> ## Documentation Index
> Fetch the complete documentation index at: https://doc.blueapi.ir/llms.txt
> Use this file to discover all available pages before exploring further.

# Quickstart: Understand EMEP in Five Minutes

> Trace a single candidate through EMEP end to end: import, compatibility, merge, evaluate, evolve. A five-minute map of the whole platform.

This quickstart traces one hypothetical candidate through every stage of EMEP so you can navigate the specification in order. No code runs; this is a guided tour of the primary research loop and the pages that own each step.

## The 8 Stages

<Steps>
  <Step title="Discover and import a base model">
    ModelLoader fetches weights, tokenizer, and configuration from the source. Every artifact is hashed. Read [Model Import](/compatibility/model-import).
  </Step>

  <Step title="Analyze compatibility with a second parent">
    ModelCompatibilityAnalyzer verifies architecture, tokenizer, tensor shapes, and precision. Emits COMPATIBLE, CONDITIONALLY\_COMPATIBLE, or INCOMPATIBLE. Read [Model Compatibility](/compatibility/model-compatibility).
  </Step>

  <Step title="Select a merge strategy and execute">
    MergeEngine runs the selected strategy (Linear, SLERP, TIES, DARE, DARE+TIES, Task Arithmetic, Franken-Merge). Read [Merge Engine](/merge/merge-engine).
  </Step>

  <Step title="Validate the merged artifact">
    Numerical checks, dtype checks, integrity hash. Read [Merge Validation](/merge/merge-validation).
  </Step>

  <Step title="Register as a candidate">
    ModelRegistry assigns an ID and links parent lineage. Read [Model Registry](/tracking/model-registry).
  </Step>

  <Step title="Evaluate on the optimization split only">
    BenchmarkEngine runs the profile. Hidden Test Set is never touched here. Read [Evaluation Framework](/evaluation/framework).
  </Step>

  <Step title="Compute fitness and update population">
    FitnessEngine aggregates metrics. EvolutionEngine ranks, selects, mutates, crosses over. Read [Evolution Engine](/evolution/evolution-engine).
  </Step>

  <Step title="Promote, quantize, sign, and deploy">
    Best models are quantized, packaged, signed, and moved through staging to production, including the offline enterprise path. Read [Deployment](/deployment/deployment-specification) and [Offline Deployment](/deployment/offline-deployment).
  </Step>
</Steps>

## Data Isolation Guarantee

```mermaid theme={null}
flowchart LR
    OPT[Optimization Set] --> EVO[Evolution]
    VAL[Validation Set] --> SEL[Candidate Selection]
    HID[Hidden Test Set] --> FIN[Final Validation Only]

    EVO -.forbidden.-> HID
    SEL -.forbidden.-> HID

    style HID fill:#fff3e0,stroke:#c66
    style FIN fill:#e8f5e9,stroke:#393
```

The EvolutionEngine never sees the Hidden Test Set. This is enforced at the data access layer. Read [Benchmark Integrity](/evaluation/benchmark-integrity).

## Reading Order

<CardGroup cols={2}>
  <Card title="If you set architecture" icon="cubes" href="/architecture/system-architecture">
    System, components, runtime, storage, GPU, security, observability.
  </Card>

  <Card title="If you implement merges" icon="code-merge" href="/merge/merge-engine">
    Strategies, tensor operations, math, validation.
  </Card>

  <Card title="If you design experiments" icon="flask" href="/evolution/evolution-engine">
    Genome, population, selection, fitness, multi-objective search.
  </Card>

  <Card title="If you run production" icon="server" href="/operations/runbook">
    Runbook, monitoring, alerting, disaster recovery, GPU health.
  </Card>
</CardGroup>