aries-explorer

ARIES Explorer

What you're doing

The user has an ARIES database — the Microsoft Access .accdb/.mdb file behind an engineering shop's reserves and economics runs. It holds a property list, production history, decline forecasts, and per-property economic assumptions, all in ARIES's own table grammar. Your job is to read it, understand it, and show it: open the binary with the bundled scripts, decode what the database author set up, and build the explorer artifact — the database's cover page.

This is NOT the dataroom flow and NOT a valuation. dataroom-extract is for a document package headed to a bid; this is one database, read on its own. The forecasts and economics inside are the database author's claims — you are displaying them, never adopting them. If the user asks what the assets are worth, that is the valuation flow (get_skill("well-forecasting") → deal_forecast_wells), which builds its own forecasts from public data. The one sanctioned exception: when the user explicitly asks to value this database's own curves, that is get_skill("aries-to-valuation") — a separate skill that translates them with full attribution and its own guardrails.

What you need

Code execution is a hard requirement. An Access file is binary — there is no by-hand fallback. With no code execution available, say so honestly, explain that reading an ARIES database needs the sandbox, and suggest the user ask their engineer for CSV/Excel exports instead. Never pretend to read the binary.

ARIES.md (bundled) is the reference — the complete table map, the AC_ECONOMIC line grammar, units, escalation codes, stream numbers. Read it before reading any dump; go back to it whenever a keyword or unit is unfamiliar. Never guess at ARIES semantics the reference doesn't cover — show the line verbatim instead.

What you're building toward

One artifact, with all arithmetic done by the bundled scripts — never by you:

1. aries_triage.py opens the database (it resolves its own reader: mdb-tools if installed, else pip install access_parser), inventories every table, and dumps the load-bearing ones to _aries/tables/*.csv.
2. aries_payload.py decodes the dumps — qualifiers, reserve categories, forecast sources, assumption clusters, lookup tables, integrity checks — and emits the payload plus a --facts digest.
3. notes.json — the judgment layer: 3–8 short observations written by you AFTER reading the digest.
4. AriesViewer.jsx is the finished, frozen viewer (you do not build, redesign, or adapt it); fill DATA/TITLE/TLDR. A worked payload is in example.json.