ApexLab Desktop

ApexLab is a local-first Windows telemetry coach for EA SPORTS F1 25 Time Trial. It captures the game's official UDP stream on the same PC, reconstructs comparable laps, identifies repeatable corner-level time loss, proposes one evidence-based driving experiment, and measures whether later laps support it.

This is a personal software-engineering diploma project, not a startup MVP. Its technical focus is reliable binary protocol handling, bounded concurrent ingestion, crash-safe persistence, deterministic replay, spatial time-series analysis, explainable coaching, uncertainty, scientific visualization, and reproducible Windows delivery.

The native Windows foundation is executable and tested: it includes the WPF shell, validated local settings, single-instance lifecycle coordination, safe SQLite schema migration, headless package smoke testing, reproducible verification/packaging scripts, strict base-F1-25 packet-envelope validation, bounded loopback UDP reception, a privacy-safe live traffic probe, integrity-bound raw evidence capture, and deterministic replay. A private real-game F1 25 Time Trial run has passed the guided manifest, replay, sequence-gap, privacy, protocol-assumption, and at-least-2x rate gates; raw telemetry and private rates remain local. Lap reconstruction and coaching begin in v0.3.

• Product and engineering design
• 30-week roadmap
• v0.1.0 implementation plan
• Changelog

comparable laps
  -> personal reference
  -> repeated evidence-backed issue
  -> one bounded experiment
  -> matched later laps
  -> supported / inconclusive / rejected outcome

The coach is deterministic and evidence-first. It must abstain when data quality, context, repeatability, or confidence is insufficient. A learned personal ranker is optional and must beat the deterministic baseline before it can be used.

• Native Windows 11 application using C#, WPF, and .NET 10 LTS.
• Base F1 25, official F1 25 v3 UDP, Time Trial, wheel and pedals.
• Same-PC loopback capture on 127.0.0.1:20777 by default.
• Fully local capture, SQLite metadata, binary telemetry chunks, replay, analysis, coaching, backup, and export.
• Self-contained win-x64 package with no paid API or mandatory service.

There is no mobile/web client, backend, account system, smartwatch, IoT device, game injection, in-game overlay, or cloud dependency in version 1.0.

Prerequisites are Windows 11 x64, PowerShell 5.1 or newer, and an SDK accepted by the policy in global.json (currently 10.0.302 or a later patch in the 10.0.3xx feature band). Restore remains locked to committed NuGet graphs.

powershell -NoProfile -ExecutionPolicy Bypass -File scripts/Verify.ps1

Verification performs the SDK policy check, locked restore, formatting check, warnings-as-errors Release build, non-soak tests, TRX/Cobertura output, and repository hygiene checks. Generated test evidence is written below ignored artifacts/test-results.

Run the desktop shell from source with:

dotnet run --project src/ApexLab.App/ApexLab.App.csproj -c Release

Configure F1 25 for base F1 25 UDP mode at 127.0.0.1:20777, then run a short offline Time Trial while this command is active:

dotnet run --project tools/ApexLab.Replay/ApexLab.Replay.csproj -c Release -- probe --duration-seconds 30

The probe uses the same bounded UDP source, sender policy, F1 adapter, and ingestion coordinator planned for the desktop capture workflow. It retains no packet payload and prints aggregate JSON only: classifier counts, packet ID/version/length counts, rate buckets, session-UID cardinality, sequence/timestamp regressions, skipped frame-identifier values, frame regressions, and player-index ranges. Skipped frame identifiers are descriptive deltas per packet family, not packet-loss claims. The probe never prints an actual session UID, sender endpoint, local path, username, or payload byte.

Exit codes are 0 for compatible traffic, 2 for invalid arguments, 3 for bind failure, 4 for no traffic, 5 for incompatible-only traffic, 6 for interruption, and 7 for an unexpected failure. Optional arguments are --port, --capacity, --max-datagram-bytes, and --duration-seconds.

Use the guided validator when you are ready to test against your own F1 25 installation. Start F1 25 first, select an offline Time Trial, enable UDP using base F1 25 v3 at 127.0.0.1:20777, and keep ApexLab closed. From a clean feature-branch worktree, run:

powershell -NoProfile -ExecutionPolicy Bypass -File scripts/ValidatePrivateF125.ps1 -GameBuild "your F1 25 build"

The command performs these stages: preflight, probe, probeEvaluation, capture, captureSelection, privateValidation, rateGate, and safeSummary. Drive during the 30-second probe. When the native ApexLab window opens, arm and stop exactly one capture, then close ApexLab. The command validates that capture twice, checks sequence-gap preservation and privacy exclusions, and finishes with a measured synthetic rate gate of at least twice the observed real peak.

On failure, read the printed stage and correction, fix that item, and run the same command again. Raw telemetry is preserved locally, while temporary diagnostics are removed. A passing commit-safe result is stored at %LOCALAPPDATA%\ApexLab\private-validation\latest-safe.json; it contains no capture ID, hash, path, sender, session UID, payload, or exact private rate.

This procedure needs no phone, SIM card, cloud account, API key, paid service, or internet upload. The Android phone is not part of this Windows capture-validation step.

Create a guarded self-contained Windows package and verify the extracted executable:

powershell -NoProfile -ExecutionPolicy Bypass -File scripts/Publish.ps1 -Version 0.1.0 -CleanKnownOutputs
powershell -NoProfile -ExecutionPolicy Bypass -File scripts/SmokeTest.ps1 -PackagePath artifacts/ApexLab-0.1.0-win-x64.zip

The publish script refuses silent overwrites, verifies Version.props, includes the README and third-party notices, and writes ApexLab-0.1.0-win-x64.zip.sha256. The smoke script extracts to new temporary storage, launches no window, enforces an exact-PID timeout, initializes real settings and SQLite schema version 1, validates the result, and proves the temporary tree can be removed.

The package is currently unsigned and may show a Windows reputation/SmartScreen warning. It is a portable ZIP, not an installer. See the v0.1.0 verification record.

Development follows GitFlow: released states live on main, integration on develop, and work on short feature/* branches. Releases use release/*, merge back to develop, and receive annotated semantic-version tags. Commits use the repository owner's Git identity and describe coherent, working changes.

Real telemetry, local databases, logs, backups, exports, vendor specification downloads, credentials, and signing material must never be committed. Only small synthetic fixtures with provenance and checksums may enter tests. See data policy.

ApexLab is an unofficial educational project. It is not affiliated with, authorized by, sponsored by, or endorsed by Electronic Arts Inc., Formula One World Championship Limited, or their affiliates. Product names and trademarks belong to their respective owners. This repository does not distribute official logos, artwork, game assets, fonts, telemetry specification attachments, or other copyrighted vendor material.

No public source-code license is granted while university and intellectual-property requirements are being confirmed.