AURELIA is an autonomous trading platform for research, validation, risk control, execution, broker verification, reconciliation, and operational readiness.
- Canonical branch:
main - Repository:
Anto-30/aurelia-trading - Live-release control:
config/LIVE_LOCK.yaml - Repository map:
AURELIA_REPOSITORY_MAP.md - Source-of-truth metadata:
AURELIA_SOURCE_OF_TRUTH.json - Operational readiness engine:
runtime/ops/readiness_orchestrator.py - Operational handoff:
docs/AURELIA_OPERATIONAL_HANDOFF_2026-10-03.md
The current repository state is authoritative for the implementation. Historical archives and old branches are not the current source of truth.
Current operational state: docs/AURELIA_CURRENT_OPERATIONAL_STATE_2026-10-08.md. The repository is browsable from main; remote GitHub read/write access has been verified for the current engineering integration. This does not grant capital authority.
- Capital plane:
capital/ - Execution boundary:
execution/ - Broker/runtime adapters:
runtime/adapters/ - Broker executor:
runtime/broker/ - Runtime controls and invariants:
runtime/core/ - Strategy governance:
runtime/strategy/ - Validation:
runtime/validation/ - Assurance:
assurance/ - Research:
research/ - Runtime entry point:
runtime/main.py - Tests:
tests/ - Deployment policy:
config/deployment_policy.json - Persistent deployment workflow:
.github/workflows/self-hosted-runtime-deploy.yml - Host deployment script:
scripts/deploy/bootstrap_and_deploy.sh - Evidence and reports:
reports/anddocs/ - Intelligence routing:
config/intelligence_source_routing.json
The canonical Options API path is the current Deriv API under api.derivws.com.
- Authorized REST credential.
- Exact account inventory and binding.
- Fresh account-specific OTP.
- Authenticated Options WebSocket.
- Session/account/currency/environment verification.
- Fresh broker balance.
- Capital snapshot.
Detailed path definition: docs/AURELIA_MODERN_DERIV_PATH_2026-10-03.md.
Legacy ws.derivws.com/Binary WS v3 endpoints are diagnostic-only and must not be used as a production fallback.
The Deriv integration must never log or commit PATs, OAuth tokens, OTPs, authenticated URLs, or other secret values.
Research and orchestration do not have capital authority. The capital plane is the only component allowed to authorize and submit capital-moving actions.
The repository preserves Risk Warden, LivePolicy, Capital Plane Gate, Execution Firewall, Account Isolation, idempotency, transaction verification, reconciliation, watchdog/kill-switch behavior, and fail-closed UNKNOWN handling.
The live lock is currently non-live by design. Passing tests or authenticated-session verification does not by itself authorize live capital.
The hard probability band is 50%–75%, inclusive, with no clipping. A probability inside the band is insufficient unless calibration, freshness, drift, instrument qualification, and positive net expected value after costs all pass.
The starting stake and contract minimum are USD 1.00. For live purchased contracts, AURELIA requires a verified USD account and computes the stake as 1% of fresh, verified available USD balance, rounded down to cents. The strategy/agent proposal cannot override the deterministic sizing rule. Since the minimum stake is USD 1.00, a 1% budget cannot fund even the starting stake until verified available balance is at least USD 100.00. Non-USD accounts and lower balances must remain blocked rather than force a currency or risk-budget exception.
Full policy: docs/AURELIA_STAKE_POLICY_2026-10-10.md.
Run the assurance suite with:
python -m unittest discover -s assurance -p 'test_*.py' -v
Run runtime tests with:
python -m unittest discover -s tests -p 'test_*.py' -v
Run the readiness report with:
python -m runtime.ops.readiness_orchestrator
Run the non-trading Deriv authentication verifier when its authorized environment secrets are configured:
python scripts/verify_deriv_session.py
A readiness report must not be interpreted as live authorization unless every mandatory control is independently verified and config/LIVE_LOCK.yaml permits capital movement.
- Current non-secret infrastructure verification:
data/runtime/AURELIA_EXTERNAL_ACCESS_STATE.json - This artifact contains historical external-access observations and must not be interpreted as capital authority.
- GitHub repository-level push capability is currently available to the connected engineering integration used for this verification.
- Railway is no longer an active AURELIA deployment dependency; its historical rejection remains documented only for auditability.
- The active deployment path is provider-neutral self-hosting on a user-controlled Linux/Docker host.
AURELIA maintains a governed registry at config/external_repo_federation.json for the external repositories used as research, engineering, tooling, or reference sources. These sources are not vendored into the capital or execution planes. The registry assigns each source a mode (RESEARCH_ONLY, SANDBOX_ONLY, TOOLCHAIN_ONLY, or REFERENCE_ONLY) and maps sources to the appropriate agents.
The weekly .github/workflows/external-repo-sync.yml workflow checks public reachability and records current head commits as an artifact. External source updates do not automatically become AURELIA dependencies or change execution behavior.
AURELIA also maintains config/intelligence_source_routing.json for bounded routing of external intelligence and engineering services.
The registry assigns specialist use to GitHub, CodeRabbit, Next Stock Outlook, The Fly Market Intelligence, Sixtyfour Intelligence, Code Tytor: Python, Notion, and Outlook/Email. These sources are non-authoritative: they may generate research, evidence, code-review findings, operational messages, or documentation, but they cannot authorize capital, modify LIVE_LOCK, read production secrets, submit broker transactions, or override deterministic validation.
The regression contract is assurance/test_intelligence_source_routing.py. External intelligence must follow the boundary:
source -> specialist agent -> normalized evidence -> deterministic validation -> existing release gate.
AURELIA maintains a pinned cross-agent engineering capability registry at config/agent_skill_federation.json. The registry covers Claude Code, Kimi K3/Kimi Code, Grok Bot, Google Agent Skills, GLM Skills, and Playwright CLI. All external capabilities are advisory/engineering-only and cannot authorize or submit capital-moving actions.
Task-aware role routing is defined in config/agent_tool_skill_assignment.json. It assigns each canonical agent its primary tools, skill bundle, eligible plugins, independent reviewer, and security boundary. The audit is enforced by scripts/validate_agent_tool_skill_assignments.py in CI. A plugin or source appearing in a registry does not prove it is connected or installed in a runtime. Dev remains inactive until an authorized device is connected and verified. The external-repository intake manifest is config/external_repo_intake_2026-10-11.json; entries are pinned references pending review, not production dependencies.