Skip to content

Project

Grilling — decision tree

Status: rounds 1 and 2 answered; Q7–Q9 were not answered individually. The later instruction to implement everything without interruptions delegated implementation choices. Defaults adopted are in...

Status: rounds 1 and 2 answered; Q7–Q9 were not answered individually. The later instruction to implement everything without interruptions delegated implementation choices. Defaults adopted are in ADR-0011 and in the ROADMAP; they are not invented answers from the owner. Implementation started under that authorization. No pre-selected option is accepted. The document remains a draft until there is shared understanding.

Method#

Ask only decisions whose prerequisites are resolved. Give a recommendation and phrase the question so that "yes" accepts it. Facts about tools, code and providers are researched, not delegated to the user. Do not implement choices still open; requested documentation may advance with clear states.

Decisions already established#

Go/Vukem + Inertia/React/Tailwind/PostgreSQL; / as dashboard; Google as the only login; visitor goes to /login; shadcn/ui sidebar; Railway.

Round 1 — answered#

Q1 — Audience: can we start as a personal terminal, restricted to the owner's Google account? Recommendation: yes, personal in the MVP. Alternative: SaaS from the start. Answer: personal use confirmed.

Q2 — Market: can we start with US stocks only? Recommendation: yes, one market. Alternatives: B3, crypto, Brazil + US. Answer: Brazil and US confirmed.

Q3 — Mode: can we start with only backtesting and paper trading? Recommendation: yes; real trading comes in a later phase. Answer: backtesting and paper confirmed.

Dependency tree#

Text
flowchart TD
  Audience[Audience: personal or SaaS] --> Access[Allowlist or onboarding and isolation]
  Audience --> Secrets[Credential and authorization model]
  Market[Initial market] --> Broker[Eligible broker and feed]
  Broker --> Instruments[Assets, calendar, precision and history]
  Mode[Paper or real in the MVP] --> Release[Activation and operation criteria]
  Instruments --> Strategies[Timeframe, strategies and supported orders]
  Strategies --> Risk[Limits and closing policy]
  Secrets --> Operation[Budget, retention, alerts and recovery]
  Risk --> Operation
  Release --> Operation
  Operation --> Review[Joint review of SDDs and ADRs]

Round 2 — answered#

Q4 — US: can we use Alpaca Paper as the first integration, with a replaceable adapter? Recommendation: yes. Alternatives: IBKR or choose later. Answer: Alpaca Paper confirmed for the US.

Q5 — B3: can we start with backtest and an internal simulator, and connect an external paper account later after proving an API? Recommendation: yes. Alternative: require integration with a B3 broker in the MVP. Answer: B3 internal simulator confirmed.

Q6 — Access: can we allow one Google account, initially authorized by verified email and later bound to the stable subject? Recommendation: yes, no public sign-up. Alternative: other accounts by invitation. Answer: one authorized Google account confirmed.

The internal simulator uses fictitious money and its own fill model; it cannot be labeled as broker execution. The B3 data source remains open even with the internal simulator already accepted in Q5.

Round 3 — current frontier#

Q7 — Assets: can we start with spot stocks, long-only without leverage, watchlist PETR4, VALE3, ITUB4, AAPL, NVDA and TSLA? Recommendation: yes, configurable; options, futures and crypto outside the MVP. Answer: pending.

Q8 — Data: can we use IEX/Alpaca in the US and import historical B3 candles via CSV with provenance/license? B3 initially historical replay; real time requires its own provider. Recommendation: yes. Answer: pending.

Q9 — Time: can we use closed 5-minute candles, regular session and planned closing before the close? Recommendation: yes, respecting the calendars of both markets. Answer: pending.

Following rounds, not yet sent#

After Q1: owner authorized by identity, invitations, recovery, isolation per account, access profiles and eventual monetization. After Q2: broker, available data, feed budget, asset universe, currency and sessions. Only after that: strategies, frequencies and order types. After Q3: definition of paper, need for an external broker, live gates, reconciliation, supervision and risk blocks. After these choices: numeric limits, no overnight positions, session closing, handling of disconnections, alerts and authority to resume. Finally: retention, RPO/RTO, availability, Railway budget, schedule and acceptance of the implementation plan.

Answer log#

RoundDecisionAnswerImpact
1AudiencePersonalNo SaaS in the MVP; owner authorization mandatory
1MarketBrazil + USTwo calendars/currencies; adapters defined in round 2
1ExecutionBacktest + paperReal capital outside the MVP
2US brokerAlpaca PaperExternal paper adapter confirmed
2B3 simulationInternal initiallyDo not promise external B3 orders
2Google admissionOne authorized accountVerified bootstrap email + stable subject
3Universe and directionPendingPrecision, capabilities and exposure
3Feeds and historyPendingIEX and B3 CSV are proposals
3Frequency/sessionPendingStrategies and calendars

The interview is not finished yet. Dates in documents do not replace acceptances.