Production-Ready
ML Solutions
A comprehensive Elixir ecosystem for machine learning, AI safety, and reliability engineering. From training to deployment, from labeling to monitoring.
Package Categories
Core Framework
ML experimentation orchestration, pipeline IR, and statistical testing infrastructure.
3 packagesMLOps Pipeline
Complete ML lifecycle: training, model registry, deployment, and feedback loops.
4 packagesTraining Infrastructure
Kitchen-based training orchestration with recipes and backend-agnostic execution.
3 packagesData Labeling
Metalworking-inspired HITL labeling stack: forge, anvil, and ingot components.
4 packagesReliability & Safety
Ensemble voting, hedging, AI firewalls, adversarial testing, and explainability.
7 packagesML Utilities
Tokenizers, embeddings, datasets, metrics, and configuration management.
10 packagesRAG
Retrieval-augmented generation with hexagonal architecture and multi-provider adapters.
4 packagesOrchestration
Asset-first data pipelines and headless multi-agent coordination frameworks.
2 packagesInfrastructure
API gateway, service discovery, database drivers, and resilience patterns.
5 packagesFlagship Packages
crucible_framework
ML experimentation orchestration with composable pipeline stages, adapters, and telemetry integration.
{:crucible_framework, "~> 0.1.0"}
LlmGuard
AI firewall for prompt injection detection, jailbreak prevention, and data leakage protection.
{:llm_guard, "~> 0.1.0"}
crucible_train
Unified ML training with supervised learning, RL, DPO, distillation, and multi-backend logging.
{:crucible_train, "~> 0.1.0"}
crucible_xai
Explainable AI toolkit with LIME, SHAP, PDP/ICE plots, and faithfulness validation.
{:crucible_xai, "~> 0.1.0"}