SigmaVault
SigmaVault handles the business logic that has to happen before funds move on-chain. The Python/FastAPI service ingests transactions from Firefly III (integrations/firefly/client.py) or CSV (pipeline/csv_adapter.py), normalizes them into a treasury model, and treasury/liquidity.py computes a deterministic assessment of operational reserves, upcoming obligations and deployable surplus. A Pydantic AI agent in treasury/treasury_agent.py reads that assessment through two read-only tools, inspect_treasury_position and inspect_arc_wallet, and returns preserve_liquidity, investigate or propose_payment with a written rationale and a human-review flag. Every recommendation becomes a SignedDecisionRecord with an Ed25519 signature appended to a Beancount journal (treasury/beancount_audit_journal.py), exportable from /api/audit/download and checkable with script/verify_audit_journal.py. integrations/arc/gateway.py wraps web3.py against Arc testnet with a chain-id check, native USDC balance reads and receipt lookups.
Backend Engineer | Math Graduate | Python, FastAPI, PostgreSQL, Docker. Contributor to Hugging Face Transformers
1 (Solo) — NDOWAH MARCEL AMBO
The Deterministic Autonomous Treasury Agent