Current state audit

Audit date: 2026-09-28, P00 import snapshot. This is a historical source audit, not a certification of runtime correctness or current implementation status. Later P01 foundation work and delivery-flow activation status are tracked in PROGRESS.

Provenance and scope

• Source: https://github.com/AlexandreZanata/brazil-fuel-prices-app
• Branch: main; clean clone at b8a52049e0294cd2d07612cedcc52b3017c5271e.
• Latest change: merge #6 (2026-09-21), structured station-address geocoding cascade. Preceded by nearest best-price station/radius work and direct navigation from home cards.
• 670 tracked files were imported using git archive from a complete clone. The new repository's existing .git and origin (AlexandreZanata/brazil-fuel-prices) were preserved. Upstream Git history is in the source repository, not grafted into the new repository.
• Destination initially had no commits or project files. Import does not create a commit, push, release or deployment.
• The supplied commercialization text is reference material. Its suggested ordering was changed to honor the user's explicit backend/infrastructure-first request. This delivery imports the base and creates planning documents; it does not implement the backend.
• No existing Android source, Gradle configuration, schema, application ID or license is changed by the planning work.

Actual stack and boundaries

settings.gradle.kts includes :domain, :application, :data, :app; package root is com.anpfuel, application ID com.anpfuel.app. App version is 3.1.0 / versionCode 4.

Versions in gradle/libs.versions.toml: Kotlin 2.0.21, AGP 8.7.3, Compose BOM 2024.12.01, Room 2.6.1, Hilt 2.53.1, WorkManager 2.10.0, OkHttp 4.12.0, Jsoup 1.18.3. Gradle wrapper is 8.10.2; Kotlin JVM toolchains request JDK 17; Android minSdk 26, compile/targetSdk 35. These are observed versions, not upgrade recommendations.

• domain: 108 production Kotlin files, 57 unit-test files. Entities, value objects, business rules, events and repository ports; no Android dependency. DomainId.generate() calls UUID directly: reuse concepts, not every implementation choice, for deterministic Go tests.
• application: 39 production Kotlin files, 29 unit-test files. Use cases and coroutine orchestration; depends on domain.
• data: 103 production Kotlin files, 30 unit-test files, 15 instrumentation-test files. Room, preferences, import/parser, geocoding and workers. Actual dependencies include BOTH domain and application.
• app: 80 production Kotlin files, 18 unit-test files, 9 instrumentation-test files. Compose, ViewModels, permissions, navigation and DI.

Counts are files matching *Test.kt, not executed test cases or coverage: 134 unit files and 24 instrumentation files total.

Implemented capabilities to preserve

UC-001…009 cover ANP discovery/import, onboarding, location selection, municipality search, averages, history, station details, settings and survey-week selection. UC-010…015 add up to three local vehicles, tank-fill estimates, optional device location, external navigation, local price-drop notifications and nearest best-price station selection.

Evidence paths:

• data/src/main/kotlin/com/anpfuel/data/parser/: custom streaming XLSX plus summary/station parsers; POI is test-only.
• data/src/main/kotlin/com/anpfuel/data/worker/SyncWorker.kt: WorkManager calls application sync use cases.
• data/src/main/kotlin/com/anpfuel/data/local/AnpFuelDatabase.kt: schema version 4, seven entity types; data/schemas/ has versions 1–4.
• data/src/main/kotlin/com/anpfuel/data/local/AnpFuelDatabaseMigrations.kt: 1→2, 2→3, 3→4 migrations.
• domain/.../PriceAmount.kt: BigDecimal rounded HALF_UP to two places; Room prices are Double/REAL.
• domain/.../Cnpj.kt: numeric-only normalization, exactly 14 digits, no check-digit validation.
• application/.../FindNearestBestPriceStationUseCase.kt: at most eight candidates, structured/free-text address cascade; BR-028 applies a 2% price band and configurable 3/5/10/15 km radius.
• data/samples/, parser tests, data/src/androidTest/: reusable input samples, import/idempotency and migration checks.

The current app is self-contained. It has no community backend, contributor authentication, photo upload, OCR contribution flow, billing verification or cloud synchronization.

Documentation reconciled with implementation

A01 — FTS mismatch. README, architecture and stack say FTS5. MunicipalityFtsEntity.kt uses Room @Fts4. Record FTS4 as baseline; do not replace search during backend work.

A02 — Schema/dependency examples are stale. docs/architecture.md contains a version-1 Room example and says data depends only on domain; actual schema is 4 and data also depends on application for worker orchestration. Keep the working dependency graph; reconcile prose in P01-T01.

A03 — Correction history differs from BR-003. Glossary promises a new history record for corrected ANP input. ADR-002 and the deterministic IDs plus OnConflictStrategy.IGNORE keep the first record. A reimport is idempotent but does not version corrections. Backend must use import revisions; Android must not silently acquire different semantics before P10.

A04 — Units and precision need an explicit bridge. PriceAmount has two decimals, while backend ingestion must preserve up to three price decimals and raw source precision. Do not migrate local REAL columns now. LPG_P13 is per 13 kg cylinder and CNG per cubic metre, not price per litre.

A05 — Product vocabulary carries a legacy misnomer. GASOLINE_PREMIUM currently maps GASOLINA ADITIVADA. Backend should use GASOLINE_ADDITIVED and a documented compatibility map, not silently claim premium/octane properties.

A06 — CNPJ cannot remain numeric-only. Current normalization drops letters. New catalog must support numeric and alphanumeric CNPJ, preserve leading zeroes, and quarantine invalid identifiers. The Receita Federal CNPJ program documents the new format. Android compatibility work is deferred to P10; no letter-to-digit coercion is acceptable.

A07 — Spreadsheet documentation is not a parser contract. docs/data-sources.md describes station header/data rows 7/8; the actual parser constants are 10/11. Fixed columns, sheet name and header assumptions need fixture-based layout detection in Go. Verify samples, not prose offsets.

A08 — Geocoding scope changed. ADR-003 describes one-shot reverse geocoding; UC-015 now performs forward-geocoding cascades. Per-device throttling does not establish a safe aggregate production load. Server-side bulk catalog geocoding requires a permitted provider and global quotas; public Nominatim is not the default bulk source.

A09 — Privacy documents describe the old product. Current policy says there is no app-owned backend, no photo collection and no account. It is valid context for the imported app, not a policy for the planned network. New backend data flows require a separately reviewed notice before a pilot.

A10 — Generic agent rules conflict with this product. Existing rules demand authentication/tenant identification for every API, events-only cross-context communication and tests without databases. The target has public reads, no tenants, explicit application interfaces and real PostgreSQL integration tests. ADR-004 resolves the scope; unit-test isolation remains mandatory.

A11 — Legacy publishing/docs references. README indexes only UC-001…014 although UC-015 exists; it links ignored .local/PROJECT_PLAN.md. Several documents reference the old TABELA-ANP-COMBUSTIVEIS name. Release scripts/descriptions still target v2.0.0 in places despite version 3.1.0. Do not run release/tag scripts to establish baseline.

A12 — License summary is imprecise. docs/license.md suggests a repository link alone suffices, while the actual MIT file requires retaining its notice. Preserve LICENSE verbatim and use it as the authoritative text; no relicensing is part of this task.

A13 — Secret scanner coverage is limited. Existing scanner covers tracked files and excludes documentation/test paths. On an unborn destination with untracked imports its success is not a meaningful full scan. It was also executed against the tracked upstream clone; broader changed-file scanning belongs in the new CI.

Validation evidence

• PASS: source clone clean; import file provenance checked against the source commit.
• PASS: bash scripts/validate-repo-baseline.sh in destination validates ignore patterns and required rule/docs files.
• PASS: bash scripts/scan-secrets.sh in tracked upstream clone, with its documented limited pattern/path coverage.
• Android unit/build baseline: attempted ./gradlew :domain:test :application:test :data:testDebugUnitTest :app:testDebugUnitTest :app:assembleDebug --no-daemon using an isolated Gradle cache. Initial attempt failed downloading Gradle due to sandbox networking. Final retry outcome is recorded in baseline evidence.
• Environment observed: active Java 21, requested toolchain 17 not installed in the inspected Java locations; SDK symlink resolves to /data/dev/android/sdk, but platform directory was absent. Do not claim APK/tests/coverage passed.
• Instrumentation, actual UI behavior, live ANP import, Play publishing and production security were not executed.
• Existing CI .github/workflows/ci.yml runs ./gradlew test --no-daemon with JDK 17; it does not explicitly run instrumentation, APK, lint, vulnerability scanning or a Go pipeline. Remote CI status was not used as a substitute for local results.

Reuse and migration risks

Preserve the four modules, 15 use cases, offline Room cache, navigation, local alerts, domain tests and sample spreadsheets. Port domain meaning through shared fixtures; Kotlin classes are not directly shared with Go. The highest-risk interfaces are fuel names/units, CNPJ identity, ANP revisions, local precision, and source/freshness presentation.

Start implementation with P01-T01 to close prerequisite and baseline gaps. Existing functionality is the regression reference. New Android functionality is blocked until the backend release gate G09.