feat(catalog): route paid xAI through Responses like SuperGrok
Switch XAI_API_KEY models from Chat Completions to /v1/responses, default both xai and xai-oauth to grok-4.5, and include reasoning.encrypted_content.
This commit is contained in:
Yang Yang
@@ -235,12 +235,13 @@ Reasoning fields are not interchangeable.
|
||||
- Compat needs both policies: disable reasoning for any tool choice, and disable
|
||||
reasoning only for forced tool choice.
|
||||
|
||||
### xAI Grok through Responses/SuperGrok
|
||||
### xAI Grok through Responses (`xai` and `xai-oauth`)
|
||||
|
||||
Keep these independent:
|
||||
Both the paid API-key provider (`xai` / `XAI_API_KEY`) and SuperGrok OAuth
|
||||
(`xai-oauth`) chat over `https://api.x.ai/v1/responses`. Keep these independent:
|
||||
|
||||
- omit `reasoning.effort`
|
||||
- include or drop encrypted reasoning replay
|
||||
- include `reasoning.encrypted_content` (request `include`) vs replay history
|
||||
- filter reasoning-history wrappers
|
||||
|
||||
Some models reject only one of those fields; do not collapse them into one
|
||||
|
||||
@@ -5,9 +5,6 @@ import type { AssistantMessage, Context, FetchImpl, Model, SimpleStreamOptions,
|
||||
import { buildOpenAICompat } from "@oh-my-pi/pi-catalog/compat/openai";
|
||||
import { getBundledModel } from "@oh-my-pi/pi-catalog/models";
|
||||
|
||||
const model = getBundledModel<"openai-completions">("xai", "grok-code-fast-1");
|
||||
if (!model) throw new Error("Expected bundled xAI Grok model");
|
||||
if (model.api !== "openai-completions") throw new Error(`Expected Chat Completions model, received ${model.api}`);
|
||||
const context: Context = { messages: [{ role: "user", content: "hello", timestamp: 0 }] };
|
||||
|
||||
const openAI56ResponsesModel = getBundledModel<"openai-responses">("openai", "gpt-5.6");
|
||||
@@ -44,7 +41,7 @@ function chatCompletionsSse(): Response {
|
||||
id: "chatcmpl-affinity",
|
||||
object: "chat.completion.chunk",
|
||||
created: 0,
|
||||
model: model.id,
|
||||
model: openAI56CompletionsModel.id,
|
||||
choices: [{ index: 0, delta, finish_reason: finishReason }],
|
||||
});
|
||||
|
||||
@@ -56,7 +53,7 @@ function chatCompletionsSse(): Response {
|
||||
|
||||
async function captureRequest(
|
||||
options: OpenAICompletionsOptions,
|
||||
requestModel: Model<"openai-completions"> = model,
|
||||
requestModel: Model<"openai-completions"> = openAI56CompletionsModel,
|
||||
requestContext: Context = context,
|
||||
): Promise<{ headers: Headers; body: Record<string, unknown> }> {
|
||||
let requestHeaders: Headers | undefined;
|
||||
@@ -83,7 +80,7 @@ async function captureRequest(
|
||||
|
||||
async function captureSimpleRequest(
|
||||
options: SimpleStreamOptions,
|
||||
requestModel: Model<"openai-completions"> = model,
|
||||
requestModel: Model<"openai-completions"> = openAI56CompletionsModel,
|
||||
requestContext: Context = context,
|
||||
): Promise<{ headers: Headers; body: Record<string, unknown> }> {
|
||||
let requestHeaders: Headers | undefined;
|
||||
@@ -104,51 +101,6 @@ async function captureSimpleRequest(
|
||||
return { headers: requestHeaders, body };
|
||||
}
|
||||
|
||||
describe("openai-completions xAI cache affinity", () => {
|
||||
const cases: Array<{
|
||||
name: string;
|
||||
options: OpenAICompletionsOptions;
|
||||
expectedHeader: string | null;
|
||||
}> = [
|
||||
{
|
||||
name: "uses sessionId when no prompt cache key is provided",
|
||||
options: { sessionId: "session-fallback" },
|
||||
expectedHeader: "session-fallback",
|
||||
},
|
||||
{
|
||||
name: "keeps the prompt cache key stable across a distinct side-channel session",
|
||||
options: { promptCacheKey: "stable-cache-key", sessionId: "side-channel-session" },
|
||||
expectedHeader: "stable-cache-key",
|
||||
},
|
||||
{
|
||||
name: "omits automatic affinity when caching is disabled",
|
||||
options: {
|
||||
promptCacheKey: "disabled-cache-key",
|
||||
sessionId: "disabled-session",
|
||||
cacheRetention: "none",
|
||||
},
|
||||
expectedHeader: null,
|
||||
},
|
||||
{
|
||||
name: "preserves a caller-provided mixed-case affinity header",
|
||||
options: {
|
||||
promptCacheKey: "automatic-cache-key",
|
||||
sessionId: "automatic-session",
|
||||
headers: { "X-Grok-Conv-Id": "caller-affinity" },
|
||||
},
|
||||
expectedHeader: "caller-affinity",
|
||||
},
|
||||
];
|
||||
|
||||
for (const { name, options, expectedHeader } of cases) {
|
||||
it(name, async () => {
|
||||
const { headers } = await captureRequest(options);
|
||||
|
||||
expect(headers.get("x-grok-conv-id")).toBe(expectedHeader);
|
||||
});
|
||||
}
|
||||
});
|
||||
|
||||
describe("OpenAI Chat Completions explicit prompt cache policy", () => {
|
||||
const historicalContext: Context = {
|
||||
messages: [
|
||||
|
||||
@@ -57,6 +57,20 @@ const xaiOAuthResponsesModel: Model<"openai-responses"> = {
|
||||
reasoning: true,
|
||||
}),
|
||||
};
|
||||
const xaiApiKeyResponsesModel: Model<"openai-responses"> = {
|
||||
...model,
|
||||
id: "grok-code-fast-1",
|
||||
name: "Grok Code Fast 1",
|
||||
provider: "xai",
|
||||
baseUrl: "https://api.x.ai/v1",
|
||||
compat: buildOpenAIResponsesCompat({
|
||||
id: "grok-code-fast-1",
|
||||
name: "Grok Code Fast 1",
|
||||
provider: "xai",
|
||||
baseUrl: "https://api.x.ai/v1",
|
||||
reasoning: true,
|
||||
}),
|
||||
};
|
||||
|
||||
const openAI56ResponsesModel: Model<"openai-responses"> = {
|
||||
...model,
|
||||
@@ -675,6 +689,16 @@ describe("openai-responses cache affinity", () => {
|
||||
}
|
||||
});
|
||||
|
||||
it("sets x-grok-conv-id cache affinity for paid xai Responses requests", async () => {
|
||||
const captured = await captureDispatchedOpenAIResponseHeaders(
|
||||
{ sessionId: "session-fallback" },
|
||||
xaiApiKeyResponsesModel,
|
||||
);
|
||||
|
||||
expect(getHeader(captured.headers, "x-grok-conv-id")).toBe("session-fallback");
|
||||
expect(captured.body?.prompt_cache_key).toBe("session-fallback");
|
||||
});
|
||||
|
||||
it("sets OpenRouter Responses session_id from sessionId in the body", async () => {
|
||||
const captured = await captureOpenAIResponseHeaders(
|
||||
{ sessionId: "workflow-123", promptCacheKey: "cache-key-123" },
|
||||
|
||||
@@ -1000,7 +1000,7 @@ describe("Generate E2E Tests", () => {
|
||||
);
|
||||
});
|
||||
|
||||
describe.skipIf(!e2eApiKey("XAI_API_KEY"))("xAI Provider (grok-code-fast-1 via OpenAI Completions)", () => {
|
||||
describe.skipIf(!e2eApiKey("XAI_API_KEY"))("xAI Provider (grok-code-fast-1 via OpenAI Responses)", () => {
|
||||
const llm = getBundledModel("xai", "grok-code-fast-1");
|
||||
|
||||
it(
|
||||
|
||||
@@ -38,6 +38,23 @@ describe("effort-dial-less reasoner encoding (regression)", () => {
|
||||
expect(grokR.thinking).toBeUndefined();
|
||||
});
|
||||
|
||||
test("paid xai/grok-code-fast-1 reasons but carries no thinking config", () => {
|
||||
const grokCodeFast = getBundledModel("xai", "grok-code-fast-1");
|
||||
if (!grokCodeFast) throw new Error("xai/grok-code-fast-1 must be in bundled models.json");
|
||||
expect(grokCodeFast.api).toBe("openai-responses");
|
||||
expect(grokCodeFast.reasoning).toBe(true);
|
||||
expect(grokCodeFast.thinking).toBeUndefined();
|
||||
expect(getSupportedEfforts(grokCodeFast)).toEqual([]);
|
||||
});
|
||||
|
||||
test("paid xai/grok-4.3 keeps its effort dial", () => {
|
||||
const grok43 = getBundledModel("xai", "grok-4.3");
|
||||
if (!grok43) throw new Error("xai/grok-4.3 must be in bundled models.json");
|
||||
expect(grok43.api).toBe("openai-responses");
|
||||
expect(grok43.thinking).toBeDefined();
|
||||
expect(getSupportedEfforts(grok43).length).toBeGreaterThan(0);
|
||||
});
|
||||
|
||||
test("the no-dial encoding stays scoped to openai-responses*", () => {
|
||||
const claude = getBundledModel("anthropic", "claude-sonnet-4-6");
|
||||
if (!claude) throw new Error("anthropic/claude-sonnet-4-6 must be in bundled models.json");
|
||||
@@ -57,6 +74,7 @@ describe("xAI OAuth Responses reasoning payload (regression)", () => {
|
||||
const { params } = buildParams(grok45, singleUserContext, undefined, undefined);
|
||||
|
||||
expect(params.reasoning).toBeUndefined();
|
||||
expect(params.include).toContain("reasoning.encrypted_content");
|
||||
});
|
||||
|
||||
test("xai-oauth/grok-4.5 omits unsupported reasoning summary", () => {
|
||||
@@ -66,5 +84,15 @@ describe("xAI OAuth Responses reasoning payload (regression)", () => {
|
||||
const { params } = buildParams(grok45, singleUserContext, { reasoning: Effort.High }, undefined);
|
||||
|
||||
expect(params.reasoning).toEqual({ effort: "high" });
|
||||
expect(params.include).toContain("reasoning.encrypted_content");
|
||||
});
|
||||
|
||||
test("paid xai/grok-4.5 requests encrypted reasoning content", () => {
|
||||
const grok45 = getBundledModel<"openai-responses">("xai", "grok-4.5");
|
||||
if (!grok45) throw new Error("xai/grok-4.5 must be in bundled models.json");
|
||||
|
||||
const { params } = buildParams(grok45, singleUserContext, { reasoning: Effort.High }, undefined);
|
||||
|
||||
expect(params.include).toContain("reasoning.encrypted_content");
|
||||
});
|
||||
});
|
||||
|
||||
@@ -2,6 +2,13 @@
|
||||
|
||||
## [Unreleased]
|
||||
|
||||
### Changed
|
||||
|
||||
- Switched the paid xAI provider (`xai` / `XAI_API_KEY`) from Chat Completions to the OpenAI Responses API (`POST https://api.x.ai/v1/responses`), matching SuperGrok `xai-oauth`. Prompt-cache affinity (`x-grok-conv-id`), reasoning-effort allowlisting, and encrypted-reasoning replay rules are now shared across both first-party xAI hosts.
|
||||
- Changed the paid xAI (`XAI_API_KEY`) default model from `grok-4-fast-non-reasoning` to `grok-4.5`.
|
||||
- Changed the SuperGrok (`xai-oauth`) default model from `grok-4.3` to `grok-4.5`.
|
||||
- Requested `reasoning.encrypted_content` on first-party xAI Responses calls (`xai` and `xai-oauth`) via the `include` parameter.
|
||||
|
||||
## [17.3.4] - 2026-08-14
|
||||
|
||||
### Added
|
||||
|
||||
@@ -684,24 +684,28 @@ export function buildOpenAIResponsesCompat(spec: OpenAIResponsesSpecLike): Resol
|
||||
const isLocalServingBackend =
|
||||
(!PROXY_OPENAI_COMPAT_PROVIDERS.has(spec.provider) && LOCAL_OPENAI_COMPAT_PROVIDERS.has(spec.provider)) ||
|
||||
hasLocalLoopbackBaseUrl(baseUrl);
|
||||
const isXaiHost = modelMatchesHost({ provider: spec.provider, baseUrl }, "xai");
|
||||
|
||||
const compat: ResolvedOpenAIResponsesCompat = {
|
||||
supportsDeveloperRole: isAzure || isOpenAIUrl || hostMatchesUrl(baseUrl, "githubCopilot"),
|
||||
supportsStrictMode: isAzure || detectStrictModeSupport(spec.provider, baseUrl),
|
||||
supportsReasoningEffort: spec.provider !== "xai-oauth" || isGrokReasoningEffortCapable(id),
|
||||
// Paid `xai` and SuperGrok `xai-oauth` share api.x.ai `/v1/responses`.
|
||||
// Only the Grok effort-capable allowlist accepts `reasoning.effort`;
|
||||
// other reasoners (grok-build, grok-code-fast-1, …) 400 if it is sent.
|
||||
supportsReasoningEffort: !isXaiHost || isGrokReasoningEffortCapable(id),
|
||||
supportsLongPromptCacheRetention: isOpenAIUrl,
|
||||
supportsPromptCacheBreakpoints,
|
||||
promptCacheBreakpointTtl: supportsPromptCacheBreakpoints ? "30m" : undefined,
|
||||
// Azure OpenAI and GitHub Copilot Responses paths require tool results
|
||||
// to strictly match prior tool calls when building Responses inputs.
|
||||
strictResponsesPairing: isAzure || spec.provider === "github-copilot",
|
||||
// GitHub Copilot and xAI OAuth reject `detail: "original"` (400 / 422).
|
||||
// Every other host preserves native-resolution frames (snapcompact relies
|
||||
// on `original`). Detect Copilot by provider id or base-URL host so a
|
||||
// model pointed at the Copilot host under a different provider id still
|
||||
// clamps; xai-oauth is provider-id only (same host family as paid `xai`).
|
||||
// GitHub Copilot and first-party xAI `/v1/responses` reject
|
||||
// `detail: "original"` (400 / 422). Every other host preserves
|
||||
// native-resolution frames (snapcompact relies on `original`). Detect
|
||||
// Copilot by provider id or base-URL host so a model pointed at the
|
||||
// Copilot host under a different provider id still clamps.
|
||||
supportsImageDetailOriginal:
|
||||
spec.provider !== "xai-oauth" && !modelMatchesHost({ provider: spec.provider, baseUrl }, "githubCopilot"),
|
||||
!isXaiHost && !modelMatchesHost({ provider: spec.provider, baseUrl }, "githubCopilot"),
|
||||
reasoningEffortMap: {},
|
||||
supportsReasoningParams: true,
|
||||
// OpenAI proprietary reasoning models (o-series, gpt-5+) reject explicit
|
||||
@@ -710,8 +714,12 @@ export function buildOpenAIResponsesCompat(spec: OpenAIResponsesSpecLike): Resol
|
||||
thinkingFormat,
|
||||
reasoningDisableMode: resolveReasoningDisableMode(thinkingFormat),
|
||||
omitReasoningEffort: false,
|
||||
includeEncryptedReasoning: spec.provider !== "xai-oauth",
|
||||
filterReasoningHistory: spec.provider === "xai-oauth" || (isOpenRouter && isAnthropicModel),
|
||||
// Ask xAI `/v1/responses` for `reasoning.encrypted_content` the same way
|
||||
// first-party OpenAI Responses does. History still drops `type:
|
||||
// "reasoning"` wrappers (`filterReasoningHistory`) independently —
|
||||
// those two flags must not be collapsed.
|
||||
includeEncryptedReasoning: true,
|
||||
filterReasoningHistory: isXaiHost || (isOpenRouter && isAnthropicModel),
|
||||
disableReasoningOnForcedToolChoice: isKimiModel,
|
||||
disableReasoningOnToolChoice: isDeepseekFamily && reasoningCapable && !isOpenRouter,
|
||||
supportsToolChoice: true,
|
||||
@@ -752,7 +760,7 @@ export function buildOpenAIResponsesCompat(spec: OpenAIResponsesSpecLike): Resol
|
||||
MINIMAX_PROVIDER_OR_ID_PATTERN.test(spec.provider) || (id ? MINIMAX_PROVIDER_OR_ID_PATTERN.test(id) : false),
|
||||
emptyLengthFinishIsContextError: spec.provider === "ollama",
|
||||
usesOpenAIToolCallIdLimit: spec.provider === "openai",
|
||||
promptCacheSessionHeader: spec.provider === "xai-oauth" ? "x-grok-conv-id" : undefined,
|
||||
promptCacheSessionHeader: isXaiHost ? "x-grok-conv-id" : undefined,
|
||||
streamFirstEventTimeoutMs: isLocalServingBackend ? 0 : spec.compat?.streamFirstEventTimeoutMs,
|
||||
streamIdleTimeoutMs: isLocalServingBackend
|
||||
? LOCAL_OPENAI_COMPAT_STREAM_IDLE_TIMEOUT_MS
|
||||
|
||||
@@ -47,7 +47,7 @@ export const KNOWN_HOSTS = {
|
||||
},
|
||||
umans: { providers: ["umans"], urlMarkers: ["api.code.umans.ai"] },
|
||||
xiaomi: { providers: ["xiaomi"], providerPrefixes: ["xiaomi-token-plan-"], urlMarkers: ["xiaomimimo.com"] },
|
||||
xai: { providers: ["xai"], urlMarkers: ["api.x.ai"] },
|
||||
xai: { providers: ["xai", "xai-oauth"], urlMarkers: ["api.x.ai"] },
|
||||
mistral: { providers: ["mistral"], urlMarkers: ["mistral.ai"] },
|
||||
together: { providers: ["together"], urlMarkers: ["api.together.xyz"] },
|
||||
baseten: { providers: ["baseten"], urlMarkers: ["baseten.co"] },
|
||||
|
||||
@@ -103716,7 +103716,7 @@
|
||||
"grok-2": {
|
||||
"id": "grok-2",
|
||||
"name": "Grok 2",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": false,
|
||||
@@ -103735,7 +103735,7 @@
|
||||
"grok-2-1212": {
|
||||
"id": "grok-2-1212",
|
||||
"name": "Grok 2 (1212)",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": false,
|
||||
@@ -103754,7 +103754,7 @@
|
||||
"grok-2-latest": {
|
||||
"id": "grok-2-latest",
|
||||
"name": "Grok 2 Latest",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": false,
|
||||
@@ -103773,7 +103773,7 @@
|
||||
"grok-2-vision": {
|
||||
"id": "grok-2-vision",
|
||||
"name": "Grok 2 Vision",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": false,
|
||||
@@ -103793,7 +103793,7 @@
|
||||
"grok-2-vision-1212": {
|
||||
"id": "grok-2-vision-1212",
|
||||
"name": "Grok 2 Vision (1212)",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": false,
|
||||
@@ -103813,7 +103813,7 @@
|
||||
"grok-2-vision-latest": {
|
||||
"id": "grok-2-vision-latest",
|
||||
"name": "Grok 2 Vision Latest",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": false,
|
||||
@@ -103833,7 +103833,7 @@
|
||||
"grok-3": {
|
||||
"id": "grok-3",
|
||||
"name": "Grok 3",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": false,
|
||||
@@ -103852,7 +103852,7 @@
|
||||
"grok-3-fast": {
|
||||
"id": "grok-3-fast",
|
||||
"name": "Grok 3 Fast",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": false,
|
||||
@@ -103871,7 +103871,7 @@
|
||||
"grok-3-fast-latest": {
|
||||
"id": "grok-3-fast-latest",
|
||||
"name": "Grok 3 Fast Latest",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": false,
|
||||
@@ -103890,7 +103890,7 @@
|
||||
"grok-3-latest": {
|
||||
"id": "grok-3-latest",
|
||||
"name": "Grok 3 Latest",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": false,
|
||||
@@ -103909,7 +103909,7 @@
|
||||
"grok-3-mini": {
|
||||
"id": "grok-3-mini",
|
||||
"name": "Grok 3 Mini",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": true,
|
||||
@@ -103937,7 +103937,7 @@
|
||||
"grok-3-mini-fast": {
|
||||
"id": "grok-3-mini-fast",
|
||||
"name": "Grok 3 Mini Fast",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": true,
|
||||
@@ -103965,7 +103965,7 @@
|
||||
"grok-3-mini-fast-latest": {
|
||||
"id": "grok-3-mini-fast-latest",
|
||||
"name": "Grok 3 Mini Fast Latest",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": true,
|
||||
@@ -103993,7 +103993,7 @@
|
||||
"grok-3-mini-latest": {
|
||||
"id": "grok-3-mini-latest",
|
||||
"name": "Grok 3 Mini Latest",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": true,
|
||||
@@ -104021,7 +104021,7 @@
|
||||
"grok-4": {
|
||||
"id": "grok-4",
|
||||
"name": "Grok 4",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": true,
|
||||
@@ -104049,7 +104049,7 @@
|
||||
"grok-4-1-fast": {
|
||||
"id": "grok-4-1-fast",
|
||||
"name": "Grok 4.1 Fast",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": true,
|
||||
@@ -104078,7 +104078,7 @@
|
||||
"grok-4-1-fast-non-reasoning": {
|
||||
"id": "grok-4-1-fast-non-reasoning",
|
||||
"name": "Grok 4.1 Fast (Non-Reasoning)",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": false,
|
||||
@@ -104098,7 +104098,7 @@
|
||||
"grok-4-fast": {
|
||||
"id": "grok-4-fast",
|
||||
"name": "Grok 4 Fast",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": true,
|
||||
@@ -104127,7 +104127,7 @@
|
||||
"grok-4-fast-non-reasoning": {
|
||||
"id": "grok-4-fast-non-reasoning",
|
||||
"name": "Grok 4 Fast (Non-Reasoning)",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": false,
|
||||
@@ -104147,7 +104147,7 @@
|
||||
"grok-4.20-0309-non-reasoning": {
|
||||
"id": "grok-4.20-0309-non-reasoning",
|
||||
"name": "Grok 4.20 (Non-Reasoning)",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": false,
|
||||
@@ -104167,7 +104167,7 @@
|
||||
"grok-4.20-0309-reasoning": {
|
||||
"id": "grok-4.20-0309-reasoning",
|
||||
"name": "Grok 4.20 (Reasoning)",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": true,
|
||||
@@ -104197,7 +104197,7 @@
|
||||
"grok-4.20-beta-latest-non-reasoning": {
|
||||
"id": "grok-4.20-beta-latest-non-reasoning",
|
||||
"name": "Grok 4.20 Beta (Non-Reasoning)",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": false,
|
||||
@@ -104217,7 +104217,7 @@
|
||||
"grok-4.20-beta-latest-reasoning": {
|
||||
"id": "grok-4.20-beta-latest-reasoning",
|
||||
"name": "Grok 4.20 Beta (Reasoning)",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": true,
|
||||
@@ -104247,7 +104247,7 @@
|
||||
"grok-4.20-multi-agent-beta-latest": {
|
||||
"id": "grok-4.20-multi-agent-beta-latest",
|
||||
"name": "Grok 4.20 Multi-Agent Beta",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": true,
|
||||
@@ -104276,7 +104276,7 @@
|
||||
"grok-4.3": {
|
||||
"id": "grok-4.3",
|
||||
"name": "Grok 4.3",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": true,
|
||||
@@ -104305,7 +104305,7 @@
|
||||
"grok-4.5": {
|
||||
"id": "grok-4.5",
|
||||
"name": "Grok 4.5",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": true,
|
||||
@@ -104363,7 +104363,7 @@
|
||||
"grok-beta": {
|
||||
"id": "grok-beta",
|
||||
"name": "Grok Beta",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": false,
|
||||
@@ -104382,7 +104382,7 @@
|
||||
"grok-build-0.1": {
|
||||
"id": "grok-build-0.1",
|
||||
"name": "Grok Build 0.1",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": true,
|
||||
@@ -104411,7 +104411,7 @@
|
||||
"grok-code-fast-1": {
|
||||
"id": "grok-code-fast-1",
|
||||
"name": "Grok Code Fast 1",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": true,
|
||||
@@ -104439,7 +104439,7 @@
|
||||
"grok-vision-beta": {
|
||||
"id": "grok-vision-beta",
|
||||
"name": "Grok Vision Beta",
|
||||
"api": "openai-completions",
|
||||
"api": "openai-responses",
|
||||
"provider": "xai",
|
||||
"baseUrl": "https://api.x.ai/v1",
|
||||
"reasoning": false,
|
||||
@@ -104483,7 +104483,7 @@
|
||||
"reasoningEffortMap": {
|
||||
"minimal": "low"
|
||||
},
|
||||
"includeEncryptedReasoning": false,
|
||||
"includeEncryptedReasoning": true,
|
||||
"filterReasoningHistory": true,
|
||||
"supportsImageDetailOriginal": false,
|
||||
"omitReasoningEffort": true,
|
||||
@@ -104515,7 +104515,7 @@
|
||||
"reasoningEffortMap": {
|
||||
"minimal": "low"
|
||||
},
|
||||
"includeEncryptedReasoning": false,
|
||||
"includeEncryptedReasoning": true,
|
||||
"filterReasoningHistory": true,
|
||||
"supportsImageDetailOriginal": false,
|
||||
"omitReasoningEffort": true,
|
||||
@@ -104559,7 +104559,7 @@
|
||||
"reasoningEffortMap": {
|
||||
"minimal": "low"
|
||||
},
|
||||
"includeEncryptedReasoning": false,
|
||||
"includeEncryptedReasoning": true,
|
||||
"filterReasoningHistory": true,
|
||||
"supportsImageDetailOriginal": false,
|
||||
"omitReasoningEffort": false,
|
||||
@@ -104604,7 +104604,7 @@
|
||||
"reasoningEffortMap": {
|
||||
"minimal": "low"
|
||||
},
|
||||
"includeEncryptedReasoning": false,
|
||||
"includeEncryptedReasoning": true,
|
||||
"filterReasoningHistory": true,
|
||||
"supportsImageDetailOriginal": false,
|
||||
"omitReasoningEffort": false,
|
||||
@@ -104649,7 +104649,7 @@
|
||||
"reasoningEffortMap": {
|
||||
"minimal": "low"
|
||||
},
|
||||
"includeEncryptedReasoning": false,
|
||||
"includeEncryptedReasoning": true,
|
||||
"filterReasoningHistory": true,
|
||||
"supportsImageDetailOriginal": false,
|
||||
"omitReasoningEffort": false,
|
||||
@@ -104709,7 +104709,7 @@
|
||||
"reasoningEffortMap": {
|
||||
"minimal": "low"
|
||||
},
|
||||
"includeEncryptedReasoning": false,
|
||||
"includeEncryptedReasoning": true,
|
||||
"filterReasoningHistory": true,
|
||||
"supportsImageDetailOriginal": false,
|
||||
"omitReasoningEffort": true,
|
||||
@@ -104741,7 +104741,7 @@
|
||||
"reasoningEffortMap": {
|
||||
"minimal": "low"
|
||||
},
|
||||
"includeEncryptedReasoning": false,
|
||||
"includeEncryptedReasoning": true,
|
||||
"filterReasoningHistory": true,
|
||||
"supportsImageDetailOriginal": false,
|
||||
"omitReasoningEffort": true,
|
||||
@@ -104772,7 +104772,7 @@
|
||||
"reasoningEffortMap": {
|
||||
"minimal": "low"
|
||||
},
|
||||
"includeEncryptedReasoning": false,
|
||||
"includeEncryptedReasoning": true,
|
||||
"filterReasoningHistory": true,
|
||||
"supportsImageDetailOriginal": false,
|
||||
"omitReasoningEffort": true,
|
||||
@@ -111996,4 +111996,4 @@
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
@@ -478,13 +478,13 @@ export const CATALOG_PROVIDERS = [
|
||||
},
|
||||
{
|
||||
id: "xai",
|
||||
defaultModel: "grok-4-fast-non-reasoning",
|
||||
defaultModel: "grok-4.5",
|
||||
envVars: ["XAI_API_KEY"],
|
||||
createModelManagerOptions: (config: ModelManagerConfig) => xaiModelManagerOptions(config),
|
||||
},
|
||||
{
|
||||
id: "xai-oauth",
|
||||
defaultModel: "grok-4.3",
|
||||
defaultModel: "grok-4.5",
|
||||
envVars: ["XAI_OAUTH_TOKEN", "XAI_API_KEY"],
|
||||
createModelManagerOptions: (config: ModelManagerConfig) => xaiOAuthModelManagerOptions(config),
|
||||
catalogDiscovery: {
|
||||
|
||||
@@ -1257,8 +1257,8 @@ export interface XaiModelManagerConfig {
|
||||
fetch?: FetchImpl;
|
||||
}
|
||||
|
||||
export function xaiModelManagerOptions(config?: XaiModelManagerConfig): ModelManagerOptions<"openai-completions"> {
|
||||
return createSimpleOpenAICompletionsOptions("xai", "https://api.x.ai/v1", config);
|
||||
export function xaiModelManagerOptions(config?: XaiModelManagerConfig): ModelManagerOptions<"openai-responses"> {
|
||||
return createSimpleOpenAIResponsesOptions("xai", "https://api.x.ai/v1", config);
|
||||
}
|
||||
|
||||
export interface XaiOAuthModelManagerConfig {
|
||||
@@ -1352,7 +1352,7 @@ const XAI_NON_CHAT_PREFIXES = ["grok-imagine-", "grok-stt-", "grok-voice-"] as c
|
||||
function withXaiOAuthCompatDefaults(model: ModelSpec<"openai-responses">): ModelSpec<"openai-responses"> {
|
||||
const compat = {
|
||||
...(model.compat ?? {}),
|
||||
includeEncryptedReasoning: model.compat?.includeEncryptedReasoning ?? false,
|
||||
includeEncryptedReasoning: model.compat?.includeEncryptedReasoning ?? true,
|
||||
filterReasoningHistory: model.compat?.filterReasoningHistory ?? true,
|
||||
supportsImageDetailOriginal: model.compat?.supportsImageDetailOriginal ?? false,
|
||||
omitReasoningEffort: model.compat?.omitReasoningEffort ?? !isGrokReasoningEffortCapable(model.id),
|
||||
@@ -1395,7 +1395,7 @@ function mergeCuratedIntoModel(
|
||||
const compat = {
|
||||
...(base.compat ?? {}),
|
||||
reasoningEffortMap: { ...XAI_REASONING_EFFORT_MAP, ...(base.compat?.reasoningEffortMap ?? {}) },
|
||||
includeEncryptedReasoning: base.compat?.includeEncryptedReasoning ?? false,
|
||||
includeEncryptedReasoning: base.compat?.includeEncryptedReasoning ?? true,
|
||||
filterReasoningHistory: base.compat?.filterReasoningHistory ?? true,
|
||||
supportsImageDetailOriginal: base.compat?.supportsImageDetailOriginal ?? false,
|
||||
omitReasoningEffort: !effortCapable,
|
||||
@@ -5713,6 +5713,15 @@ function openAiCompletionsDescriptor(
|
||||
return simpleModelsDevDescriptor(modelsDevKey, providerId, "openai-completions", baseUrl, options);
|
||||
}
|
||||
|
||||
function openAiResponsesDescriptor(
|
||||
modelsDevKey: string,
|
||||
providerId: string,
|
||||
baseUrl: string,
|
||||
options: Omit<ModelsDevProviderDescriptor, "modelsDevKey" | "providerId" | "api" | "baseUrl"> = {},
|
||||
): ModelsDevProviderDescriptor {
|
||||
return simpleModelsDevDescriptor(modelsDevKey, providerId, "openai-responses", baseUrl, options);
|
||||
}
|
||||
|
||||
function anthropicMessagesDescriptor(
|
||||
modelsDevKey: string,
|
||||
providerId: string,
|
||||
@@ -5837,7 +5846,7 @@ const MODELS_DEV_PROVIDER_DESCRIPTORS_CORE: readonly ModelsDevProviderDescriptor
|
||||
defaultContextWindow: 131072,
|
||||
}),
|
||||
// --- xAI ---
|
||||
openAiCompletionsDescriptor("xai", "xai", "https://api.x.ai/v1"),
|
||||
openAiResponsesDescriptor("xai", "xai", "https://api.x.ai/v1"),
|
||||
// --- DeepSeek ---
|
||||
openAiCompletionsDescriptor("deepseek", "deepseek", "https://api.deepseek.com", {
|
||||
// Only ship the v4 family as built-ins; older deepseek-chat / deepseek-reasoner
|
||||
|
||||
@@ -223,12 +223,15 @@ describe("buildModel", () => {
|
||||
});
|
||||
});
|
||||
|
||||
describe("xAI-OAuth Responses reasoning-effort suppression", () => {
|
||||
const grokResponsesSpec = (id: string): ModelSpec<"openai-responses"> => ({
|
||||
describe("xAI Responses reasoning-effort suppression", () => {
|
||||
const grokResponsesSpec = (
|
||||
id: string,
|
||||
provider: "xai" | "xai-oauth" = "xai-oauth",
|
||||
): ModelSpec<"openai-responses"> => ({
|
||||
id,
|
||||
name: id,
|
||||
api: "openai-responses",
|
||||
provider: "xai-oauth",
|
||||
provider,
|
||||
baseUrl: "https://api.x.ai/v1",
|
||||
reasoning: true,
|
||||
input: ["text"],
|
||||
@@ -248,6 +251,28 @@ describe("xAI-OAuth Responses reasoning-effort suppression", () => {
|
||||
expect(buildOpenAIResponsesCompat(grokResponsesSpec("grok-4.3")).supportsReasoningEffort).toBe(true);
|
||||
});
|
||||
|
||||
it("applies the same Responses dialect to paid xai and xai-oauth", () => {
|
||||
const paid = buildOpenAIResponsesCompat(grokResponsesSpec("grok-4.3", "xai"));
|
||||
const oauth = buildOpenAIResponsesCompat(grokResponsesSpec("grok-4.3", "xai-oauth"));
|
||||
expect(paid.promptCacheSessionHeader).toBe("x-grok-conv-id");
|
||||
expect(oauth.promptCacheSessionHeader).toBe("x-grok-conv-id");
|
||||
expect(paid.includeEncryptedReasoning).toBe(true);
|
||||
expect(oauth.includeEncryptedReasoning).toBe(true);
|
||||
expect(paid.filterReasoningHistory).toBe(true);
|
||||
expect(oauth.filterReasoningHistory).toBe(true);
|
||||
expect(paid.supportsImageDetailOriginal).toBe(false);
|
||||
expect(oauth.supportsImageDetailOriginal).toBe(false);
|
||||
expect(paid.supportsReasoningEffort).toBe(true);
|
||||
expect(oauth.supportsReasoningEffort).toBe(true);
|
||||
});
|
||||
|
||||
it("omits effort for paid xai models off the Grok allowlist", () => {
|
||||
const compat = buildOpenAIResponsesCompat(grokResponsesSpec("grok-code-fast-1", "xai"));
|
||||
expect(compat.supportsReasoningEffort).toBe(false);
|
||||
expect(compat.omitReasoningEffort).toBe(true);
|
||||
expect(buildModel(grokResponsesSpec("grok-code-fast-1", "xai")).thinking).toBeUndefined();
|
||||
});
|
||||
|
||||
it("lets an explicit compat.supportsReasoningEffort override the allowlist default", () => {
|
||||
const compat = buildOpenAIResponsesCompat({
|
||||
...grokResponsesSpec("grok-build"),
|
||||
@@ -256,7 +281,7 @@ describe("xAI-OAuth Responses reasoning-effort suppression", () => {
|
||||
expect(compat.supportsReasoningEffort).toBe(true);
|
||||
});
|
||||
|
||||
it("does not suppress effort for a non-xai-oauth provider with a grok-like id", () => {
|
||||
it("does not suppress effort for a non-xAI provider with a grok-like id", () => {
|
||||
const compat = buildOpenAIResponsesCompat({
|
||||
...grokResponsesSpec("grok-build"),
|
||||
provider: "openai",
|
||||
|
||||
@@ -0,0 +1,25 @@
|
||||
import { describe, expect, it } from "bun:test";
|
||||
import { getBundledModels } from "@oh-my-pi/pi-catalog/models";
|
||||
import { CATALOG_PROVIDERS } from "@oh-my-pi/pi-catalog/provider-models/descriptors";
|
||||
import { xaiModelManagerOptions } from "@oh-my-pi/pi-catalog/provider-models/openai-compat";
|
||||
|
||||
describe("paid xai (XAI_API_KEY) Responses contract", () => {
|
||||
it("registers xai on the catalog Responses discovery path", () => {
|
||||
const entry = CATALOG_PROVIDERS.find(provider => provider.id === "xai");
|
||||
expect(entry, "xai catalog descriptor").toBeDefined();
|
||||
expect(entry!.defaultModel).toBe("grok-4.5");
|
||||
expect(entry!.envVars).toContain("XAI_API_KEY");
|
||||
const options = xaiModelManagerOptions({ apiKey: "test-key" });
|
||||
expect(options.providerId).toBe("xai");
|
||||
expect(options.fetchDynamicModels, "live /v1/models overlay").toBeTypeOf("function");
|
||||
});
|
||||
|
||||
it("bundles every paid xai chat model on openai-responses", () => {
|
||||
const models = getBundledModels("xai");
|
||||
expect(models.length).toBeGreaterThan(0);
|
||||
for (const model of models) {
|
||||
expect(model.api, `${model.provider}/${model.id}`).toBe("openai-responses");
|
||||
expect(model.baseUrl).toBe("https://api.x.ai/v1");
|
||||
}
|
||||
});
|
||||
});
|
||||
@@ -0,0 +1,86 @@
|
||||
import { describe, expect, it } from "bun:test";
|
||||
import MODELS_JSON from "@oh-my-pi/pi-catalog/models.json" with { type: "json" };
|
||||
import { CATALOG_PROVIDERS, DEFAULT_MODEL_PER_PROVIDER } from "@oh-my-pi/pi-catalog/provider-models/descriptors";
|
||||
import { buildXaiOAuthStaticSeed } from "@oh-my-pi/pi-catalog/provider-models/openai-compat";
|
||||
import type { ModelSpec } from "@oh-my-pi/pi-catalog/types";
|
||||
|
||||
// Pins the invariant: bundled `models.json` carries every entry the runtime
|
||||
// curated catalog (XAI_OAUTH_CURATED_MODELS, surfaced via
|
||||
// buildXaiOAuthStaticSeed) emits. Without this, editing the curated list
|
||||
// without regenerating `models.json` silently regresses the boot-time
|
||||
// default-model resolver — the registry sees the runtime seed only after
|
||||
// `refresh()`, but interactive boot resolves the persisted default
|
||||
// synchronously from `#loadModels()`, which reads only `models.json`.
|
||||
//
|
||||
// Failure here means: run `bun run gen:models` and commit the diff.
|
||||
describe("xai-oauth bundled catalog (regression)", () => {
|
||||
const bundled =
|
||||
(MODELS_JSON as unknown as Record<string, Record<string, ModelSpec<"openai-responses">>>)["xai-oauth"] ?? {};
|
||||
const seed = buildXaiOAuthStaticSeed();
|
||||
|
||||
it("defaults SuperGrok selection to grok-4.5", () => {
|
||||
const entry = CATALOG_PROVIDERS.find(provider => provider.id === "xai-oauth");
|
||||
expect(entry?.defaultModel).toBe("grok-4.5");
|
||||
expect(DEFAULT_MODEL_PER_PROVIDER["xai-oauth"]).toBe("grok-4.5");
|
||||
expect(bundled["grok-4.5"], "xai-oauth/grok-4.5 must be bundled for the default").toBeDefined();
|
||||
});
|
||||
|
||||
it("bundles every curated id", () => {
|
||||
const seededIds = seed.map(model => model.id).sort();
|
||||
const bundledIds = Object.keys(bundled).sort();
|
||||
expect(bundledIds).toEqual(seededIds);
|
||||
});
|
||||
|
||||
for (const seededModel of seed) {
|
||||
it(`matches contract for ${seededModel.id}`, () => {
|
||||
const bundledEntry = bundled[seededModel.id];
|
||||
expect(bundledEntry, `xai-oauth/${seededModel.id} missing from models.json`).toBeDefined();
|
||||
expect(bundledEntry.id).toBe(seededModel.id);
|
||||
expect(bundledEntry.name).toBe(seededModel.name);
|
||||
expect(bundledEntry.provider).toBe("xai-oauth");
|
||||
expect(bundledEntry.api).toBe("openai-responses");
|
||||
expect(bundledEntry.contextWindow).toBe(seededModel.contextWindow);
|
||||
expect(bundledEntry.reasoning).toBe(seededModel.reasoning);
|
||||
// Input modality must survive both the curated seed and the bundle.
|
||||
// Without this the static fallback used on offline boot strips
|
||||
// vision capability silently (Codex PR #1127 review).
|
||||
expect(bundledEntry.input).toEqual(seededModel.input);
|
||||
expect(bundledEntry.compat?.supportsReasoningEffort).toBe(seededModel.compat?.supportsReasoningEffort);
|
||||
});
|
||||
}
|
||||
|
||||
// Absolute contract for the user-specified SuperGrok addition. The parity
|
||||
// loop above can't catch a value typo (e.g. 2_000_000) or a flipped
|
||||
// reasoning flag — both sides regenerate from the same seed together — so
|
||||
// pin the literal attributes here.
|
||||
it("exposes grok-composer-2.5-fast as a non-reasoning 200K text model", () => {
|
||||
const composer = seed.find(model => model.id === "grok-composer-2.5-fast");
|
||||
expect(composer, "grok-composer-2.5-fast must be in the SuperGrok curated seed").toBeDefined();
|
||||
expect(composer!.reasoning).toBe(false);
|
||||
expect(composer!.contextWindow).toBe(200_000);
|
||||
expect(composer!.input).toEqual(["text"]);
|
||||
// The bundled models.json entry is byte-identical to the generator's
|
||||
// deterministic xai-oauth output: gen:models pushes
|
||||
// buildXaiOAuthStaticSeed() (offline — xai-oauth has no upstream catalog
|
||||
// source) and applyGeneratedModelPolicies(), so a regen reproduces these
|
||||
// exact bytes; only unrelated other-provider network churn was excluded
|
||||
// to keep the diff scoped. Pin its zero-cost invariant (overlay-stable
|
||||
// for the SuperGrok subscription), which the parity loop above never
|
||||
// compares. (maxTokens is pinned by the maxTokens-equals-contextWindow
|
||||
// test below.)
|
||||
expect(bundled["grok-composer-2.5-fast"]?.cost).toEqual({ input: 0, output: 0, cacheRead: 0, cacheWrite: 0 });
|
||||
});
|
||||
|
||||
// The OAuth surface's /v1/models reports no per-request output limit, so the
|
||||
// curated catalog owns maxTokens — set to mirror each model's contextWindow
|
||||
// (the openai-responses wire still clamps the actual request to
|
||||
// OPENAI_MAX_OUTPUT_TOKENS). Pin maxTokens === contextWindow on both the
|
||||
// static-seed and bundled paths so a null placeholder can
|
||||
// never silently leak back into the bundle.
|
||||
it("sets maxTokens equal to contextWindow for every xai-oauth model", () => {
|
||||
for (const model of seed) {
|
||||
expect(model.maxTokens, `seed ${model.id} maxTokens`).toBe(model.contextWindow);
|
||||
expect(bundled[model.id]?.maxTokens, `bundled ${model.id} maxTokens`).toBe(model.contextWindow);
|
||||
}
|
||||
});
|
||||
});
|
||||
@@ -2,6 +2,13 @@
|
||||
|
||||
## [Unreleased]
|
||||
|
||||
### Changed
|
||||
|
||||
- Routed paid xAI models (`XAI_API_KEY` / `xai/…`) through the Responses API used by SuperGrok OAuth instead of Chat Completions.
|
||||
- Changed the default model for `XAI_API_KEY` (`xai`) from `grok-4-fast-non-reasoning` to `grok-4.5`.
|
||||
- Changed the default model for SuperGrok OAuth (`xai-oauth`) from `grok-4.3` to `grok-4.5`.
|
||||
- Included `reasoning.encrypted_content` in Responses `include` for paid xAI and SuperGrok OAuth models.
|
||||
|
||||
## [17.3.4] - 2026-08-14
|
||||
|
||||
### Changed
|
||||
|
||||
Reference in New Issue
Block a user