Welcome to F1 StratLab

Open-source multi-agent system that fuses seven machine-learning models, six LangGraph sub-agents and one strategy orchestrator into a single Formula 1 strategy recommender. Shipped under Apache-2.0.

This is the canonical technical reference for the F1 StratLab codebase. It is hand-curated and complements two sibling resources: the public landing at f1stratlab.com tells the project story for non-technical visitors, and the auto-generated DeepWiki gives a notebook-per-notebook tour of the source tree. The pages here focus on the narratives those two cannot: how the layers connect, why the contracts look the way they do, and what to do when something breaks.

The system at a glance

graph TD
    subgraph Sources["Race data"]
        FF[FastF1 sessions]
        OF1[OpenF1 radios]
        RCM[Race Control messages]
    end

    subgraph Engine["Replay engine"]
        RRE[RaceReplayEngine]
        RSM[RaceStateManager]
    end

    subgraph Models["ML core · 7 models"]
        M1[Pace XGBoost · N06]
        M2[Tire TCN · N07-N10]
        M3[Overtake LightGBM · N12]
        M4[Safety-car LightGBM · N14]
        M5[Pit duration HistGBT · N15]
        M6[Undercut LightGBM · N16]
        M7[Circuit clusters · N30]
    end

    subgraph Agents["LangGraph sub-agents · 6"]
        A1[Pace · N25]
        A2[Tire · N26]
        A3[Race Situation · N27]
        A4[Pit Strategy · N28]
        A5[Radio · N29]
        A6[RAG]
    end

    ORCH[N31 Strategy Orchestrator]

    subgraph Surfaces["Operator surfaces"]
        CLI[Headless CLI]
        ARC[Arcade dashboard]
        STR[Streamlit web app]
    end

    Sources --> RRE
    RRE --> RSM
    RSM --> ORCH
    Models -.consumed by.-> Agents
    Agents -->|structured outputs| ORCH
    ORCH -->|recommendation| CLI
    ORCH -->|recommendation| ARC
    ORCH -->|recommendation| STR

Three layers carry the system from raw telemetry to a strategy call: a machine-learning core, a multi-agent reasoning layer, and three operator surfaces. Each layer lives behind a documented contract, so any one can be swapped without disturbing the others.

What lives where

The narratives on this site stop at the contract level. For per-file deep-dives — every function in src/agents/, every notebook from N06 to N34, every helper in src/arcade/ — jump to the F1 StratLab DeepWiki. It is regenerated on every push to main.

Project status

The current focus is documentation polish for the thesis defence; see the project changelog for the full history.