Changelog

Dated notes for the Socaity SDK, APIPod, and the hosted platform. Current versions: socaity 0.1.6, apipod 1.0.4.

socaity (Python SDK)

• Replicate-backed models reachable via from socaity.sdk.replicate.<vendor> import <model>. Example: from socaity.sdk.replicate.black_forest_labs import flux_schnell.
• Official services face2face and speechcraft reachable via from socaity import ....
• Job.get_result(timeout_s=...) returns None on timeout and re-raises TaskException on failure.
• Polling defaults: 1 s interval, 3600 s total timeout. Up to 4 consecutive polling errors tolerated, raising on the fifth.
• runtime_info returns (delay_seconds, execution_seconds) for RunPod and Replicate jobs.

apipod

• CLI deploy subcommands: socaity scan, socaity build [FILE], socaity start (delegate to APIPod).
• @app.endpoint(path, queue_size=500) on the FastAPI router. get() and post() shortcuts default queue_size=100.
• JobProgress.set_status(progress, message) takes progress as a float in [0, 1.0].
• Default orchestrator local, default compute dedicated, default provider localhost. Default port 8000, host 0.0.0.0.
• Working providers: localhost, runpod. scaleway and azure raise NotImplementedError (planned).

Platform

• Marketplace catalog at socaity.ai with 1,200+ services across Text, Audio, Image, Video, Animation, and Misc.
• Active-only billing: charged for GPU-seconds in PROCESSING state. Idle, queued, cold-start, failed, and cancelled jobs are not billed.
• Subscription plans (Plus, Pro, Ultimate) and pay-per-call pricing listed at socaity.ai/Pricing.
• Serverless GPU on RunPod (EU and US regions). Dedicated GPU on Socaity-managed European infrastructure.

JavaScript SDK

• npm install socaity ships low-level helpers: setApiKey, getAvailableModels, getJobs, cancelJob.
• High-level methods (chat, text2img, text2voice) are planned.
• Runs in Node.js and the browser.

Models at launch

Items below are planned, not shipped. Each will get its own dated entry above when it lands.