t3nets-sdk 0.1.0

Public SDK for building t3nets practices — contracts, models, interfaces, test doubles.

t3nets-sdk

Public SDK for building t3nets practices.

A practice is a bundle of skills, pages, and prompts that customizes a t3nets deployment for a specific domain (engineering, therapy, ops, etc.). Practices typically live in their own repos — this SDK is the only t3nets dependency they need.

What's in here

• Models (t3nets_sdk.models) — RequestContext, Tenant, TenantUser, TenantSettings, InboundMessage, OutboundMessage, channel types.
• Interfaces (t3nets_sdk.interfaces) — BlobStore, SecretsProvider, EventBus, ConversationStore. Abstract ports a skill worker may receive at runtime.
• Contracts (t3nets_sdk.contracts) — typed SkillContext / SkillResult worker contract. Workers receive a SkillContext and return a SkillResult carrying either text (verbatim) or render_prompt (router AI formatter).
• Manifest validators (t3nets_sdk.manifest) — pydantic validators for practice.yaml / skill.yaml.
• Test doubles (t3nets_sdk.testing) — in-memory Mock* implementations of every interface, for offline skill tests.
• CLI (t3nets) — practice init, practice validate, practice package, practice run-local.

Design rules

• Zero cloud SDKs. No boto3, no anthropic, no starlette. The SDK must stay light enough to drop into any practice repo.
• Stable surface. SemVer strict. Practices pin a major.
• Stdlib dataclasses on the hot path. Pydantic only for manifest parsing — keeps Lambda cold starts lean.

Installation

pip install t3nets-sdk

For local development against the t3nets monorepo:

pip install -e ./sdk