Skip to content

Clarify cua-ai runtime spec docs #13

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
rgarcia merged 1 commit into from
May 14, 2026
Merged

Clarify cua-ai runtime spec docs #13

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 7 additions & 9 deletions packages/ai/src/runtime-spec.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,16 +7,15 @@ import * as tzafon from "./providers/tzafon/index";
import * as yutori from "./providers/yutori/index";

/**
* Provider-resolved runtime defaults for CUA execution.
* Runtime configuration for a supported CUA model.
*
* `@onkernel/cua-agent` consumes this shape so it can stay provider-neutral:
* it only needs model, tool definitions, and optional payload middleware,
* without branching on provider-specific quirks.
* Use this to pair a model with the tool definitions, baseline prompt, and
* request payload middleware expected by its provider.
*/
export interface CuaRuntimeSpec {
model: Model<Api>;
provider: CuaProvider;
/** Canonical CUA tool definitions exposed to the model. */
/** Model-facing CUA tool definitions for this provider. */
toolDefinitions: Tool[];
/** Provider-tuned baseline prompt for browser control behavior. */
defaultSystemPrompt: string;
Expand All @@ -27,11 +26,10 @@ export interface CuaRuntimeSpec {
export type CuaRuntimeSpecInput = CuaModelRef | Model<Api>;

/**
* Resolve provider-owned policy from either a CUA model ref or a concrete model.
* Resolve provider defaults from either a CUA model ref or a concrete model.
*
* This is intentionally the only place where provider-specific defaults are
* selected. Callers should consume the returned spec and avoid branching on
* provider names directly.
* Use the returned spec to build computer-use requests without hard-coding
* model-provider rules in your application.
*/
export function resolveCuaRuntimeSpec(input: CuaRuntimeSpecInput): CuaRuntimeSpec {
const model = typeof input === "string" ? getCuaModel(input) : input;
Expand Down
Loading