HMM state — "MEAN_REVERTING", "TRENDING", "HIGH_VOL"
metrics.composite_confidence	float	Combined confidence after all scalars applied
metrics.entropy	float	Shannon entropy of return distribution (0–1)
metrics.vpin_at_time	float	VPIN at scan time — > 0.80 = predatory flow
metrics.entropy_scalar	float	Confidence multiplier from entropy (1.0 = no reduction)
metrics.vpin_scalar	float	Confidence multiplier from order-flow toxicity
metrics.macro_z_score	float	Z-score vs macro driver spread
metrics.lock3_score	float	Transformer sequence validation score

client.check_lock(instrument, injected_data=None)

Binary EXECUTE / VETO decision. Use this in your order router immediately before placing a trade — reads pre-computed state from Redis so response is sub-millisecond on the server side.

decision = client.check_lock("EUR_USD")
if decision["authorized"]:
    place_order()

Field	Type	Description
authorized	bool	True = EXECUTE, False = BLOCK
decision	str	"EXECUTE" or "BLOCK"
forensics.failed_lock	str	Which lock blocked, or "NONE"
forensics.vpin_toxicity	float	Current VPIN — > 0.75 = toxic
forensics.pip_multiplier	int	Correct pip factor for this instrument
forensics.execution_intelligence.execution_benchmarks.vwap_deviation_bps	float	Distance from VWAP in basis points
forensics.execution_intelligence.capacity_metrics.suggested_max_clip_usd	float	Max recommended trade size

client.get_session_signal_quality(instrument)

Returns the DML-calibrated session signal quality for an instrument. Based on a Double Machine Learning study across 32 G10 FX, metals, and bond instruments — identifies whether order flow imbalance carries structural predictive content in the current session, or whether it is predominantly spurious macro co-movement.

Field	Type	Description
current_session	str	"Asian", "London", or "NewYork"
current_quality	str	"STRUCTURAL", "MIXED", "SPURIOUS", or "UNKNOWN"
current_multiplier	float	Lock 3 threshold multiplier currently active (1.0, 1.1, or 1.3)
sessions	dict	Per-session breakdown: quality, multiplier, ic_reduction_pct, structural_ic
last_calibrated	str	Date of most recent DML calibration run

Multiplier guide:

Quality	Multiplier	Meaning
STRUCTURAL	1.0×	OFI signal is genuine — standard threshold
MIXED	1.1×	Partial spurious component — threshold raised 10%
SPURIOUS	1.3×	Predominantly macro noise — threshold raised 30%

Research note: The Asian session finding is the sharpest result. EUR/USD in Asian hours (00–08 UTC) loses 99.6% of its apparent OFI signal to confounder exposure — it is almost entirely spurious. USD/JPY in the same session loses only 2.7% — Tokyo institutional participation is genuine. The difference is not in the signal. It is in who is trading. Source: DeepAlgo Sovereign Research, Bletchley Intelligence Ltd.

client.get_entry_signal(instrument)

Returns OU-calibrated entry timing quality. Derived from 60-day IC analysis across 76 instruments — identifies whether the current moment is a statistically optimal entry point, not just whether the regime is right.

Field	Type	Description
entry_quality	str	HIGH, MEDIUM, LOW, or NO_DATA
current_ou_z	float	Live OU z-score — deviation from equilibrium in σ
ou_z_threshold	float	Calibrated threshold from IC analysis
recommended_delay_candles	int	Optimal M5 candle wait after signal touch
best_confirmation_signal	str	Highest-IC confirmation signal at optimal lag
calibrated	bool	False if instrument has no calibration data yet

client.validate_sequence(asset, candles)

Runs the full validation stack using 60 M5 candles of OHLCV data.

result = client.validate_sequence("EUR_USD", candles=[...])  # list of 60 OHLCV dicts
# {"is_authorized": True, "verdict": "AUTHORIZED"}

Supported Instruments

The API covers 76+ instruments across four asset classes:

• FX: EUR_USD, GBP_USD, USD_JPY, EUR_JPY, GBP_JPY, AUD_USD, and 30+ major/minor/exotic pairs
• Metals: XAU_USD, XAG_USD, XAU_EUR, XAU_JPY, XAU_AUD
• Indices: SPX500_USD, NAS100_USD, US30_USD, EU50_EUR, DE30_EUR, JP225_USD
• Commodities: NATGAS_USD, SUGAR_USD, WHEAT_USD, SOYBN_USD, CORN_USD

Tier Access

Feature	Free	Professional	Institutional
Instruments	4	15	All 76+
API calls / day	50	Unlimited	Unlimited
verify_regime()	✓	✓	✓
get_entry_signal()	—	✓	✓
get_session_signal_quality()	—	✓	✓
OU-calibrated entry timing	—	✓	✓
DML session quality (32 instruments)	—	✓	✓
Institutional order-flow VPIN	—	—	✓
Cross-client Network Intelligence	—	—	✓

Upgrade at gateway.deepalgo.co.uk

Health Check

import requests
r = requests.get("https://api.deepalgo.co.uk/health")
print(r.json())

Support

• Documentation: deepalgo.co.uk
• Email: [e-mail]

---

DeepAlgo Sovereign SDK v0.2.2 — MIT License