Product-specific documentation for the reusable provider connectivity layer behind Beka Forge AI-enabled tools. It covers provider assets, request transport, local key handling, runtime/editor request paths, and the public API core.
AI Providers owns provider assets, request DTOs, request execution, local project-scoped key lookup, timeout/retry behavior, and provider-specific response extraction. It does not own graph authoring workflows.
Import flow, provider asset creation, local API key entry, and the first editor/runtime prompt send.
Built-in provider families, request DTOs, service ownership, and response extraction boundaries.
PlayerPrefs-based local secret storage, public API surfaces, blocking/runtime tradeoffs, and security caveats.
The package is dependency-free and can be reused by tools outside Dialogue Graph System.
Each provider subclass owns its request shape and response parsing instead of relying on one monolithic service switch.
Provider assets are shareable, but keys stay local and are scoped by project hash plus provider GUID.
TrySendPrompt is blocking and best described as editor-oriented. Runtime callers should generally use the coroutine path.
Import the asset, create a provider asset, and send the first request.
Service-driven transport, provider abstraction, and DTO ownership.
Built-in provider families and what their asset layer actually stores.
Blocking and coroutine request paths, timeout options, and response normalization.
Shared asset config, local keys, and editor-only preference state.
The reusable service, provider base class, DTOs, and built-in provider types.
