@aexhq/sdk 0.45.2-canary → 1.0.1-canary

package/dist/_contracts/index.d.ts

@@ -1,10 +1,9 @@
-export * from "./provider-support.js";
 export * from "./provider-fault.js";
 export * from "./models.js";
 export { SESSION_LIFECYCLE_STATUSES, SESSION_STATUSES, SESSION_TERMINAL_OUTCOMES, isTerminalSessionStatus } from "./status.js";
 export type { SessionLifecycleStatus, SessionStatus, SessionTerminalOutcome } from "./status.js";
-export { AEX_RESERVED_ENV_PREFIX, BUILTIN_TOOL_NAMES, BuiltinTools, DEFAULT_BUILTIN_TOOLS, DEFAULT_OUTPUT_MODE, DEFAULT_PROVIDER, ENV_VARS_MAX_ENTRIES, ENV_VARS_MAX_TOTAL_BYTES, ENV_VARS_MAX_VALUE_BYTES, OUTPUT_MODES, PLATFORM_PACKAGE_ECOSYSTEMS, Providers, RESPONSE_FORMAT_KINDS, PROVIDERS, SECRETS_KEY, SECRET_ENV_NAME_PATTERN, SECRET_HANDLE_PATTERN, SKILLS_TOOL_DEFINITION, SKILLS_TOOL_NAME, STREAMABLE_SHAPES, assertStreamableOutputMode, crossValidateSecretEnvAndValues, isStreamableProvider, packageInstallString, parseApprovalGate, parseInlineSecrets, parseResponseFormat, parseSessionLimits, parseProviderName, parseSessionWebhook, parseSubmission, resolveBuiltinToolNames } from "./submission.js";
-export type { ApprovalGate, BuiltinToolName, JsonPrimitive, JsonValue, OutputMode, PlatformEnvironment, PlatformEnvironmentInput, PlatformInlineSecrets, PlatformInjectionConfig, PlatformMcpServerSecret, PlatformNetworking, PlatformFileCaptureConfig, PlatformPackage, PlatformPackageEcosystem, PlatformPackageInput, PlatformSecretEnvEntry, PlatformSubmission, ResponseFormat, ResponseFormatKind, SessionLimits, ProviderName, SessionWebhookSpec, StreamableShape } from "./submission.js";
+export { AEX_RESERVED_ENV_PREFIX, BUILTIN_TOOL_NAMES, BuiltinTools, DEFAULT_BUILTIN_TOOLS, DEFAULT_OUTPUT_MODE, ENV_VARS_MAX_ENTRIES, ENV_VARS_MAX_TOTAL_BYTES, ENV_VARS_MAX_VALUE_BYTES, OUTPUT_MODES, PLATFORM_PACKAGE_ECOSYSTEMS, RESPONSE_FORMAT_KINDS, SECRETS_KEY, SECRET_ENV_NAME_PATTERN, SECRET_HANDLE_PATTERN, SKILLS_TOOL_DEFINITION, SKILLS_TOOL_NAME, crossValidateSecretEnvAndValues, packageInstallString, parseApprovalGate, parseInlineSecrets, parseResponseFormat, parseSessionLimits, parseSessionWebhook, parseSubmission, resolveBuiltinToolNames } from "./submission.js";
+export type { ApprovalGate, BuiltinToolName, JsonPrimitive, JsonValue, OutputMode, PlatformEnvironment, PlatformEnvironmentInput, PlatformInlineSecrets, PlatformInjectionConfig, PlatformMcpServerSecret, PlatformNetworking, PlatformFileCaptureConfig, PlatformPackage, PlatformPackageEcosystem, PlatformPackageInput, PlatformSecretEnvEntry, PlatformSubmission, ResponseFormat, ResponseFormatKind, SessionLimits, ProviderName, SessionWebhookSpec } from "./submission.js";
 export * from "./runtime-sizes.js";
 export * from "./runtime-kind.js";
 export * from "./runner-event.js";

package/dist/_contracts/index.js

@@ -1,8 +1,7 @@
-export * from "./provider-support.js";
 export * from "./provider-fault.js";
 export * from "./models.js";
 export { SESSION_LIFECYCLE_STATUSES, SESSION_STATUSES, SESSION_TERMINAL_OUTCOMES, isTerminalSessionStatus } from "./status.js";
-export { AEX_RESERVED_ENV_PREFIX, BUILTIN_TOOL_NAMES, BuiltinTools, DEFAULT_BUILTIN_TOOLS, DEFAULT_OUTPUT_MODE, DEFAULT_PROVIDER, ENV_VARS_MAX_ENTRIES, ENV_VARS_MAX_TOTAL_BYTES, ENV_VARS_MAX_VALUE_BYTES, OUTPUT_MODES, PLATFORM_PACKAGE_ECOSYSTEMS, Providers, RESPONSE_FORMAT_KINDS, PROVIDERS, SECRETS_KEY, SECRET_ENV_NAME_PATTERN, SECRET_HANDLE_PATTERN, SKILLS_TOOL_DEFINITION, SKILLS_TOOL_NAME, STREAMABLE_SHAPES, assertStreamableOutputMode, crossValidateSecretEnvAndValues, isStreamableProvider, packageInstallString, parseApprovalGate, parseInlineSecrets, parseResponseFormat, parseSessionLimits, parseProviderName, parseSessionWebhook, parseSubmission, resolveBuiltinToolNames } from "./submission.js";
+export { AEX_RESERVED_ENV_PREFIX, BUILTIN_TOOL_NAMES, BuiltinTools, DEFAULT_BUILTIN_TOOLS, DEFAULT_OUTPUT_MODE, ENV_VARS_MAX_ENTRIES, ENV_VARS_MAX_TOTAL_BYTES, ENV_VARS_MAX_VALUE_BYTES, OUTPUT_MODES, PLATFORM_PACKAGE_ECOSYSTEMS, RESPONSE_FORMAT_KINDS, SECRETS_KEY, SECRET_ENV_NAME_PATTERN, SECRET_HANDLE_PATTERN, SKILLS_TOOL_DEFINITION, SKILLS_TOOL_NAME, crossValidateSecretEnvAndValues, packageInstallString, parseApprovalGate, parseInlineSecrets, parseResponseFormat, parseSessionLimits, parseSessionWebhook, parseSubmission, resolveBuiltinToolNames } from "./submission.js";
 export * from "./runtime-sizes.js";
 export * from "./runtime-kind.js";
 export * from "./runner-event.js";

package/dist/_contracts/models.d.ts

@@ -1,160 +1,29 @@
-import type { ProviderName } from "./submission.js";
 /**
- * Source of truth for the closed model set: each canonical model id maps to the
- * upstream providers that can serve it and the **provider-native** model string
- * each one expects.
+ * Public model ids are plain **Vercel AI Gateway `creator/model` slug strings**.
  *
- * `Models.*` / `SUPPORTED_MODELS` are aex's own **canonical, provider-neutral**
- * identifiers — they are NOT the strings sent to a provider. The platform
- * translates a `(canonical model, provider)` pair to the native id via
- * {@link resolveProviderModelId} when it builds the session's session manifest. The
- * same canonical model can therefore be served by more than one provider (e.g.
- * `gpt-4o-mini` via `openai` *or* `openrouter`), with a different native string
- * per provider.
- *
- * Ordering matters: the **first** provider listed for a model is its default
- * (see {@link providerForModel}) — list the native vendor before `openrouter`.
- * Additions belong here first so SDK types, CLI validation, docs examples, and
- * platform parsing all move together.
- */
-export declare const MODEL_PROVIDER_IDS: {
-    readonly "claude-haiku-4-5": {
-        readonly anthropic: "claude-haiku-4-5";
-    };
-    readonly "claude-3-5-haiku-latest": {
-        readonly anthropic: "claude-3-5-haiku-latest";
-    };
-    readonly "claude-3-5-sonnet-latest": {
-        readonly anthropic: "claude-3-5-sonnet-latest";
-    };
-    readonly "claude-sonnet-4-6": {
-        readonly anthropic: "claude-sonnet-4-6";
-    };
-    readonly "deepseek-v4-flash": {
-        readonly deepseek: "deepseek-v4-flash";
-    };
-    readonly "deepseek-v4-pro": {
-        readonly deepseek: "deepseek-v4-pro";
-    };
-    readonly "gpt-4.1": {
-        readonly openai: "gpt-4.1";
-    };
-    readonly "gpt-4o-mini": {
-        readonly openai: "gpt-4o-mini";
-        readonly openrouter: "openai/gpt-4o-mini";
-    };
-    readonly "gpt-4o": {
-        readonly openrouter: "openai/gpt-4o";
-    };
-    readonly "gemini-2.0-flash": {
-        readonly openrouter: "google/gemini-2.0-flash-001";
-    };
-    readonly "gemini-2.5-flash": {
-        readonly gemini: "gemini-2.5-flash";
-    };
-    readonly "mistral-large-latest": {
-        readonly mistral: "mistral-large-latest";
-    };
-    readonly "mistral-small-latest": {
-        readonly mistral: "mistral-small-latest";
-    };
-    readonly "doubao-seed-pro": {
-        readonly doubao: "doubao-seed-1-8-251228";
-    };
-    readonly "doubao-seed-flash": {
-        readonly doubao: "doubao-seed-1-6-flash-250828";
-    };
-};
-/**
- * Closed set of canonical model ids accepted by the public session-submission
- * schema. Derived from {@link MODEL_PROVIDER_IDS} so the two never drift.
- */
-export type ModelName = keyof typeof MODEL_PROVIDER_IDS;
-export declare const SUPPORTED_MODELS: readonly ModelName[];
-/**
- * Symbol-style accessors for the closed model set. Prefer these over raw
- * strings so an invalid token is a compile error, not a runtime 400 — e.g.
- * `Models.CLAUDE_HAIKU_4_5`. These are aex's **canonical** ids, not the native
- * strings sent upstream; the platform translates them per provider (see
- * {@link MODEL_PROVIDER_IDS} / {@link resolveProviderModelId}). When a model is
- * served by more than one provider, pair it with an explicit {@link Providers}
- * value; otherwise the single (default) provider is used.
- */
-export declare const Models: {
-    /** Claude Haiku 4.5 — Anthropic. */
-    readonly CLAUDE_HAIKU_4_5: "claude-haiku-4-5";
-    /** Claude 3.5 Haiku (latest) — Anthropic. */
-    readonly CLAUDE_3_5_HAIKU_LATEST: "claude-3-5-haiku-latest";
-    /** Claude 3.5 Sonnet (latest) — Anthropic. */
-    readonly CLAUDE_3_5_SONNET_LATEST: "claude-3-5-sonnet-latest";
-    /** Claude Sonnet 4.6 — Anthropic (1M context, reasoning-capable). */
-    readonly CLAUDE_SONNET_4_6: "claude-sonnet-4-6";
-    /** DeepSeek V4 Flash — DeepSeek (non-thinking, fast). */
-    readonly DEEPSEEK_V4_FLASH: "deepseek-v4-flash";
-    /** DeepSeek V4 Pro — DeepSeek (reasoning-heavy). */
-    readonly DEEPSEEK_V4_PRO: "deepseek-v4-pro";
-    /** GPT-4.1 — OpenAI. */
-    readonly GPT_4_1: "gpt-4.1";
-    /** GPT-4o mini — OpenAI, or via OpenRouter (`provider: Providers.OPENROUTER`). */
-    readonly GPT_4O_MINI: "gpt-4o-mini";
-    /** GPT-4o — via OpenRouter (`provider: Providers.OPENROUTER`). */
-    readonly GPT_4O: "gpt-4o";
-    /** Gemini 2.0 Flash — via OpenRouter (`provider: Providers.OPENROUTER`). */
-    readonly GEMINI_2_0_FLASH: "gemini-2.0-flash";
-    /** Gemini 2.5 Flash — Gemini. */
-    readonly GEMINI_2_5_FLASH: "gemini-2.5-flash";
-    /** Mistral Large (latest) — Mistral. */
-    readonly MISTRAL_LARGE_LATEST: "mistral-large-latest";
-    /** Mistral Small (latest) — Mistral. */
-    readonly MISTRAL_SMALL_LATEST: "mistral-small-latest";
-    /**
-     * Doubao Seed 1.8 — ByteDance, via the international BytePlus gateway.
-     */
-    readonly DOUBAO_SEED_PRO: "doubao-seed-pro";
-    /**
-     * Doubao Seed 1.6 Flash — ByteDance, via BytePlus (fast/cheap).
-     */
-    readonly DOUBAO_SEED_FLASH: "doubao-seed-flash";
-};
-/**
- * Provider → canonical models that provider can serve. Derived from
- * {@link MODEL_PROVIDER_IDS}; every provider currently serves at least one
- * model, so all {@link ProviderName} keys are present.
- */
-export declare const SUPPORTED_MODELS_BY_PROVIDER: Readonly<Record<ProviderName, readonly ModelName[]>>;
-/**
- * All upstream providers that can serve a model id, in declaration order.
- * Returns `[]` for an unknown model.
- */
-export declare function providersForModel(model: string): readonly ProviderName[];
-/**
- * The default upstream provider for a model id — the first provider declared
- * for it in {@link MODEL_PROVIDER_IDS}. Returns `undefined` when the input is
- * not a known {@link ModelName} (so the SDK can fall back to the default and let
- * the server reject the model).
- */
-export declare function providerForModel(model: string): ProviderName | undefined;
-/**
- * Translate a canonical model id + provider into the provider-native model
- * string the upstream API expects (e.g. `("gpt-4o-mini", "openrouter")` →
- * `"openai/gpt-4o-mini"`). Throws when the provider does not serve the model.
- */
-export declare function resolveProviderModelId(model: string, provider: ProviderName): string;
-/**
- * The single model→provider resolver shared by the SDK and the CLI, wrapping
- * {@link providersForModel} with the forward-compat unknown-model allowance:
- *
- *   - `provider` given: it is honored. If `model` is a KNOWN id, the provider
- *     must serve it (else throw). If `model` is UNKNOWN, it is allowed through
- *     so a slightly-old client can still run a newly-launched model (the server
- *     arbitrates).
- *   - `provider` omitted: a known model resolves to its DEFAULT provider (first
- *     declared). An UNKNOWN model with no provider throws a `did you mean?`
- *     hint — you must name a provider to run a model this client doesn't know.
- *
- * Returns the resolved {@link ProviderName}.
- */
-export declare function resolveModelProvider(model: string, provider?: ProviderName): ProviderName;
-export declare function isModelName(input: unknown): input is ModelName;
-export declare function parseModelName(input: unknown, field?: string): ModelName;
-export declare function assertModelNameMatchesProvider(provider: ProviderName, model: ModelName, field?: string): void;
+ * There is no closed model set and no provider concept: a customer names any
+ * slug the managed gateway catalog serves (e.g. `anthropic/claude-haiku-4-5`,
+ * `deepseek/deepseek-v4-flash`) and the platform routes it through the single
+ * managed gateway key. Adding a catalog model is a ZERO-code change — the slug
+ * just works. The boundary validation is purely structural (see
+ * {@link parseModelSlug}); an unknown-but-well-formed slug is allowed through so
+ * a slightly-old client can still run a newly-launched model (the gateway
+ * arbitrates at submit time).
+ */
+export type ModelName = string;
+/**
+ * Structural gateway model-slug shape: a lowercase `creator`, a `/`, then the
+ * model segment. Matches Vercel AI Gateway `creator/model` slugs
+ * (e.g. `anthropic/claude-sonnet-4-6`, `openai/gpt-4.1`, `x-ai/grok-2`).
+ */
+export declare const MODEL_SLUG_PATTERN: RegExp;
+/** True when `input` is a structurally valid `creator/model` gateway slug. */
+export declare function isModelSlug(input: unknown): input is string;
+/**
+ * Validate a public model id as a gateway `creator/model` slug string. This is
+ * a boundary shape gate ONLY — it does not consult a catalog, so a well-formed
+ * slug the running client does not recognize is accepted (the gateway rejects a
+ * truly-absent model at submit time). Throws a {@link import("./contract-parse-error.js").ContractParseError}
+ * with `field` context on a malformed value.
+ */
+export declare function parseModelSlug(input: unknown, field?: string): string;

package/dist/_contracts/models.js

@@ -1,196 +1,32 @@
-import { suggest } from "./suggest.js";
 import { rethrowContractParseError } from "./contract-parse-error.js";
 /**
- * Source of truth for the closed model set: each canonical model id maps to the
- * upstream providers that can serve it and the **provider-native** model string
- * each one expects.
- *
- * `Models.*` / `SUPPORTED_MODELS` are aex's own **canonical, provider-neutral**
- * identifiers — they are NOT the strings sent to a provider. The platform
- * translates a `(canonical model, provider)` pair to the native id via
- * {@link resolveProviderModelId} when it builds the session's session manifest. The
- * same canonical model can therefore be served by more than one provider (e.g.
- * `gpt-4o-mini` via `openai` *or* `openrouter`), with a different native string
- * per provider.
- *
- * Ordering matters: the **first** provider listed for a model is its default
- * (see {@link providerForModel}) — list the native vendor before `openrouter`.
- * Additions belong here first so SDK types, CLI validation, docs examples, and
- * platform parsing all move together.
+ * Structural gateway model-slug shape: a lowercase `creator`, a `/`, then the
+ * model segment. Matches Vercel AI Gateway `creator/model` slugs
+ * (e.g. `anthropic/claude-sonnet-4-6`, `openai/gpt-4.1`, `x-ai/grok-2`).
  */
-export const MODEL_PROVIDER_IDS = {
-    "claude-haiku-4-5": { anthropic: "claude-haiku-4-5" },
-    "claude-3-5-haiku-latest": { anthropic: "claude-3-5-haiku-latest" },
-    "claude-3-5-sonnet-latest": { anthropic: "claude-3-5-sonnet-latest" },
-    "claude-sonnet-4-6": { anthropic: "claude-sonnet-4-6" },
-    "deepseek-v4-flash": { deepseek: "deepseek-v4-flash" },
-    "deepseek-v4-pro": { deepseek: "deepseek-v4-pro" },
-    "gpt-4.1": { openai: "gpt-4.1" },
-    "gpt-4o-mini": { openai: "gpt-4o-mini", openrouter: "openai/gpt-4o-mini" },
-    "gpt-4o": { openrouter: "openai/gpt-4o" },
-    "gemini-2.0-flash": { openrouter: "google/gemini-2.0-flash-001" },
-    "gemini-2.5-flash": { gemini: "gemini-2.5-flash" },
-    "mistral-large-latest": { mistral: "mistral-large-latest" },
-    "mistral-small-latest": { mistral: "mistral-small-latest" },
-    // Doubao (ByteDance) via the official international BytePlus ModelArk API.
-    // Ark accepts the API-format model NAME directly in the chat-completions
-    // `model` field (no `ep-…` inference-endpoint id).
-    //   pro   — Doubao Seed 1.8 (flagship, 256K context).
-    //   flash — Doubao Seed 1.6 Flash (fast/cheap, 256K context).
-    "doubao-seed-pro": { doubao: "doubao-seed-1-8-251228" },
-    "doubao-seed-flash": { doubao: "doubao-seed-1-6-flash-250828" }
-};
-export const SUPPORTED_MODELS = Object.keys(MODEL_PROVIDER_IDS);
-/**
- * Symbol-style accessors for the closed model set. Prefer these over raw
- * strings so an invalid token is a compile error, not a runtime 400 — e.g.
- * `Models.CLAUDE_HAIKU_4_5`. These are aex's **canonical** ids, not the native
- * strings sent upstream; the platform translates them per provider (see
- * {@link MODEL_PROVIDER_IDS} / {@link resolveProviderModelId}). When a model is
- * served by more than one provider, pair it with an explicit {@link Providers}
- * value; otherwise the single (default) provider is used.
- */
-export const Models = {
-    /** Claude Haiku 4.5 — Anthropic. */
-    CLAUDE_HAIKU_4_5: "claude-haiku-4-5",
-    /** Claude 3.5 Haiku (latest) — Anthropic. */
-    CLAUDE_3_5_HAIKU_LATEST: "claude-3-5-haiku-latest",
-    /** Claude 3.5 Sonnet (latest) — Anthropic. */
-    CLAUDE_3_5_SONNET_LATEST: "claude-3-5-sonnet-latest",
-    /** Claude Sonnet 4.6 — Anthropic (1M context, reasoning-capable). */
-    CLAUDE_SONNET_4_6: "claude-sonnet-4-6",
-    /** DeepSeek V4 Flash — DeepSeek (non-thinking, fast). */
-    DEEPSEEK_V4_FLASH: "deepseek-v4-flash",
-    /** DeepSeek V4 Pro — DeepSeek (reasoning-heavy). */
-    DEEPSEEK_V4_PRO: "deepseek-v4-pro",
-    /** GPT-4.1 — OpenAI. */
-    GPT_4_1: "gpt-4.1",
-    /** GPT-4o mini — OpenAI, or via OpenRouter (`provider: Providers.OPENROUTER`). */
-    GPT_4O_MINI: "gpt-4o-mini",
-    /** GPT-4o — via OpenRouter (`provider: Providers.OPENROUTER`). */
-    GPT_4O: "gpt-4o",
-    /** Gemini 2.0 Flash — via OpenRouter (`provider: Providers.OPENROUTER`). */
-    GEMINI_2_0_FLASH: "gemini-2.0-flash",
-    /** Gemini 2.5 Flash — Gemini. */
-    GEMINI_2_5_FLASH: "gemini-2.5-flash",
-    /** Mistral Large (latest) — Mistral. */
-    MISTRAL_LARGE_LATEST: "mistral-large-latest",
-    /** Mistral Small (latest) — Mistral. */
-    MISTRAL_SMALL_LATEST: "mistral-small-latest",
-    /**
-     * Doubao Seed 1.8 — ByteDance, via the international BytePlus gateway.
-     */
-    DOUBAO_SEED_PRO: "doubao-seed-pro",
-    /**
-     * Doubao Seed 1.6 Flash — ByteDance, via BytePlus (fast/cheap).
-     */
-    DOUBAO_SEED_FLASH: "doubao-seed-flash"
-};
-/**
- * Per-model provider lists, in declaration order. Derived from
- * {@link MODEL_PROVIDER_IDS} so the two never drift.
- */
-const PROVIDERS_BY_MODEL = (() => {
-    const map = {};
-    for (const [model, providers] of Object.entries(MODEL_PROVIDER_IDS)) {
-        map[model] = Object.keys(providers);
-    }
-    return map;
-})();
-/**
- * Provider → canonical models that provider can serve. Derived from
- * {@link MODEL_PROVIDER_IDS}; every provider currently serves at least one
- * model, so all {@link ProviderName} keys are present.
- */
-export const SUPPORTED_MODELS_BY_PROVIDER = (() => {
-    const map = {};
-    for (const [model, providers] of Object.entries(MODEL_PROVIDER_IDS)) {
-        for (const provider of Object.keys(providers)) {
-            (map[provider] ??= []).push(model);
-        }
-    }
-    return map;
-})();
-/**
- * All upstream providers that can serve a model id, in declaration order.
- * Returns `[]` for an unknown model.
- */
-export function providersForModel(model) {
-    return PROVIDERS_BY_MODEL[model] ?? [];
-}
-/**
- * The default upstream provider for a model id — the first provider declared
- * for it in {@link MODEL_PROVIDER_IDS}. Returns `undefined` when the input is
- * not a known {@link ModelName} (so the SDK can fall back to the default and let
- * the server reject the model).
- */
-export function providerForModel(model) {
-    return providersForModel(model)[0];
+export const MODEL_SLUG_PATTERN = /^[a-z0-9-]+\/[A-Za-z0-9._:-]+$/;
+/** True when `input` is a structurally valid `creator/model` gateway slug. */
+export function isModelSlug(input) {
+    return typeof input === "string" && MODEL_SLUG_PATTERN.test(input);
 }
 /**
- * Translate a canonical model id + provider into the provider-native model
- * string the upstream API expects (e.g. `("gpt-4o-mini", "openrouter")` →
- * `"openai/gpt-4o-mini"`). Throws when the provider does not serve the model.
+ * Validate a public model id as a gateway `creator/model` slug string. This is
+ * a boundary shape gate ONLY — it does not consult a catalog, so a well-formed
+ * slug the running client does not recognize is accepted (the gateway rejects a
+ * truly-absent model at submit time). Throws a {@link import("./contract-parse-error.js").ContractParseError}
+ * with `field` context on a malformed value.
  */
-export function resolveProviderModelId(model, provider) {
-    const entry = MODEL_PROVIDER_IDS[model];
-    const native = entry?.[provider];
-    if (native === undefined) {
-        throw new Error(`resolveProviderModelId: model ${JSON.stringify(model)} is not available for provider ${JSON.stringify(provider)}; ` +
-            `available: ${providersForModel(model).join(", ") || "(none)"}`);
-    }
-    return native;
-}
-/**
- * The single model→provider resolver shared by the SDK and the CLI, wrapping
- * {@link providersForModel} with the forward-compat unknown-model allowance:
- *
- *   - `provider` given: it is honored. If `model` is a KNOWN id, the provider
- *     must serve it (else throw). If `model` is UNKNOWN, it is allowed through
- *     so a slightly-old client can still run a newly-launched model (the server
- *     arbitrates).
- *   - `provider` omitted: a known model resolves to its DEFAULT provider (first
- *     declared). An UNKNOWN model with no provider throws a `did you mean?`
- *     hint — you must name a provider to run a model this client doesn't know.
- *
- * Returns the resolved {@link ProviderName}.
- */
-export function resolveModelProvider(model, provider) {
-    const providers = providersForModel(model);
-    if (provider !== undefined) {
-        if (providers.length > 0 && !providers.includes(provider)) {
-            throw new Error(`model ${JSON.stringify(model)} is not available for provider ${provider}; ` +
-                `available: ${providers.join(", ")}`);
-        }
-        return provider;
-    }
-    const inferred = providers[0];
-    if (inferred === undefined) {
-        const hint = suggest(model, SUPPORTED_MODELS);
-        throw new Error(`${JSON.stringify(model)} is not a known model id` +
-            (hint ? ` (did you mean ${JSON.stringify(hint)}?)` : "") +
-            "; pass provider explicitly to run it");
-    }
-    return inferred;
-}
-export function isModelName(input) {
-    return typeof input === "string" && SUPPORTED_MODELS.includes(input);
-}
-export function parseModelName(input, field = "submission.model") {
+export function parseModelSlug(input, field = "submission.model") {
     try {
-        if (!isModelName(input)) {
-            throw new Error(`${field} must be one of: ${SUPPORTED_MODELS.join(", ")}`);
+        if (typeof input !== "string" || input.length === 0) {
+            throw new Error(`${field} must be a non-empty gateway model slug string ("creator/model")`);
+        }
+        if (!MODEL_SLUG_PATTERN.test(input)) {
+            throw new Error(`${field} must be a gateway model slug of the form "creator/model" matching ${MODEL_SLUG_PATTERN.source} (got ${JSON.stringify(input)})`);
         }
         return input;
     }
     catch (error) {
-        rethrowContractParseError(error, "parseModelName");
-    }
-}
-export function assertModelNameMatchesProvider(provider, model, field = "submission.model") {
-    const providers = providersForModel(model);
-    if (!providers.includes(provider)) {
-        throw new Error(`${field} ${JSON.stringify(model)} is not supported for provider ${provider}; ` +
-            `expected one of: ${providers.join(", ")}`);
+        rethrowContractParseError(error, "parseModelSlug");
     }
 }

package/dist/_contracts/runtime-types.d.ts

@@ -3,7 +3,7 @@ import type { RuntimeSize } from "./runtime-sizes.js";
 import type { RuntimeKind } from "./runtime-kind.js";
 import type { SessionCostProviderUsage } from "./session-cost.js";
 import type { ProviderFault } from "./provider-fault.js";
-import type { PlatformInlineSecrets, PlatformSubmission, ProviderName, SessionLimits, SessionWebhookSpec } from "./submission.js";
+import type { PlatformInlineSecrets, PlatformSubmission, SessionLimits, SessionWebhookSpec } from "./submission.js";
 /** Public run phases. Internal finalization remains behind the consistency barrier. */
 export declare const SESSION_RUN_PHASES: readonly ["queued", "starting", "running", "finished", "error"];
 export type SessionRunPhase = (typeof SESSION_RUN_PHASES)[number];
@@ -114,7 +114,6 @@ export type SessionSubmission = Omit<PlatformSubmission, "prompt"> & {
     readonly prompt?: readonly string[];
 };
 export interface SessionCreateRequest {
-    readonly provider: ProviderName;
     readonly submission: SessionSubmission;
     readonly runtimeSize?: RuntimeSize;
     readonly runtimeKind?: RuntimeKind;
@@ -181,7 +180,7 @@ export interface TurnResult {
     readonly status: SessionRunOutcome;
     /** True only when the run completed successfully. */
     readonly ok: boolean;
-    /** AEX showback estimate (USD, >= 0). Excludes BYOK provider spend. */
+    /** AEX showback estimate (USD, >= 0). Includes managed-gateway model tokens. */
     readonly costUsd: number;
     /** Aggregate token usage, derived from `costTelemetry.providerUsage`. */
     readonly usage: UsageSummary;

package/dist/_contracts/session-config.js

@@ -30,7 +30,7 @@
  * boundary.
  */
 import { rethrowContractParseError, withContractParseError } from "./contract-parse-error.js";
-import { parseModelName } from "./models.js";
+import { parseModelSlug } from "./models.js";
 import { assertAllowedKeys, defineAllowedKeys } from "./allowed-keys.js";
 // ---------------------------------------------------------------------------
 // Skill ID + name format
@@ -447,7 +447,7 @@ export function rejectStdioMcpShape(record) {
  * literal is a routable public address (or not an IP literal at all — name
  * resolution is the caller's concern). This numeric-range deny-list is kept
  * in parity across the public contract parser and platform shared parser so
- * the MCP parser, BYOK proxy handlers, and `submission.parseProxyBaseUrl`
+ * the MCP parser, the egress proxy handlers, and `submission.parseProxyBaseUrl`
  * classify the same bytes.
  *
  * `host` is the already-bracket-stripped, lowercased hostname.
@@ -630,7 +630,7 @@ export function parseSessionRequestConfig(input) {
         const record = input;
         const allowed = defineAllowedKeys()("model", "system", "prompt", "mcpServers", "environment", "runtimeSize", "timeout", "metadata");
         assertAllowedKeys(record, allowed, (key) => new Error(`session request config contains unexpected field: ${key}`));
-        const model = parseModelName(record.model, "session request config model");
+        const model = parseModelSlug(record.model, "session request config model");
         const system = record.system;
         if (system !== undefined && typeof system !== "string") {
             throw new Error("session request config system, when provided, must be a string");

package/dist/_contracts/session-cost.d.ts

@@ -117,8 +117,9 @@ export interface SessionCostTelemetry {
     /**
      * Customer-facing AEX cost of serving this session, USD — a REPORTED ESTIMATE,
      * not a charge (telemetry/showback only; no invoicing or credit deduction).
-     * = rawCostUsd × marginMultiplier (margin currently a global 1.0). EXCLUDES
-     * the customer's BYOK provider spend. The raw (pre-margin) figure is kept
+     * = rawCostUsd × marginMultiplier (margin currently a global 1.0). INCLUDES
+     * the run's managed-gateway model tokens, which aex serves on its own key and
+     * bills as a usage dimension. The raw (pre-margin) figure is kept
      * internal and never appears on this public-safe shape. A plain number, so it
      * passes the session-record public-safe archive scan. Absent when the session incurred
      * no priced AEX usage.