@agi-cli/server 0.1.139 → 0.1.140

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@agi-cli/server",
-	"version": "0.1.139",
+	"version": "0.1.140",
 	"description": "HTTP API server for AGI CLI",
 	"type": "module",
 	"main": "./src/index.ts",
@@ -29,8 +29,8 @@
 		"typecheck": "tsc --noEmit"
 	},
 	"dependencies": {
-		"@agi-cli/sdk": "0.1.139",
-		"@agi-cli/database": "0.1.139",
+		"@agi-cli/sdk": "0.1.140",
+		"@agi-cli/database": "0.1.140",
 		"drizzle-orm": "^0.44.5",
 		"hono": "^4.9.9",
 		"zod": "^4.1.8"

package/src/openapi/schemas.ts

@@ -91,8 +91,8 @@ export const schemas = {
 			createdAt: { type: 'integer', format: 'int64' },
 			completedAt: { type: 'integer', format: 'int64', nullable: true },
 			latencyMs: { type: 'integer', nullable: true },
-			promptTokens: { type: 'integer', nullable: true },
-			completionTokens: { type: 'integer', nullable: true },
+			inputTokens: { type: 'integer', nullable: true },
+			outputTokens: { type: 'integer', nullable: true },
 			totalTokens: { type: 'integer', nullable: true },
 			error: { type: 'string', nullable: true },
 		},
@@ -206,7 +206,7 @@ export const schemas = {
 			id: { type: 'string' },
 			label: { type: 'string' },
 			toolCall: { type: 'boolean' },
-			reasoning: { type: 'boolean' },
+			reasoningText: { type: 'boolean' },
 		},
 		required: ['id', 'label'],
 	},

package/src/routes/config/models.ts

@@ -59,7 +59,7 @@ export function registerModelsRoutes(app: Hono) {
 					id: m.id,
 					label: m.label || m.id,
 					toolCall: m.toolCall,
-					reasoning: m.reasoning,
+					reasoningText: m.reasoningText,
 					vision: m.modalities?.input?.includes('image') ?? false,
 				})),
 				default: getDefault(
@@ -97,7 +97,7 @@ export function registerModelsRoutes(app: Hono) {
 						id: string;
 						label: string;
 						toolCall?: boolean;
-						reasoning?: boolean;
+						reasoningText?: boolean;
 					}>;
 				}
 			> = {};
@@ -122,7 +122,7 @@ export function registerModelsRoutes(app: Hono) {
 							id: m.id,
 							label: m.label || m.id,
 							toolCall: m.toolCall,
-							reasoning: m.reasoning,
+							reasoningText: m.reasoningText,
 							vision: m.modalities?.input?.includes('image') ?? false,
 						})),
 					};

package/src/routes/git/commit.ts

@@ -159,7 +159,7 @@ Commit message:`;
 				model,
 				system: systemPrompt,
 				prompt: userPrompt,
-				maxTokens: 500,
+				maxOutputTokens: 500,
 			});
 
 			const message = text.trim();

package/src/routes/git/schemas.ts

@@ -1,4 +1,4 @@
-import { z } from 'zod';
+import { z } from 'zod/v3';
 
 export const gitStatusSchema = z.object({
 	project: z.string().optional(),

package/src/routes/session-messages.ts

@@ -122,7 +122,7 @@ export function registerSessionMessagesRoutes(app: Hono) {
 				typeOf: typeof userContext,
 			});
 
-			const reasoning = body?.reasoning === true;
+			const reasoning = body?.reasoningText === true;
 
 			// Validate model capabilities if tools are allowed for this agent
 			const wantsToolCalls = true; // agent toolset may be non-empty
@@ -156,7 +156,7 @@ export function registerSessionMessagesRoutes(app: Hono) {
 				content,
 				oneShot: Boolean(body?.oneShot),
 				userContext,
-				reasoning,
+				reasoningText: reasoning,
 				images,
 				files,
 			});

package/src/runtime/agent/runner-setup.ts

@@ -1,4 +1,4 @@
-import { loadConfig, catalog } from '@agi-cli/sdk';
+import { loadConfig, getUnderlyingProviderKey } from '@agi-cli/sdk';
 import { getDb } from '@agi-cli/database';
 import { sessions } from '@agi-cli/database/schema';
 import { eq } from 'drizzle-orm';
@@ -40,16 +40,6 @@ export interface SetupResult {
 
 const THINKING_BUDGET = 16000;
 
-function getSolforgeUnderlyingProvider(
-	model: string,
-): 'anthropic' | 'openai' | null {
-	const entry = catalog.solforge?.models?.find((m) => m.id === model);
-	const npm = entry?.provider?.npm;
-	if (npm === '@ai-sdk/anthropic') return 'anthropic';
-	if (npm === '@ai-sdk/openai') return 'openai';
-	return null;
-}
-
 export async function setupRunner(opts: RunOpts): Promise<SetupResult> {
 	const cfgTimer = time('runner:loadConfig+db');
 	const cfg = await loadConfig(opts.projectRoot);
@@ -231,36 +221,31 @@ export async function setupRunner(opts: RunOpts): Promise<SetupResult> {
 	const providerOptions: Record<string, unknown> = {};
 	let effectiveMaxOutputTokens = maxOutputTokens;
 
-	if (opts.reasoning) {
-		if (opts.provider === 'anthropic') {
+	if (opts.reasoningText) {
+		const underlyingProvider = getUnderlyingProviderKey(
+			opts.provider,
+			opts.model,
+		);
+
+		if (underlyingProvider === 'anthropic') {
 			providerOptions.anthropic = {
 				thinking: { type: 'enabled', budgetTokens: THINKING_BUDGET },
 			};
 			if (maxOutputTokens && maxOutputTokens > THINKING_BUDGET) {
 				effectiveMaxOutputTokens = maxOutputTokens - THINKING_BUDGET;
 			}
-		} else if (opts.provider === 'openai') {
+		} else if (underlyingProvider === 'openai') {
 			providerOptions.openai = {
 				reasoningSummary: 'auto',
 			};
-		} else if (opts.provider === 'google') {
+		} else if (underlyingProvider === 'google') {
 			providerOptions.google = {
 				thinkingConfig: { thinkingBudget: THINKING_BUDGET },
 			};
-		} else if (opts.provider === 'solforge') {
-			const underlying = getSolforgeUnderlyingProvider(opts.model);
-			if (underlying === 'anthropic') {
-				providerOptions.anthropic = {
-					thinking: { type: 'enabled', budgetTokens: THINKING_BUDGET },
-				};
-				if (maxOutputTokens && maxOutputTokens > THINKING_BUDGET) {
-					effectiveMaxOutputTokens = maxOutputTokens - THINKING_BUDGET;
-				}
-			} else if (underlying === 'openai') {
-				providerOptions.openai = {
-					reasoningSummary: 'auto',
-				};
-			}
+		} else if (underlyingProvider === 'openai-compatible') {
+			providerOptions['openai-compatible'] = {
+				reasoningEffort: 'high',
+			};
 		}
 	}
 

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

@@ -9,7 +9,7 @@ export interface TokenUsage {
 	output: number;
 	cacheRead?: number;
 	cacheWrite?: number;
-	reasoning?: number;
+	reasoningText?: number;
 }
 
 export interface ModelLimits {
@@ -17,7 +17,10 @@ export interface ModelLimits {
 	output: number;
 }
 
-export function isOverflow(tokens: TokenUsage, limits: ModelLimits): boolean {
+export function isOverflow(
+	tokens: LanguageModelUsage,
+	limits: ModelLimits,
+): boolean {
 	if (limits.context === 0) return false;
 
 	const count =

package/src/runtime/message/history-builder.ts

@@ -211,7 +211,7 @@ export async function buildHistoryMessages(
 		}
 	}
 
-	return convertToModelMessages(ui);
+	return await convertToModelMessages(ui);
 }
 
 async function logPendingToolParts(

package/src/runtime/message/service.ts

@@ -23,7 +23,7 @@ type DispatchOptions = {
 	content: string;
 	oneShot?: boolean;
 	userContext?: string;
-	reasoning?: boolean;
+	reasoningText?: boolean;
 	images?: Array<{ data: string; mediaType: string }>;
 	files?: Array<{
 		type: 'image' | 'pdf' | 'text';
@@ -47,7 +47,7 @@ export async function dispatchAssistantMessage(
 		content,
 		oneShot,
 		userContext,
-		reasoning,
+		reasoningText,
 		images,
 		files,
 	} = options;
@@ -181,7 +181,7 @@ export async function dispatchAssistantMessage(
 			projectRoot: cfg.projectRoot,
 			oneShot: Boolean(oneShot),
 			userContext,
-			reasoning,
+			reasoningText,
 			isCompactCommand: isCompact,
 			compactionContext,
 		},

package/src/runtime/session/branch.ts

@@ -115,8 +115,8 @@ export async function createBranch({
 			createdAt: msg.createdAt,
 			completedAt: msg.completedAt,
 			latencyMs: msg.latencyMs,
-			promptTokens: msg.promptTokens,
-			completionTokens: msg.completionTokens,
+			inputTokens: msg.inputTokens,
+			outputTokens: msg.outputTokens,
 			totalTokens: msg.totalTokens,
 			cachedInputTokens: msg.cachedInputTokens,
 			cacheCreationInputTokens: msg.cacheCreationInputTokens,

package/src/runtime/session/db-operations.ts

@@ -27,7 +27,7 @@ export type ProviderMetadata = Record<string, unknown> & {
 
 export function normalizeUsage(
 	usage: UsageData,
-	providerMetadata: ProviderMetadata | undefined,
+	providerOptions: ProviderMetadata | undefined,
 	provider: ProviderId,
 ): UsageData {
 	const rawInputTokens = Number(usage.inputTokens ?? 0);
@@ -37,17 +37,17 @@ export function normalizeUsage(
 	const cachedInputTokens =
 		usage.cachedInputTokens != null
 			? Number(usage.cachedInputTokens)
-			: providerMetadata?.openai?.cachedPromptTokens != null
-				? Number(providerMetadata.openai.cachedPromptTokens)
-				: providerMetadata?.anthropic?.cacheReadInputTokens != null
-					? Number(providerMetadata.anthropic.cacheReadInputTokens)
+			: providerOptions?.openai?.cachedPromptTokens != null
+				? Number(providerOptions.openai.cachedPromptTokens)
+				: providerOptions?.anthropic?.cacheReadInputTokens != null
+					? Number(providerOptions.anthropic.cacheReadInputTokens)
 					: undefined;
 
 	const cacheCreationInputTokens =
 		usage.cacheCreationInputTokens != null
 			? Number(usage.cacheCreationInputTokens)
-			: providerMetadata?.anthropic?.cacheCreationInputTokens != null
-				? Number(providerMetadata.anthropic.cacheCreationInputTokens)
+			: providerOptions?.anthropic?.cacheCreationInputTokens != null
+				? Number(providerOptions.anthropic.cacheCreationInputTokens)
 				: undefined;
 
 	const cachedValue = cachedInputTokens ?? 0;
@@ -99,18 +99,14 @@ export function resolveUsageProvider(
  */
 export async function updateSessionTokensIncremental(
 	usage: UsageData,
-	providerMetadata: ProviderMetadata | undefined,
+	providerOptions: ProviderMetadata | undefined,
 	opts: RunOpts,
 	db: Awaited<ReturnType<typeof getDb>>,
 ) {
 	if (!usage || !db) return;
 
 	const usageProvider = resolveUsageProvider(opts.provider, opts.model);
-	const normalizedUsage = normalizeUsage(
-		usage,
-		providerMetadata,
-		usageProvider,
-	);
+	const normalizedUsage = normalizeUsage(usage, providerOptions, usageProvider);
 
 	// Read session totals
 	const sessRows = await db
@@ -134,8 +130,8 @@ export async function updateSessionTokensIncremental(
 		.where(eq(messages.id, opts.assistantMessageId));
 
 	const msg = msgRows[0];
-	const priorPromptMsg = Number(msg?.promptTokens ?? 0);
-	const priorCompletionMsg = Number(msg?.completionTokens ?? 0);
+	const priorPromptMsg = Number(msg?.inputTokens ?? 0);
+	const priorCompletionMsg = Number(msg?.outputTokens ?? 0);
 	const priorCachedMsg = Number(msg?.cachedInputTokens ?? 0);
 	const priorCacheCreationMsg = Number(msg?.cacheCreationInputTokens ?? 0);
 	const priorReasoningMsg = Number(msg?.reasoningTokens ?? 0);
@@ -231,18 +227,14 @@ export async function updateSessionTokens(
  */
 export async function updateMessageTokensIncremental(
 	usage: UsageData,
-	providerMetadata: ProviderMetadata | undefined,
+	providerOptions: ProviderMetadata | undefined,
 	opts: RunOpts,
 	db: Awaited<ReturnType<typeof getDb>>,
 ) {
 	if (!usage || !db) return;
 
 	const usageProvider = resolveUsageProvider(opts.provider, opts.model);
-	const normalizedUsage = normalizeUsage(
-		usage,
-		providerMetadata,
-		usageProvider,
-	);
+	const normalizedUsage = normalizeUsage(usage, providerOptions, usageProvider);
 
 	const msgRows = await db
 		.select()
@@ -251,8 +243,8 @@ export async function updateMessageTokensIncremental(
 
 	if (msgRows.length > 0 && msgRows[0]) {
 		const msg = msgRows[0];
-		const priorPrompt = Number(msg.promptTokens ?? 0);
-		const priorCompletion = Number(msg.completionTokens ?? 0);
+		const priorPrompt = Number(msg.inputTokens ?? 0);
+		const priorCompletion = Number(msg.outputTokens ?? 0);
 		const priorCached = Number(msg.cachedInputTokens ?? 0);
 		const priorCacheCreation = Number(msg.cacheCreationInputTokens ?? 0);
 		const priorReasoning = Number(msg.reasoningTokens ?? 0);
@@ -287,8 +279,8 @@ export async function updateMessageTokensIncremental(
 		await db
 			.update(messages)
 			.set({
-				promptTokens: cumPrompt,
-				completionTokens: cumCompletion,
+				inputTokens: cumPrompt,
+				outputTokens: cumCompletion,
 				totalTokens: cumTotal,
 				cachedInputTokens: cumCached,
 				cacheCreationInputTokens: cumCacheCreation,

package/src/runtime/session/queue.ts

@@ -10,7 +10,7 @@ export type RunOpts = {
 	projectRoot: string;
 	oneShot?: boolean;
 	userContext?: string;
-	reasoning?: boolean;
+	reasoningText?: boolean;
 	abortSignal?: AbortSignal;
 	isCompactCommand?: boolean;
 	compactionContext?: string;

package/src/runtime/stream/finish-handler.ts

@@ -73,8 +73,8 @@ export function createFinishHandler(
 
 		const usage = sessRows[0]
 			? {
-					inputTokens: Number(sessRows[0].promptTokens ?? 0),
-					outputTokens: Number(sessRows[0].completionTokens ?? 0),
+					inputTokens: Number(sessRows[0].inputTokens ?? 0),
+					outputTokens: Number(sessRows[0].outputTokens ?? 0),
 					totalTokens: Number(sessRows[0].totalTokens ?? 0),
 					cachedInputTokens: Number(sessRows[0].cachedInputTokens ?? 0),
 					cacheCreationInputTokens: Number(
@@ -97,7 +97,7 @@ export function createFinishHandler(
 			try {
 				const limits = getModelLimits(opts.provider, opts.model);
 				if (limits) {
-					const tokenUsage: TokenUsage = {
+					const tokenUsage: LanguageModelUsage = {
 						input: usage.inputTokens ?? 0,
 						output: usage.outputTokens ?? 0,
 						cacheRead:

package/src/runtime/stream/step-finish.ts

@@ -18,13 +18,13 @@ export function createStepFinishHandler(
 	sharedCtx: ToolAdapterContext,
 	updateSessionTokensIncrementalFn: (
 		usage: UsageData,
-		providerMetadata: ProviderMetadata | undefined,
+		providerOptions: ProviderMetadata | undefined,
 		opts: RunOpts,
 		db: Awaited<ReturnType<typeof getDb>>,
 	) => Promise<void>,
 	updateMessageTokensIncrementalFn: (
 		usage: UsageData,
-		providerMetadata: ProviderMetadata | undefined,
+		providerOptions: ProviderMetadata | undefined,
 		opts: RunOpts,
 		db: Awaited<ReturnType<typeof getDb>>,
 	) => Promise<void>,

package/src/tools/database/get-parent-session.ts

@@ -1,5 +1,5 @@
 import { tool } from 'ai';
-import { z } from 'zod';
+import { z } from 'zod/v3';
 import { getDb } from '@agi-cli/database';
 import { sessions, messages, messageParts } from '@agi-cli/database/schema';
 import { eq, asc, count } from 'drizzle-orm';

package/src/tools/database/get-session-context.ts

@@ -1,5 +1,5 @@
 import { tool } from 'ai';
-import { z } from 'zod';
+import { z } from 'zod/v3';
 import { getDb } from '@agi-cli/database';
 import { sessions, messages, messageParts } from '@agi-cli/database/schema';
 import { eq, asc, count } from 'drizzle-orm';

package/src/tools/database/present-session-links.ts

@@ -1,5 +1,5 @@
 import { tool } from 'ai';
-import { z } from 'zod';
+import { z } from 'zod/v3';
 
 const sessionLinkSchema = z.object({
 	sessionId: z.string().describe('The session ID to link to'),

package/src/tools/database/query-messages.ts

@@ -1,5 +1,5 @@
 import { tool } from 'ai';
-import { z } from 'zod';
+import { z } from 'zod/v3';
 import { getDb } from '@agi-cli/database';
 import { sessions, messages, messageParts } from '@agi-cli/database/schema';
 import { eq, desc, asc, gte, lte, and, like, count, sql } from 'drizzle-orm';

package/src/tools/database/query-sessions.ts

@@ -1,5 +1,5 @@
 import { tool } from 'ai';
-import { z } from 'zod';
+import { z } from 'zod/v3';
 import { getDb } from '@agi-cli/database';
 import { sessions, messages } from '@agi-cli/database/schema';
 import { eq, desc, asc, gte, lte, and, sql, count } from 'drizzle-orm';

package/src/tools/database/search-history.ts

@@ -1,5 +1,5 @@
 import { tool } from 'ai';
-import { z } from 'zod';
+import { z } from 'zod/v3';
 import { getDb } from '@agi-cli/database';
 import { sessions, messages, messageParts } from '@agi-cli/database/schema';
 import { eq, desc, asc, like, and, sql } from 'drizzle-orm';