@ottocode/server 0.1.263 → 0.1.265

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@ottocode/server",
-	"version": "0.1.263",
+	"version": "0.1.265",
 	"description": "HTTP API server for ottocode",
 	"type": "module",
 	"main": "./src/index.ts",
@@ -61,8 +61,8 @@
 		"typecheck": "tsc --noEmit"
 	},
 	"dependencies": {
-		"@ottocode/database": "0.1.263",
-		"@ottocode/sdk": "0.1.263",
+		"@ottocode/database": "0.1.265",
+		"@ottocode/sdk": "0.1.265",
 		"@hono/zod-openapi": "^1.1.5",
 		"ai-sdk-ollama": "^3.8.3",
 		"drizzle-orm": "^0.44.5",

package/src/runtime/agent/runner.ts

@@ -26,7 +26,6 @@ import {
 } from '../stream/handlers.ts';
 import {
 	pruneSession,
-	getModelLimits,
 	shouldAutoCompactBeforeOverflow,
 } from '../message/compaction.ts';
 import { triggerDeferredTitleGeneration } from '../message/service.ts';
@@ -131,7 +130,6 @@ async function shouldPreemptivelyAutoCompact(
 	opts: RunOpts,
 	threshold: number | null | undefined,
 ): Promise<boolean> {
-	const limits = getModelLimits(opts.provider, opts.model);
 	const sessionRows = await db
 		.select({ currentContextTokens: sessions.currentContextTokens })
 		.from(sessions)
@@ -140,7 +138,6 @@ async function shouldPreemptivelyAutoCompact(
 
 	return shouldAutoCompactBeforeOverflow({
 		autoCompactThresholdTokens: threshold,
-		modelContextWindow: limits?.context ?? null,
 		currentContextTokens: sessionRows[0]?.currentContextTokens ?? 0,
 		estimatedInputTokens: opts.estimatedInputTokens ?? 0,
 		isCompactCommand: opts.isCompactCommand,

package/src/runtime/ask/service.ts

@@ -140,6 +140,7 @@ async function processAskRequest(
 				opencode: { enabled: true },
 				copilot: { enabled: true },
 				ottorouter: { enabled: true },
+				xai: { enabled: true },
 				zai: { enabled: true },
 				'zai-coding': { enabled: true },
 				moonshot: { enabled: true },

package/src/runtime/message/compaction-limits.ts

@@ -14,7 +14,6 @@ export interface ModelLimits {
 
 export function shouldAutoCompactBeforeOverflow(args: {
 	autoCompactThresholdTokens?: number | null;
-	modelContextWindow?: number | null;
 	currentContextTokens?: number | null;
 	estimatedInputTokens?: number | null;
 	isCompactCommand?: boolean;
@@ -31,11 +30,6 @@ export function shouldAutoCompactBeforeOverflow(args: {
 		return false;
 	}
 
-	const modelContextWindow = Number(args.modelContextWindow ?? 0);
-	if (!Number.isFinite(modelContextWindow) || modelContextWindow <= threshold) {
-		return false;
-	}
-
 	const currentContextTokens = Math.max(
 		0,
 		Math.floor(Number(args.currentContextTokens ?? 0)),

package/src/runtime/provider/index.ts

@@ -19,6 +19,7 @@ import {
 	resolveOttoRouterModel,
 	type ResolveOttoRouterModelOptions,
 } from './ottorouter.ts';
+import { getXaiInstance } from './xai.ts';
 import { getZaiInstance, getZaiCodingInstance } from './zai.ts';
 import { resolveOpencodeModel } from './opencode.ts';
 import { getMoonshotInstance } from './moonshot.ts';
@@ -73,6 +74,9 @@ export async function resolveModel(
 			autoPayThresholdUsd: options?.autoPayThresholdUsd,
 		});
 	}
+	if (provider === 'xai') {
+		return getXaiInstance(cfg, model);
+	}
 	if (provider === 'zai') {
 		return getZaiInstance(cfg, model);
 	}

package/src/runtime/provider/reasoning.ts

@@ -190,6 +190,7 @@ function getReasoningProviderTarget(
 	const npmBinding = getModelNpmBinding(provider, model);
 	if (npmBinding === '@ai-sdk/anthropic') return 'anthropic';
 	if (npmBinding === '@ai-sdk/openai') return 'openai';
+	if (npmBinding === '@ai-sdk/xai') return 'openai';
 	if (npmBinding === '@ai-sdk/google') return 'google';
 	if (npmBinding === 'ai-sdk-ollama') return 'ollama';
 	if (npmBinding === '@ai-sdk/openai-compatible') return 'openai-compatible';

package/src/runtime/provider/xai.ts

@@ -0,0 +1,8 @@
+import type { OttoConfig } from '@ottocode/sdk';
+import { getAuth, createXaiModel } from '@ottocode/sdk';
+
+export async function getXaiInstance(cfg: OttoConfig, model: string) {
+	const auth = await getAuth('xai', cfg.projectRoot);
+	const apiKey = auth?.type === 'api' ? auth.key : undefined;
+	return createXaiModel(model, { apiKey });
+}