# EMEP

- [EMEP: Evolutionary Model Engineering Platform](https://doc.blueapi.ir/index.md): EMEP creates new high-performing LLMs from existing open-weight models via compatibility analysis, tensor merging, evaluation, and evolutionary search.
- [Quickstart: Navigate the EMEP Documentation](https://doc.blueapi.ir/quickstart.md): How to read the EMEP specification set, what to read first for each role, and the documentation quality gates that must pass before implementation.
- [EMEP Context: What the Platform Is and Is Not](https://doc.blueapi.ir/overview/context.md): Define EMEP's purpose as a platform for creating new LLMs through compatibility analysis, tensor merging, evaluation, and evolutionary optimization. Clarify explicit exclusions.
- [EMEP Vision and Scope: Scientific and Engineering Objectives](https://doc.blueapi.ir/overview/vision-and-scope.md): State the scientific hypothesis, engineering objective, commercial objective, phase boundaries, and explicit exclusions that govern EMEP development.
- [EMEP Project Constitution: 20 Non-Negotiable Scientific Rules](https://doc.blueapi.ir/overview/project-constitution.md): The 20 non-negotiable rules governing EMEP design, execution, evaluation, and deployment. Each rule includes rationale and consequences for violation.
- [EMEP Glossary: Terms and Definitions](https://doc.blueapi.ir/overview/glossary.md): Alphabetical definitions of EMEP terminology organized by category: platform concepts, model lifecycle, compatibility, merging, evolution, evaluation, deployment, and operations.
- [EMEP Architecture Decision Records (ADRs)](https://doc.blueapi.ir/overview/decisions.md): Ratified architecture decisions governing EMEP design: modular architecture, documentation-first execution, hidden test isolation, fine-tuning optionality, RAG placement, diagram format, provenance, compatibility states, and Pareto selection.
- [EMEP Annotated Bibliography: Research References](https://doc.blueapi.ir/research/references.md): Annotated bibliography of foundational papers in model merging, task arithmetic, evolutionary optimization, quantization, evaluation, and reproducibility. Grouped by research area with relevance to EMEP.
- [EMEP Evidence Map: Claims to Sources](https://doc.blueapi.ir/research/evidence-map.md): Maps every EMEP claim to its evidence source, evidence type, and affected component. Includes a Mermaid concept map linking research areas to platform components.
- [EMEP Component Mapping: Research Dependencies](https://doc.blueapi.ir/research/component-mapping.md): Reverse mapping of each EMEP component to the research papers and methods it depends on. Includes a Mermaid dependency diagram.
- [EMEP System Dependency Map: End-to-End Pipeline](https://doc.blueapi.ir/maps/system-dependency-map.md): Full system dependency map showing the primary model pipeline from discovery through deployment, with optional side branches for fine-tuning and downstream RAG consumption.
- [EMEP Component Dependency Graph: 17 Internal Abstractions](https://doc.blueapi.ir/maps/component-dependency-graph.md): Component-level dependencies between the 17 internal EMEP abstractions. Shows which components call, depend on, or produce outputs for others.
- [EMEP Data Flow Map: Weights, Vectors, and Artifacts](https://doc.blueapi.ir/maps/data-flow-map.md): Data flow diagram showing how model weights, task vectors, candidates, evaluation results, and artifacts move through EMEP. Includes storage tiers and transformation stages.
- [EMEP Experiment Lifecycle: States and Transitions](https://doc.blueapi.ir/maps/experiment-lifecycle.md): Experiment state machine with all valid and invalid transitions. Includes Mermaid state diagram and sequence diagram of experiment execution.
- [EMEP System Architecture: Three-Level Design](https://doc.blueapi.ir/architecture/system-architecture.md): Three-level architecture for EMEP: executive blocks, logical abstractions, and internal component details. Covers model sourcing, compatibility, merging, evolution, evaluation, and deployment.
- [EMEP Technical Requirements: Functional and Non-Functional](https://doc.blueapi.ir/architecture/technical-requirements.md): Complete functional and non-functional requirements for EMEP. Covers performance, reliability, security, reproducibility, observability, offline operation, and portability.
- [EMEP Component Architecture: 17 Internal Abstractions](https://doc.blueapi.ir/architecture/component-architecture.md): Detailed specification of all 17 EMEP internal abstractions. Covers purpose, interface, responsibilities, dependencies, data ownership, persistence, and failure modes.
- [EMEP Runtime Architecture: Processes, Workers, and Message Flow](https://doc.blueapi.ir/architecture/runtime-architecture.md): Runtime topology for EMEP covering API server, scheduler, worker pools, message flow, and queue design. Includes deployment-style and sequence diagrams.
- [EMEP Security Architecture: Trust Boundaries and Key Hierarchy](https://doc.blueapi.ir/architecture/security-architecture.md): Security architecture for EMEP covering trust boundaries, principal types, secrets management, artifact signing hierarchy, and network zones including air-gap deployment.
- [EMEP Storage Architecture: Tiers, Layout, and Retention](https://doc.blueapi.ir/architecture/storage-architecture.md): Storage architecture for EMEP covering hot SSD/NVMe, warm NAS/object, and cold archive tiers. Includes layout for artifacts, datasets, experiment DB, metrics, and logs.
- [EMEP GPU Orchestration: Scheduling, Queues, and Topology](https://doc.blueapi.ir/architecture/gpu-orchestration.md): GPU scheduling architecture for EMEP covering priority queues, preemption, fair-share allocation, multi-node support, and NVLink topology awareness.
- [EMEP Observability: Logs, Metrics, Traces, and Alerting](https://doc.blueapi.ir/architecture/observability.md): Observability architecture for EMEP covering logs, metrics, traces, model-level metrics, evaluation metrics, GPU monitoring, and alerting hooks.
- [EMEP Diagram Index: Complete Visual Reference](https://doc.blueapi.ir/diagrams/diagram-index.md): Master index of every diagram in the EMEP documentation site, organized by category with diagram type, location, corresponding specification, phase, and tracking identifiers.
- [EMEP Architecture Diagrams: Three-Level Reference](https://doc.blueapi.ir/diagrams/architecture-diagrams.md): Consolidated architecture-only view of EMEP with Level 1, 2, and 3 diagrams plus deployment, infrastructure, storage, GPU, security, and supply chain views.
- [EMEP Algorithm Flowchart Index: All Processes](https://doc.blueapi.ir/diagrams/algorithm-flowchart-index.md): Complete index of every algorithm flowchart in EMEP, from model discovery through evolution, evaluation, deployment, and recovery, with links to specs and tracking IDs.
- [ModelCompatibilityAnalyzer: Full Specification](https://doc.blueapi.ir/compatibility/model-compatibility.md): Complete specification for the ModelCompatibilityAnalyzer component, including the full decision tree, all compatibility checks, and the three output states used by EMEP.
- [Model Import Algorithm and Pipeline Specification](https://doc.blueapi.ir/compatibility/model-import.md): Full specification for the Model Import pipeline, including discovery, verification, download, hash validation, metadata extraction, format detection, loading, and registration.
- [Tokenizer Compatibility Algorithm and Decision Flow](https://doc.blueapi.ir/compatibility/tokenizer-compatibility.md): Full specification for tokenizer compatibility analysis in EMEP, including vocabulary intersection, special token alignment, format detection, and decision states.
- [Architecture Compatibility Specification and Config Comparison](https://doc.blueapi.ir/compatibility/architecture-compatibility.md): Specification for architecture compatibility analysis in EMEP, including the full config comparison table and the sequence diagram for the compatibility check.
- [Tensor Shape Validation Algorithm and Per-Tensor Checks](https://doc.blueapi.ir/compatibility/tensor-shape-validation.md): Full specification for tensor shape validation in EMEP, including per-tensor iteration, name matching, shape matching, dtype matching, and the complete decision flowchart.
- [MergeEngine: Architecture and Component Specification](https://doc.blueapi.ir/merge/merge-engine.md): Complete specification for the EMEP MergeEngine, including plugin architecture, TensorEngine integration, validation, artifact writeout, and the full execution sequence diagram.
- [Merge Strategies: Overview, Comparison, and Selection](https://doc.blueapi.ir/merge/merge-strategies.md): Overview of all EMEP merge strategies including Linear, SLERP, TIES, DARE, DARE+TIES, Task Arithmetic, Franken-Merge, and Passthrough, with comparison table and decision tree.
- [Tensor Operations: Primitives, Flowcharts, and Validation](https://doc.blueapi.ir/merge/tensor-operations.md): Specification for every tensor operation in EMEP including element-wise ops, normalization, interpolation, sparsification, masking, and concatenation, each with validation flowcharts.
- [Merge Validation: Post-Merge Checks and Integrity Verification](https://doc.blueapi.ir/merge/merge-validation.md): Specification for merge validation in EMEP, including NaN/Inf detection, norm bounds, parameter count preservation, tensor integrity, and checksum verification with full flowchart.
- [Merge Mathematics: Scalars, Vectors, Matrices, and Tensors](https://doc.blueapi.ir/math/merge-math.md): Overview of the mathematical foundations of model merging in EMEP, covering scalars, vectors, matrices, tensors, weight tensors, parameter deltas, task vectors, and norms.
- [SLERP Math: Spherical Linear Interpolation for Model Merging](https://doc.blueapi.ir/math/slerp-math.md): Formal definition of SLERP (Shoemake 1985) for EMEP model merging, including formula, symbol table, shape expectations, numerical stability, and fallback to LERP.
- [TIES-Merging Math: Trim, Elect Sign, and Disjoint Merge](https://doc.blueapi.ir/math/ties-math.md): Formal specification of TIES-Merging (Yadav et al. 2023) for EMEP, including trim, sign election, disjoint merge, symbols, shapes, and a worked example.
- [DARE Math: Drop And REscale for Task Vector Merging](https://doc.blueapi.ir/math/dare-math.md): Formal specification of DARE (Yu et al. 2023) for EMEP, including random drop, rescale factor, symbols, shapes, and interaction with TIES-Merging.
- [Task Arithmetic: Task Vector Addition and Negation for Model Merging](https://doc.blueapi.ir/math/task-arithmetic.md): Formal specification of Task Arithmetic (Ilharco et al. 2022) for EMEP, including task vector definition, addition, negation, and multi-task composition with symbols and examples.
- [Tensor Math: Broadcasting, Dtype Promotion, Norms, and Precision](https://doc.blueapi.ir/math/tensor-math.md): Specification for tensor primitives in EMEP including broadcasting rules, dtype promotion, norm computations, element-wise operations, and precision effects across fp32, bf16, and fp16.
- [Optional Fine-Tuning: Secondary Pipeline Specification](https://doc.blueapi.ir/finetuning/optional-finetuning.md): Specification for the optional fine-tuning pipeline in EMEP, including full flowchart, catastrophic forgetting controls, and explicit statement that fine-tuning is secondary to merge and evolution.
- [PEFT Overview: Methods, Selection, and Integration in EMEP](https://doc.blueapi.ir/finetuning/peft.md): Overview of Parameter-Efficient Fine-Tuning methods in EMEP including LoRA, QLoRA, prefix tuning, and adapters, with guidance on when to use each method.
- [LoRA Specification: Low-Rank Adaptation for EMEP](https://doc.blueapi.ir/finetuning/lora.md): Full specification for LoRA (Hu et al. 2021) in EMEP, including low-rank decomposition, rank and alpha scaling, shapes, algorithm, and integration flowchart.
- [QLoRA Specification: Quantized Low-Rank Adaptation for EMEP](https://doc.blueapi.ir/finetuning/qlora.md): Full specification for QLoRA (Dettmers et al. 2023) in EMEP, including NF4 quantization, double quantization, paged optimizers, and precision considerations.
- [Catastrophic Forgetting Controls: Detection, Mitigation, and Rejection](https://doc.blueapi.ir/finetuning/catastrophic-forgetting-controls.md): Specification for catastrophic forgetting controls in EMEP including rehearsal, replay, low learning rate, early stopping, regularization, distillation, and mandatory regression evaluation before and after any fine-tune.
- [EvolutionEngine: Full Specification and Master Flowchart](https://doc.blueapi.ir/evolution/evolution-engine.md): Complete specification for the EMEP EvolutionEngine. Defines initialization, evaluation, selection, mutation, crossover, and termination for single-objective and multi-objective evolutionary search.
- [Genome Schema: Encoding Merge Configurations for Evolution](https://doc.blueapi.ir/evolution/genome.md): Specification of the EMEP Genome data structure. Defines the JSON schema, fields, and class diagram for encoding parent models, merge strategies, coefficients, and structural parameters.
- [Population Data Structure and Initialization Strategies](https://doc.blueapi.ir/evolution/population.md): Specification of the EMEP Population data structure, initialization strategies, diversity metrics, and aging mechanisms for evolutionary model merging.
- [Mutation Operators: Perturbing Genomes in Evolutionary Search](https://doc.blueapi.ir/evolution/mutation.md): Specification of EMEP mutation operators for evolutionary model merging. Defines perturbation, parent swap, density change, and layer routing resampling with full flowcharts and validation.
- [Crossover Operators: Recombining Genomes in Evolutionary Search](https://doc.blueapi.ir/evolution/crossover.md): Specification of EMEP crossover operators for evolutionary model merging. Defines one-point, uniform, parent-set, and per-layer crossover with full flowcharts and validation rules.
- [Selection Operators: Choosing Parents for Reproduction](https://doc.blueapi.ir/evolution/selection.md): Specification of EMEP selection operators including tournament, truncation, rank-based, and NSGA-II selection. Defines elitism, flowcharts, and integration with single-objective and multi-objective modes.
- [Fitness Calculation: Aggregating Evaluation Metrics into Scores](https://doc.blueapi.ir/evolution/fitness.md): Specification of EMEP fitness calculation. Defines metric normalization, weighted aggregation for single-objective, vector fitness for multi-objective, regression penalties, and constraint handling.
- [Multi-Objective Optimization: Pareto Fronts and NSGA-II Selection](https://doc.blueapi.ir/evolution/multi-objective-optimization.md): Specification of EMEP multi-objective optimization using Pareto dominance, non-dominated sorting, crowding distance, and constraint handling for evolutionary model merging.
- [EvaluationEngine: Full Specification and Master Flowchart](https://doc.blueapi.ir/evaluation/framework.md): Complete specification for the EMEP EvaluationEngine. Defines the candidate evaluation pipeline from integrity checks through benchmark execution, scoring, regression detection, and candidate status assignment.
- [Benchmark Specification: Splits, Schema, and Evaluation Profile](https://doc.blueapi.ir/evaluation/benchmark-specification.md): Specification of the EMEP benchmark schema, dataset split architecture, and evaluation profile structure. Defines Optimization, Validation, and Hidden Test sets with strict access controls.
- [Benchmark Catalog: Categories and Dataset Registry](https://doc.blueapi.ir/evaluation/benchmark-catalog.md): Specification of the EMEP benchmark catalog structure organized by category. Defines slots for Persian, English, reasoning, safety, and enterprise benchmarks pending Phase 4 selection.
- [Persian Evaluation Profile: Script, Register, and Cultural Considerations](https://doc.blueapi.ir/evaluation/persian-evaluation.md): Specification of the EMEP Persian language evaluation profile. Covers script and orthography, formal and conversational registers, cultural knowledge, and translation quality with placeholder benchmarks.
- [Safety Evaluation: Refusals, Harm, Jailbreak, Bias, and Hallucination](https://doc.blueapi.ir/evaluation/safety-evaluation.md): Specification of EMEP safety evaluation categories including refusals, harmful content, jailbreak resistance, bias, and hallucination detection. Defines scoring, thresholds, and regression gates.
- [Regression Detection: Baseline Comparison and Statistical Significance](https://doc.blueapi.ir/evaluation/regression-evaluation.md): Specification of EMEP regression detection. Defines per-metric baseline comparison, bootstrap confidence intervals, threshold-based flagging, and REGRESSION status assignment.
- [Benchmark Integrity: Contamination, Leakage, and Overfitting Controls](https://doc.blueapi.ir/evaluation/benchmark-integrity.md): Specification of EMEP benchmark integrity controls. Covers contamination detection, data leakage prevention, overfitting mitigation, prompt sensitivity, evaluator bias, and metric instability.
- [ExperimentTracker: Record and Lifecycle Specification](https://doc.blueapi.ir/tracking/experiment-tracking.md): Specification for the ExperimentTracker component that records every experiment run, captures full provenance metadata, and manages experiment lifecycle states from creation through archival.
- [Experiment Schema: JSON Record Structure and Constraints](https://doc.blueapi.ir/tracking/experiment-schema.md): Complete JSON schema specification for experiment records in EMEP, including field types, constraints, validation rules, and entity relationships with models, datasets, and artifacts.
- [ModelRegistry: State Machine and Lifecycle Management](https://doc.blueapi.ir/tracking/model-registry.md): Specification for the ModelRegistry component that manages model lifecycle states, lineage tracking, hash-based lookup, and search across all registered models in EMEP.
- [DatasetRegistry: Schema, Versioning, and Split Integrity](https://doc.blueapi.ir/tracking/dataset-registry.md): Specification for the DatasetRegistry that manages dataset schemas, versioning, content hashing, split declarations, and split leak detection for all evaluation datasets in EMEP.
- [ArtifactStore: Artifact Types and Lifecycle Management](https://doc.blueapi.ir/tracking/artifact-registry.md): Specification for the ArtifactStore that manages all EMEP artifacts including weights, tokenizers, configs, merge recipes, evaluation results, quantized weights, and deployment packages.
- [Model Lineage: Ancestry, Descendants, and Common Ancestors](https://doc.blueapi.ir/tracking/model-lineage.md): Specification for model lineage tracking in EMEP, including ancestor traversal, descendant discovery, common ancestor queries, and the graph structure that records every merge and evolution step.
- [Reproducibility: Requirements for Reproducible Experiments](https://doc.blueapi.ir/tracking/reproducibility.md): Specification for reproducibility requirements in EMEP, defining the complete set of inputs needed to reproduce a model, merge, benchmark, experiment, or deployment with bit-exact or tolerance-verified results.
- [Reproducibility Protocol: Step-by-Step Reproduction Flow](https://doc.blueapi.ir/tracking/reproducibility-protocol.md): Step-by-step specification for reproducing any EMEP experiment, from locating the experiment record through environment rebuild, rerun, metric comparison, tolerance checking, and final report generation.
- [Model Provenance: Source, License, and Supply Chain Tracking](https://doc.blueapi.ir/tracking/model-provenance.md): Specification for model provenance tracking in EMEP, covering source attribution, license verification, upstream lineage, cryptographic signatures, and the full supply chain from model source to production deployment.
- [QuantizationEngine: Reduce Model Precision for Efficient Deployment](https://doc.blueapi.ir/deployment/quantization.md): QuantizationEngine reduces model precision from FP32 to FP16, BF16, FP8, INT8, or INT4. Learn supported methods, selection logic, execution flow, and regression gates.
- [InferenceBackend: Serve Models Across Hardware Targets](https://doc.blueapi.ir/deployment/inference.md): InferenceBackend serves quantized models via PyTorch, vLLM, TGI, llama.cpp, or ONNX Runtime. Learn backend selection, batching, caching, and speculative decoding.
- [DeploymentManager: Release Models to Staging, Canary, and Production](https://doc.blueapi.ir/deployment/deployment-specification.md): DeploymentManager orchestrates model releases across staging, canary, and production. Learn deployment strategies, approval gates, and the deployment state machine.
- [Offline Deployment: Air-Gapped Model Release Procedures](https://doc.blueapi.ir/deployment/offline-deployment.md): Deploy EMEP models into air-gapped environments with secure transfer, integrity verification, and staged acceptance. Full flowchart and sequence diagram included.
- [Performance Validation: Benchmark and Accept Deployed Models](https://doc.blueapi.ir/deployment/performance-validation.md): PerformanceValidation measures latency, throughput, memory, tokens per second, and TTFT. Learn the benchmark protocol, statistical rigor, and regression gate.
- [Enterprise Model Versioning for EMEP](https://doc.blueapi.ir/enterprise/enterprise-model.md): Specification for Enterprise Model versioning in EMEP, covering semantic versioning rules, lineage tracking, compatibility guarantees, and deprecation policies for production model releases.
- [Air-Gapped Operations for EMEP Enterprise](https://doc.blueapi.ir/enterprise/offline-operations.md): Specification for running EMEP in air-gapped environments with no network egress, covering model transfer, dataset ingress, monitoring, and forbidden operations in isolated deployments.
- [Enterprise Model Update Protocol for EMEP](https://doc.blueapi.ir/enterprise/model-update-protocol.md): Governed workflow for promoting new model versions into EMEP production, with approval gates, security review, staging validation, and automated rollback triggers.
- [Enterprise Model Rollback Protocol for EMEP](https://doc.blueapi.ir/enterprise/model-rollback.md): Specification for rolling back Enterprise Model versions in EMEP production, covering automated triggers, manual procedures, verification steps, and post-mortem requirements.
- [Secure Model Packaging Specification for EMEP](https://doc.blueapi.ir/enterprise/secure-model-packaging.md): Specification for signed release manifests, checksums, dependency pinning, and offline bundles for EMEP Enterprise Model artifacts.
- [EMEP Optional RAG Architecture for Downstream Knowledge](https://doc.blueapi.ir/rag/optional-rag-architecture.md): Document the optional downstream RAG layer for EMEP, covering ingestion, chunking, embedding, retrieval, reranking, and citation without modifying model weights or replacing merge-based evolution.
- [EMEP Model Supply Chain Security and Verification](https://doc.blueapi.ir/security/model-supply-chain.md): Specify the end-to-end EMEP model supply chain from source discovery through deployment, with checkpoint gates, hash and signature verification, registry entry, and attack vector mitigations.
- [EMEP License Compliance and Compatibility Matrix](https://doc.blueapi.ir/security/license-compliance.md): Define license compatibility rules for EMEP model merging, including a compatibility matrix, attribution requirements, redistribution rules, and a decision tree for license classification.
- [EMEP Security Requirements: Authentication to Model Scanning](https://doc.blueapi.ir/security/security-requirements.md): Comprehensive security requirements for EMEP covering authentication, authorization, encryption, secret management, audit logging, malicious model detection, dependency scanning, container scanning, and offline operation controls.
- [EMEP Artifact Signing: Key Hierarchy and Verification](https://doc.blueapi.ir/security/artifact-signing.md): EMEP artifact signing uses a three-tier Ed25519 key hierarchy with optional Sigstore cosign integration, annual rotation, revocation, and air-gapped ceremony support.
- [EMEP Dependency Safety: Scanning and SBOM Compliance](https://doc.blueapi.ir/security/dependency-safety.md): EMEP dependency safety covers exact version pinning, SBOM generation in CycloneDX or SPDX, CVE monitoring, offline mirrors, and license blocking for supply chain integrity.
- [EMEP Offline Security: Air-Gap Boundaries and Controls](https://doc.blueapi.ir/security/offline-security.md): EMEP offline security defines air-gap boundaries, no-egress guarantees, physical media handling, incident containment, one-way log transfer, offline HSM custody, and signing ceremony steps.
- [EMEP Operations Runbook: Startup, Shutdown, and Common Tasks](https://doc.blueapi.ir/operations/runbook.md): Operational procedures for starting, stopping, and managing the EMEP platform. Covers cold start, warm start, graceful shutdown, forced shutdown, and common commands.
- [EMEP Disaster Recovery Plan and Recovery Procedures](https://doc.blueapi.ir/operations/disaster-recovery.md): Disaster recovery plan for EMEP covering failure classification, recovery procedures, RPO/RTO targets, and DR drill cadence.
- [EMEP Backup Strategy: Targets, Schedules, and Restore Procedures](https://doc.blueapi.ir/operations/backup-strategy.md): Backup targets, schedules, retention policies, and restore procedures for EMEP. Covers model artifacts, experiment data, datasets, metrics, registry metadata, and signing keys.
- [EMEP Monitoring: Metrics, Dashboards, and Data Pipeline](https://doc.blueapi.ir/operations/monitoring.md): Monitoring architecture for EMEP including metrics catalog, dashboard specifications, and the OpenTelemetry-based data pipeline from producers to dashboards.
- [EMEP Alerting: Severity Levels, Routing, and Escalation](https://doc.blueapi.ir/operations/alerting.md): Alert catalog, severity levels, routing rules, and escalation procedures for the EMEP platform. Links alerts to runbooks and on-call rotations.
- [EMEP GPU Health Checks: Metrics, Monitoring, and Auto-Remediation](https://doc.blueapi.ir/operations/gpu-health.md): GPU health monitoring for EMEP including ECC errors, thermal checks, XID errors, memory failures, PCIe errors, and the health-check flowchart with auto-remediation policies.
- [EMEP Storage Operations: Cleanup, Retention, and Cold Tier Migration](https://doc.blueapi.ir/operations/storage-operations.md): Storage operations for EMEP including artifact cleanup, retention policies, cold tier migration, capacity monitoring, and the artifact lifecycle flowchart.
- [EMEP API Specification: Model Registration to Deployment](https://doc.blueapi.ir/interfaces/api-specification.md): REST/gRPC-agnostic API specification for EMEP covering model registration, compatibility checks, merge execution, experiment lifecycle, evaluation, evolution, and artifact retrieval.
- [EMEP CLI Specification: Command Reference and Execution Flow](https://doc.blueapi.ir/interfaces/cli-specification.md): Complete specification for the emep CLI covering model operations, merge execution, experiment orchestration, evolution runs, and evaluation commands with arguments and exit codes.
- [EMEP Configuration Schema: YAML Spec and Resolution Order](https://doc.blueapi.ir/interfaces/configuration-specification.md): YAML configuration file schema for EMEP covering registry, storage, GPU, evaluation, evolution, and security sections with precedence rules and validation.
- [EMEP Testing Strategy: Pyramid, Coverage, and Quality Gates](https://doc.blueapi.ir/testing/testing-strategy.md): Comprehensive testing strategy for EMEP covering unit, integration, and end-to-end test layers with coverage targets, quality gates, and test automation workflows.
- [EMEP Unit Test Plan: Component Coverage and Test Cases](https://doc.blueapi.ir/testing/unit-test-plan.md): Detailed unit test plan for all 17 EMEP abstractions with test scope, coverage targets, and key test cases per component.
- [EMEP Integration Test Plan: Cross-Component Workflows](https://doc.blueapi.ir/testing/integration-test-plan.md): Integration test scenarios for EMEP covering model import through deployment workflows with cross-component validation and sequence diagrams.
- [EMEP Model Merge Tests: Strategy Validation and Regression](https://doc.blueapi.ir/testing/model-merge-tests.md): Merge-specific testing strategy for EMEP covering per-strategy numerical validation, tolerance thresholds, regression fixtures, and cross-hardware consistency.
- [EMEP Numerical Testing: Tolerance, Determinism, and Stability](https://doc.blueapi.ir/testing/numerical-testing.md): Numerical testing policy for EMEP defining absolute and relative tolerances per dtype, determinism requirements, seed control, and cross-hardware validation standards.
- [EMEP Evaluation Testing: Benchmark Integrity and Split Isolation](https://doc.blueapi.ir/testing/evaluation-testing.md): Evaluation testing strategy for EMEP covering benchmark harness correctness, metric computation validation, and strict split isolation with Hidden Test Set protection.
- [EMEP Performance Testing: Latency, Throughput, and Regression Gates](https://doc.blueapi.ir/testing/performance-testing.md): Performance testing strategy for EMEP covering latency benchmarks, throughput measurements, memory profiling, and regression gates with baseline tracking.
- [EMEP Security Testing: Artifacts, Models, Dependencies, and Auth](https://doc.blueapi.ir/testing/security-testing.md): Security testing strategy for EMEP covering artifact signature verification, tampered model rejection, malicious model detection, dependency scanning, and authentication validation.
- [EMEP Failure Injection: Chaos Testing and Recovery Protocols](https://doc.blueapi.ir/testing/failure-injection.md): Failure injection and chaos testing strategy for EMEP covering GPU failure, storage failure, network partition, corrupted models, invalid tokenizers, and worker crash scenarios with recovery flowcharts.
- [EMEP Failure Modes: FMEA Analysis and Containment Strategies](https://doc.blueapi.ir/risk/failure-modes.md): Failure mode and effects analysis for EMEP covering merge quality regression, capability interference, catastrophic forgetting, benchmark overfitting, and 12 additional critical failure modes with detection and recovery protocols.
- [EMEP Risk Management: Identification, Assessment, and Mitigation](https://doc.blueapi.ir/risk/risk-management.md): Risk management process for EMEP covering identification, assessment, mitigation, monitoring, and review with workflow diagrams and role assignments.
- [EMEP Risk Register: Identified Risks with Likelihood, Impact, and Mitigation](https://doc.blueapi.ir/risk/risk-register.md): Comprehensive risk register for EMEP cataloging all identified risks with likelihood, impact, ownership, mitigation strategies, and current status.
- [EMEP Roadmap: 10-Phase Plan for Evolutionary Model Engineering](https://doc.blueapi.ir/project/roadmap.md): EMEP builds high-performing LLMs from open-weight models via compatibility analysis, tensor merging, evaluation, and evolutionary optimization. This roadmap defines ten phases from research to enterprise deployment.
- [EMEP Project Management: Milestones, Reviews, and Delivery](https://doc.blueapi.ir/project/project-management.md): EMEP uses phase-gated delivery with milestone-based milestones, definition-of-done criteria, and structured review cadence to manage a complex multi-disciplinary engineering effort.
- [EMEP Task Catalog: Seed Tasks Across All Epics](https://doc.blueapi.ir/project/tasks.md): The EMEP task catalog defines required fields for every task and seeds 15 representative tasks across registry, compatibility, merge, math, evaluation, evolution, quantization, deployment, security, operations, documentation, and testing epics.
- [EMEP KPI Specification: Metrics, Formulas, and Ownership](https://doc.blueapi.ir/project/kpi-specification.md): EMEP tracks eight key performance indicators covering candidate throughput, evaluation efficiency, reproducibility, quality, deployment speed, reliability, and cost. Each KPI includes formula, source metric, target, and owner.
- [EMEP Phase 0: Research, Architecture, and Documentation](https://doc.blueapi.ir/project/phases/phase-0.md): Phase 0 establishes the foundational specification for EMEP. Implementation is blocked until the consistency audit passes. Deliverables include architecture, algorithms, APIs, and a fully documented design.
- [EMEP Phase 1: Model Registry and Compatibility Analysis](https://doc.blueapi.ir/project/phases/phase-1.md): Phase 1 implements the ModelRegistry and ModelCompatibilityAnalyzer. Models are imported, validated, and classified as COMPATIBLE, CONDITIONALLY_COMPATIBLE, or INCOMPATIBLE for merging.
- [EMEP Phase 2: Basic Merge Engine Implementation](https://doc.blueapi.ir/project/phases/phase-2.md): Phase 2 implements the MergeEngine with Linear Merge as the baseline strategy. It validates the tensor operation pipeline, produces merged candidates, and establishes numerical stability controls.
- [EMEP Phase 3: Advanced Merge Strategies Implementation](https://doc.blueapi.ir/project/phases/phase-3.md): Phase 3 extends the MergeEngine with SLERP, TIES, DARE, DARE + TIES, Task Arithmetic, and Franken-Merge. Each strategy is implemented with full algorithmic specifications and numerical stability controls.
- [EMEP Phase 4: Evaluation and Benchmarking Framework](https://doc.blueapi.ir/project/phases/phase-4.md): Phase 4 implements the EvaluationEngine and BenchmarkEngine with three-split integrity controls. The Hidden Test Set is isolated from the EvolutionEngine to preserve evaluation validity.
- [EMEP Phase 5: Evolution Engine Implementation](https://doc.blueapi.ir/project/phases/phase-5.md): Phase 5 implements the EvolutionEngine with genome encoding, population management, mutation, crossover, selection, and fitness scoring. This closes the primary loop: Models → Compatibility → Merge → Candidate → Evaluation → Fitness → Evolution → Best Model.
- [EMEP Phase 6: Quantization and Inference Optimization](https://doc.blueapi.ir/project/phases/phase-6.md): Phase 6 implements the QuantizationEngine and InferenceBackend. Promoted candidates are quantized with configurable precision, validated against quality thresholds, and prepared for deployment.
- [EMEP Phase 7: Offline Deployment Implementation](https://doc.blueapi.ir/project/phases/phase-7.md): Phase 7 implements the DeploymentManager for air-gapped environments. Models are packaged with integrity verification, transferred offline, and validated in the target environment.
- [EMEP Phase 8: Enterprise Pilot and Production Readiness](https://doc.blueapi.ir/project/phases/phase-8.md): Phase 8 validates EMEP in an enterprise environment through a controlled pilot. It integrates all prior phases, establishes operational procedures, and confirms production readiness.
- [EMEP Phase 9: Optional RAG Architecture Extension](https://doc.blueapi.ir/project/phases/phase-9.md): Phase 9 is an optional extension that adds retrieval-augmented generation capabilities to EMEP-produced models. It is not part of the primary evolutionary loop.
- [EMEP Consistency Audit Report Template and Process](https://doc.blueapi.ir/audit/consistency-report.md): The EMEP consistency audit verifies that terminology, architecture, APIs, math, benchmarks, security, and task alignment are consistent across all specification pages. Implementation is blocked until the audit passes.
- [EMEP Traceability Matrix: Requirements to Tests to KPIs](https://doc.blueapi.ir/audit/traceability-matrix.md): The EMEP traceability matrix links requirements to architecture components, specification pages, implementation tasks, tests, and KPIs. It ensures every requirement is implemented, tested, and measured.
- [EMEP Open Questions Log: Known Unknowns and Decisions Pending](https://doc.blueapi.ir/audit/open-questions.md): The EMEP open questions log tracks unresolved decisions, research uncertainties, and engineering ambiguities that must be resolved before or during implementation. Each question has an owner, category, status, and impact assessment.