Architecture Overview
What AgentiCraft's two published libraries are for, and where LLM provider management lives.
Package Architecture
AgentiCraft publishes two focused libraries, each with a clear responsibility and neither depending on the other:
┌─────────────────────────────────────────────────────┐
│ Your Application │
└─────────────┬─────────────────────────┬─────────────┘
│ │
v v
┌─────────────────────────┐ ┌─────────────────────────┐
│ agenticraft-types │ │ agenticraft- │
│ │ │ foundation │
│ Pydantic models │ │ │
│ Protocols │ │ CSP algebra │
│ Error hierarchy │ │ Topology │
│ Config schemas │ │ Verification │
│ │ │ Proofs │
└────────────┬────────────┘ └────────────┬────────────┘
│ │
┌────────┴────────┐ ┌────────┴────────┐
│ pydantic >=2.0 │ │ numpy >=1.24 │
└─────────────────┘ └─────────────────┘
agenticraft-types
The contract layer. Defines shared interfaces:
- Pydantic models —
CompletionRequest,CompletionResponse,Message,Usage,StreamChunk,ProviderConfig - Protocol classes —
Provider,StreamingProvider,Router,CircuitBreakerLike,KeyPoolLike - Error hierarchy — 12 exception types with structured error details
- Enums —
ProviderName,FailureType,RoutingStrategy,MessageRole,CircuitBreakerState - Configuration —
RetryConfig,CircuitBreakerConfig,RateLimitConfig
Design constraints: under 1,000 LOC, zero business logic, only dependency is pydantic>=2.0.
agenticraft-foundation
The mathematical layer. Formal verification and analysis for multi-agent systems:
- CSP algebra — Process calculus for specifying and verifying agent communication
- Topology analysis — Network graph metrics, Laplacian eigenvalues, connectivity analysis
- Stochastic primitives — Quality functions, ensemble judges, cost routing formalization
- Proofs — Machine-checkable correctness proofs for consensus protocols
LLM Provider Management
agenticraft-llm is archived. It was a third package, the operational layer between an application and its LLM providers: provider adapters, cost-aware routing, key rotation, circuit breaking and rate limiting. Those capabilities were absorbed into the mesh LLM stack that runs inside the AgentiCraft platform. The stack is not distributed as a library, so no package sits beside the two above for provider management.
Next Steps
- Getting Started — install the libraries
- Multi-Agent Patterns — coordination patterns
- Production Deployment — checklists for running LLM-backed agents