Public API

Symbols re-exported from aire (src/aire/__init__.py) and the AI facade (src/aire/ai.py).

Top-level exports

Symbol	Role
AI	Unified facade
Agent, AgentConfig, AgentResult, AgentStatus	Agent types
Team, TeamResult	Multi-agent team
tool, Tool, ToolSpec, ToolResult, SideEffect	Tools
Knowledge, Answer, Citation	RAG
Assistant	Fluent project builder (AI.project)
Model, EmbeddingModel, ModelInfo, GenerationRequest, GenerationResult	Models
Dataset, Record	Data
EvalCase, EvalReport, Evaluator	Evaluation
KnowledgeGraph	GraphRAG
LongTermMemory, MemoryEntry	Memory
Workflow, WorkflowResult	Workflows
Runtime, Settings	Core runtime/config
Message, content types (TextContent, …)	Content
Capability, HealthStatus, Manifest, Ref, Usage	Shared types
AireError, ConfigurationError, PermissionDeniedError, ProviderError	Errors
__version__	Package version (0.4.0)

Import what you need from aire first; deeper modules are stable for advanced use but not all are re-exported.

AI facade namespaces

Namespace	Highlights
AI.models	use / use_sync, embedder, router, cache, register_callable, describe
AI.data	load, chunker, describe
AI.rag	create → Knowledge, vector_store, incremental, describe
AI.graph	create, store, communities
AI.memory	long-term memory helpers
AI.mcp	MCP server/client helpers
AI.agents	create / create_sync, team, builder, pattern, toolkit, approver, topologies
AI.observe	tracer, metrics, analytics, costs
AI.deploy	api, artifacts, scale
AI.gateway	create, serve, endpoints
AI.workflows	create, HITL helpers
AI.training	LoRA / quantize / distill / HPO / foundation (toy)
AI.ml	estimators / arch
AI.safety	guardrails, redact, policy
AI.skills	skill registry
AI.schedule	scheduler
AI.workers	in-process / file queue workers
AI.recipes	one-call scaffolds
AI.locks	project lock files
AI.vision / AI.audio / AI.docs	multimodal helpers (partially stubbed)

Class methods on AI

• AI.runtime() / AI.configure(...) — process-wide Runtime
• AI.project(name) / AI.from_config(path) → Assistant
• AI.workflow(name) / AI.recipe(name) / AI.topologies()
• AI.tool(...) — alias for registering tools
• AI.evaluate(target, dataset, ...) — evaluation runner
• AI.describe() — top-level manifest

Typical usage

from aire import AI, tool

model = AI.models.use_sync("mock:echo")

@tool
def add(a: int, b: int) -> int:
    """Add two integers."""
    return a + b

agent = AI.agents.create_sync(model, tools=[add])
print(agent.run_sync("hello").output)