document_id	AI4B-DOC-README-001
title	AI4BINANCE EnterpriseAI vNext Repository README
document_type	REGISTRY
version	1.0.1
status	ACTIVE
owner	Enterprise Knowledge Governance
authority_level	NORMATIVE
authority_layer	L5_REGISTRIES_ROADMAP
authority_scope	repository_readme
content_role	AUTHORITATIVE
source_of_truth	true
source_of_truth_scope	canonical
machine_enforceable	true
audit_required	true
classification	INTERNAL
canonical_path	README.md

AI4BINANCE EnterpriseAI vNext

ELI10

This repository is the governed AI4BINANCE workspace. It explains what the system can do, which evidence is real, and why live trading remains blocked unless every validation and risk gate passes.

AI4BINANCE reviews Binance Spot data, tests strategy ideas, and remains a research platform that does not open operations when evidence is insufficient. It does not guarantee profit. The canonical source for Core vNext governance, schema, entity/relationship rules, output format, and the policy-as-code backbone is: AI4BINANCE Core vNext Governance Framework. This backbone defines AI4BINANCE with a pyramid authority model, evidence-based fail-closed operation, LOOPS, Web Intelligence Radar, research-only development candidates, and the transparent digital-company operating model.

ELI10: What does the system do?

Think of it as a control tower:

1. It only retrieves market data from Binance.
2. Checks whether the data is broken or outdated.
3. Expert agents examine the same data snapshot.
4. The strategy engine only creates research candidates.
5. The risk engine writes all missing proofs as blockers.
6. Backtest, walk-forward, and tuning candidates are tested against historical data.
7. If the proof is insufficient, the result is NO_TRADE.

Public data -> immutable snapshot -> agents -> candidate -> risk
            -> backtest/walk-forward/tuning -> human review

Current Real Situation

Area	Status
Python	3.14.7
Agent Catalog	49 platform definitions; 34 logical analytical capabilities
Technical Analysis	10 core, 23 advanced, and 1 trend-events agents
Strategy	20 playbooks; 10 calculation-generating playbooks
Supertrend	ATR14 + OHLC4 + multiplier 2, research-only
Backtest	Event-based Spot long engine is available
Walk-forward/OOS	Code is present; real long-term evidence is missing
Tuning	Limited and human-validated governance is present
Research integrity	Kline/aggTrade revision, scalable integrity, queue-fill replay, SPA/MCS, Monte Carlo and CPCV are present
Paper lifecycle	Present; automatic CLI command is not
Wallet	Salt-read HMAC REST adapter is present, not CLI-dependent
WebSocket/Ed25519	Signing, session.logon, read-only account/orders and gate-enforced order methods are ready; real testnet/production session authority is pending
Live order	Adapter shell is present; real testnet authentication and end-to-end live write proof are missing; blocked

Verified quality baseline:

Pytest pass count: evidence-bound
Coverage: evidence-bound
Ruff format/lint: evidence-bound
MyPy: evidence-bound
Bandit: evidence-bound
Financial/privacy leak guards: evidence-bound

Machine-readable current quality evidence: runtime/artifacts/quality/gate/latest.json. This file contains pytest_pass_count, coverage_percent, coverage_source, execution_allowed=false, promotion_status=RESEARCH_ONLY, and live_eligibility_status=LIVE_ORDER_BLOCKED.

Setup

python -m venv .venv
.\.venv\Scripts\python.exe -m pip install -r requirements.txt

Step-by-step usage

1. First, check the quality

powershell.exe -NoProfile -ExecutionPolicy Bypass -File `
  .agents\skills\quality-gate-loop\scripts\invoke_gate.ps1 `
  -RepositoryRoot (Resolve-Path .).Path

2. View secure default state

.\.venv\Scripts\python.exe -m ai4binance.cli status

Expected secure result:

NO_TRADE
LIVE_ORDER_BLOCKED

3. View Agent Catalog

.\.venv\Scripts\python.exe -m ai4binance.cli agents

3a. QAQC-Agent continuous improvement report view

.\.venv\Scripts\python.exe -m ai4binance.cli qaqc-agent

QAQC-Agent, 5S, Hoshin Kanri, Kaizen, Six Sigma, and Poka-Yoke via Generates a report-only edit/suggestion; does not issue an order or live permission.

4. Run Public Spot analysis

.\.venv\Scripts\python.exe -m ai4binance.cli analyze-public

Do not use API key or order permission.

WHALE-FUSION Phases 1-2-3-4-5-6-7-8

Think of this motor as a shared data dictionary of three separate detectors. First three phases:

1. Whale Fusion identifies the names and source evidence of social media and derivative events.
2. Binance USD-M public data source for OI, long/short ratios, funding, taker flow, Reads the difference between mark/index, order-book balance, and large orders.
3. OI change, z-score, percentile, and price-OI regime are deterministically calculated.
4. Validate the on-chain transfer from the provider; Binance, DEX, bridge, staking, token-unlock, market-maker, stablecoin, new wallet and partial Separates transfer events.
5. Validate canonical posts from social accounts in the allowlist; category, event, stance, entity, freshness, duplicate, and conflict checks applies these checks.
6. On-chain, social, and derivatives channels with time-decay and fixed weights merges; averages the duplicates of the same channel and applies the conflict penalty.
7. Binds the fusion result to an immutable market snapshot with the same snapshot_id, Assigns research-only whale and records it in an idempotent JSONL audit.
8. Application service same cycle's canonical proofs fusion, audit, snapshot, Orchestrator passes through the reporting chain; produces a secure CLI output.

This data is only auxiliary research evidence for the spot decision. On its own, it is a signal, Does not generate risk approval or order authority. Phase 4 does not connect to the external provider; provider Secure kernel that will be used by the adapters. Phase 5 for real social platform Not applicable; verifies events explicitly labeled by the provider adapter. Phase 6 combined generates a score but this score is not a final spot signal or execution permission.

Default fusion weights:

on-chain 40% | social 25% | derivatives 35%

Returns INDEPENDENT_FUSION_CHANNELS_INSUFFICIENT if there are fewer than two independent channels. Fusion agent only produces supplementary evidence; confluence, validation and NO_TRADE security flow does not alter ownership.

Secure Phase 8 smoke command without a provider:

.\.venv\Scripts\python.exe -m ai4binance.cli whale-fusion-research

Default empty cycle, neutral score and INDEPENDENT_FUSION_CHANNELS_INSUFFICIENT blocker with NO_TRADE is returned.

External Intelligence & Evidence Fabric

EIEF connects X, News, Reddit, Telegram, GitHub, Security, and Regulatory radars to a shared evidence, validation, manipulation/copy scoring, fusion, and advisory risk-impact contract. The first MVP returns explicit DATA_UNAVAILABLE when a provider is absent and connects GitHub Radar to the existing capability-first engine through an adapter.

.\.venv\Scripts\python.exe -m ai4binance.external_intel scan --symbol HOTUSDT
.\.venv\Scripts\python.exe -m ai4binance.cli external-intel --symbol HOTUSDT
.\.venv\Scripts\python.exe -m ai4binance.cli external-intel-open-web --format json

EIEF sends this information to the evidence factory; it is not a trade signal or order engine. Open Web Radar discovers allowlisted RSS/Atom feeds and seed URLs without collecting hosted LLM or API tokens; it does not store full article text and keeps all outputs within RESEARCH_ONLY and LIVE_ORDER_BLOCKED boundaries.

5. Archive and verify data

.\.venv\Scripts\python.exe -m ai4binance.cli archive-public
.\.venv\Scripts\python.exe -m ai4binance.cli validate-research

Security Rules

• The default mode is paper/manual.
• Spot SELL can only reduce the available inventory; there is no naked short.
• Agents cannot send orders.
• A method cannot be a hard gate if it does not have an OOS proof.
• Private keys and API secrets must not be logged or added to Git.
• --confirm-live by itself does not issue any command.

Documents

• ELI10 document map
• Core vNext Governance Framework
• Architecture
• Roadmap
• Compliance matrix
• External Intelligence & Evidence Fabric
• Backtest explanation
• Secrets security

Live compatibility

execution_allowed=false
LIVE_ORDER_BLOCKED