socaity-schemas

Shared Pydantic models for APIPod services and the socaity product APIs. One definition, used on both sides of the wire, so a service and a client stay on the same field names.

Import the module that defines the type. The package root does not re-export symbols.

public is what anyone calling a service needs:

• public.inference.language: chat, completion, embedding, tools, usage, and stream chunks.
• public.inference.generation: image, video, audio, vision, and 3D.
• public.inference.media: FileModel and its typed subclasses.
• public.providers: Socaity, RunPod, and Replicate job envelopes, plus the StreamingResponse sentinel.
• public.spec.address: ServiceAddress variants and URL helpers.
• public.spec.endpoint: parameters, Endpoint, and ServiceContract.

platform is the socaity product:

• platform.catalog: Service (category, endpoint meta, ServiceDetails), hosting (Deployment, ContainerConfig, ServiceConnector), AIModel, and public pricing.
• platform.jobs: Job and Project.
• platform.conversations: chat transcript and the interrupt inbox.
• platform.workflows: the workflow document and the catalog (revision, run, trace, upsert, estimate).
• platform.account: User, ApiKeyView, and billing (Balance, Purchase, Subscription, SpendBreakdown).
• platform.credentials, platform.files, platform.context, platform.agent.

from socaity_schemas.public.inference.language import ChatCompletionRequest
from socaity_schemas.public.spec.endpoint import ServiceContract
from socaity_schemas.platform.catalog.service import Service
from socaity_schemas.platform.catalog.model import AIModel

The inference shapes follow the OpenAI API so OpenAI-compatible clients work without translation. That is the wire format only. Any provider that matches the shape is a valid backend.

Install

pip install socaity-schemas

Apache-2.0.