GPT OSS (not working)

convex/action/lifecycle/private.ts

@@ -241,6 +241,8 @@ export const _resolve = internalMutation({
 				skills: result.reactions.map((reaction) => ({
 					skillKey: reaction.skillKey,
 					args: reaction.args,
+					status: reaction.status,
+					result: reaction.result,
 				})),
 			});
 		}

convex/action/private.ts

@@ -22,7 +22,7 @@ export const _add = internalMutation({
 			owner,
 			depth,
 			shouldReopen,
-			skills: [{ skillKey, args }],
+			skills: [{ skillKey, args, status: 'enqueued' }],
 		});
 
 		return actionIds[0];
@@ -39,6 +39,8 @@ export const _addMany = internalMutation({
 			z.object({
 				skillKey: z.string().describe('The key of the skill to use'),
 				args: z.record(z.any()),
+				status: z.enum(['enqueued', 'succeeded']),
+				result: z.string().optional(),
 			}),
 		),
 	},
@@ -53,22 +55,41 @@ export const _addMany = internalMutation({
 
 		// reopen if needed and requested
 		if (!task.isActive && shouldReopen) {
-			skills.unshift({ skillKey: 'reopen', args: {} });
+			skills.unshift({ skillKey: 'reopen', args: {}, status: 'enqueued' });
 		}
 
 		const actionIds = await Promise.all(
-			skills.map((skill) =>
-				ctx.db.insert('actions', {
+			skills.map((skill) => {
+				if (skill.status === 'enqueued') {
+					return ctx.db.insert('actions', {
+						taskId,
+						author,
+						owner,
+						depth,
+						status: 'enqueued',
+						result: null,
+						skillKey: skill.skillKey,
+						args: skill.args,
+					});
+				}
+
+				if (!skill.result) throw new Error('Skill result is required for succeeded actions.');
+
+				return ctx.db.insert('actions', {
 					taskId,
 					author,
 					owner,
 					depth,
-					status: 'enqueued',
-					result: null,
+					status: skill.status,
+					result: {
+						text: skill.result,
+						reactions: [],
+					},
 					skillKey: skill.skillKey,
 					args: skill.args,
-				}),
-			),
+					costs: [],
+				});
+			}),
 		);
 
 		await _runNextActionIfNeeded(ctx, taskId);

convex/action/public.ts

@@ -26,6 +26,7 @@ export const act = mutation({
 			author: currentUser._id,
 			owner: currentUser._id,
 			shouldReopen,
+			status: 'enqueued',
 		});
 	},
 });

convex/magicRock.tsx

@@ -124,8 +124,33 @@ export async function _askMagicRock(args: MagicRockContext) {
 		usage,
 		warnings,
 		providerMetadata,
+		reasoning,
+		reasoningDetails,
+		sources,
+		files,
 		//
-	} = await generateText(args);
+	} = await generateText({
+		...args,
+		providerOptions: {
+			openai: {
+				reasoningEffort: 'high',
+				reasoningSummary: 'detailed',
+			},
+			openrouter: {
+				reasoning: { effort: 'high' },
+				provider: {
+					// only: ['baseten'], // "Not Found", no tools support
+					// only: ['novitaai'], // "Not Found", no tools support
+					// only: ['together'], // broken, possibly related to Harmony.
+					// only: ['fireworks'], // sort of works, but no nice. Possibly related to Harmony.
+					// only: ['parasail'], // "Not Found", no tools support
+					only: ['groq'], // the best yet, but not reliable.
+					allow_fallbacks: false,
+					require_parameters: true,
+				},
+			},
+		},
+	});
 
 	const result = {
 		finishReason,
@@ -134,6 +159,10 @@ export async function _askMagicRock(args: MagicRockContext) {
 		usage,
 		warnings,
 		providerMetadata,
+		reasoning,
+		reasoningDetails,
+		sources,
+		files,
 	};
 
 	console.debug('askMagicRock', result);
@@ -174,6 +203,9 @@ function languageModelFrom(
 		// OpenAI
 		// 'openai/gpt-4o': openai('gpt-4o', openAIconfig),
 		// 'openai/gpt-4o-mini': openai('gpt-4o-mini', openAIconfig),
+		'openai/gpt-5': openai('gpt-5', openAIconfig),
+		'openai/gpt-5-mini': openai('gpt-5-mini', openAIconfig),
+		'openai/gpt-5-nano': openai('gpt-5-nano', openAIconfig),
 		'openai/gpt-4.1': openai('gpt-4.1', openAIconfig),
 		'openai/gpt-4.1-mini': openai('gpt-4.1-mini', openAIconfig),
 		'openai/gpt-4.1-nano': openai('gpt-4.1-nano', openAIconfig),
@@ -193,6 +225,12 @@ function languageModelFrom(
 		// 'groq/llama-4-scout': groq('meta-llama/llama-4-scout-17b-16e-instruct'),
 		// 'groq/llama-4-maverick': groq('meta-llama/llama-4-maverick-17b-128e-instruct'),
 		'groq/qwen3-32b': groq('qwen/qwen3-32b'),
+		// 'openai/gpt-oss-120b': openrouter('openai/gpt-oss-120b', {
+		// 	// reasoning: { enabled: false, effort: 'high' },
+		// }),
+		// 'openai/gpt-oss-20b': openrouter('openai/gpt-oss-20b', {
+		// 	// reasoning: { enabled: false, effort: 'high' },
+		// }),
 
 		// DeepSeek
 		'deepseek/deepseek-v3': deepseek('deepseek-chat'),

convex/schedules/lifecycle.ts

@@ -28,6 +28,7 @@ export const _executeOneTime = internalMutation({
 			skillKey,
 			args,
 			depth,
+			status: 'enqueued',
 		});
 
 		await ctx.db.delete(scheduleId);
@@ -55,6 +56,7 @@ export const _executeRecurring = internalMutation({
 			skillKey,
 			args,
 			depth,
+			status: 'enqueued',
 		});
 
 		const nextRunAt = computeNextRun(cronExpression, timeZone);

convex/schemas/actionSchema.tsx

@@ -10,7 +10,7 @@ export const newActionSchema = z.object({
 	skillKey: z.string().describe('The key of the skill to use'),
 	args: z.record(z.any()),
 	depth: z.number().min(0).max(1000),
-	status: z.enum(['enqueued', 'succeeded']).default('enqueued').optional(),
+	status: z.enum(['enqueued', 'succeeded']).default('enqueued'),
 	result: z.string().optional(),
 });
 

convex/schemas/skillSchema.tsx

@@ -68,6 +68,9 @@ export const modelsSchema = z.enum([
 	'openai/gpt-4.1',
 	'openai/gpt-4.1-mini',
 	'openai/gpt-4.1-nano',
+	'openai/gpt-5',
+	'openai/gpt-5-mini',
+	'openai/gpt-5-nano',
 	'openai/gpt-oss-120b',
 	'openai/gpt-oss-20b',
 
@@ -84,6 +87,8 @@ export const modelsSchema = z.enum([
 	// 'groq/llama-4-scout',
 	// 'groq/llama-4-maverick',
 	'groq/qwen3-32b',
+	'openai/gpt-oss-120b',
+	// 'openai/gpt-oss-20b',
 
 	// DeepSeek
 	'deepseek/deepseek-v3',
@@ -140,6 +145,12 @@ export function pricingFor(model: z.infer<typeof modelsSchema>): {
 			return pricePerMillionTokens({ input: 0.4, output: 1.6 });
 		case 'openai/gpt-4.1-nano':
 			return pricePerMillionTokens({ input: 0.1, output: 0.4 });
+		case 'openai/gpt-5':
+			return pricePerMillionTokens({ input: 1.25, output: 10 });
+		case 'openai/gpt-5-mini':
+			return pricePerMillionTokens({ input: 0.25, output: 2 });
+		case 'openai/gpt-5-nano':
+			return pricePerMillionTokens({ input: 0.05, output: 0.4 });
 		case 'openai/gpt-oss-120b':
 			return pricePerMillionTokens({ input: 0.25, output: 0.75 });
 		case 'openai/gpt-oss-20b':
@@ -166,6 +177,8 @@ export function pricingFor(model: z.infer<typeof modelsSchema>): {
 		// 	return pricePerMillionTokens({ input: 0.2, output: 0.6 });
 		case 'groq/qwen3-32b':
 			return pricePerMillionTokens({ input: 0.29, output: 0.59 });
+		// case 'openai/gpt-oss-20b':
+		// 	return pricePerMillionTokens({ input: 0.1, output: 0.5 });
 
 		// DeepSeek
 		case 'deepseek/deepseek-v3':

convex/skills/createAITool.ts

@@ -40,6 +40,10 @@ export function createAITool(
 				usage,
 				warnings,
 				providerMetadata,
+				reasoning,
+				reasoningDetails,
+				sources,
+				files,
 				//
 			} = await _askMagicRock(context);
 
@@ -55,6 +59,7 @@ export function createAITool(
 				author: action._id,
 				owner: task.owner,
 				depth: action.depth + 1,
+				status: 'enqueued',
 			});
 
 			let reason = finishReason;
@@ -65,6 +70,19 @@ export function createAITool(
 				);
 			}
 
+			if (reasoning) {
+				reactions.push({
+					skillKey: 'reason',
+					args: { reasoning },
+					taskId: task._id,
+					author: action._id,
+					owner: task.owner,
+					depth: action.depth + 1,
+					status: 'succeeded',
+					result: reasoning,
+				});
+			}
+
 			switch (reason) {
 				//
 				case 'tool-calls':
@@ -94,14 +112,15 @@ export function createAITool(
 						author: action._id,
 						owner: task.owner,
 						depth: action.depth + 1,
+						status: 'enqueued',
 					});
 					break;
 
 				// prettier-ignore
 				case 'stop': say(text); break;
 
 				// prettier-ignore
-				case 'error': say(text); break;
+				case 'error': say(`Error: ${text}`); break;
 
 				// prettier-ignore
 				case 'content-filter': say(`[damn @sama] Content filter hit: ${warnings}`); break;

convex/skills/createReactions.ts

@@ -1,4 +1,6 @@
+import { z } from 'zod';
 import { Doc, Id } from '../_generated/dataModel';
+import { newActionSchema } from '../schemas/actionSchema';
 
 export function createReactions(
 	action: Doc<'actions'>,
@@ -7,7 +9,8 @@ export function createReactions(
 		args: Record<string, any>;
 		condition?: 'owner' | 'companion' | 'any';
 	}>,
-) {
+): Array<z.infer<typeof newActionSchema>> {
+	//
 	return (reactions ?? [])
 		.filter((reaction) => {
 			// prettier-ignore
@@ -24,5 +27,6 @@ export function createReactions(
 			owner: action.owner,
 			depth: action.depth + 1,
 			author: action._id as Id<'actions'> | Id<'users'>, // I have no idea why I need that cast, as it expects a union of Id<'actions'> or Id<'users'>
+			status: 'enqueued' as const,
 		}));
 }

convex/skills/public.ts

@@ -107,6 +107,42 @@ export const availableIntelligences = query({
 		return {
 			default: env.DEFAULT_MODEL,
 			recommended: [
+				{
+					key: 'openai/gpt-5',
+					name: 'GPT-5',
+					provider: 'OpenAI',
+					description: 'Best performance per energy',
+				},
+				{
+					key: 'openai/gpt-5-mini',
+					name: 'GPT-5 Mini',
+					provider: 'OpenAI',
+					description: 'Best performance per energy',
+				},
+				{
+					key: 'openai/gpt-5-nano',
+					name: 'GPT-5 Nano',
+					provider: 'OpenAI',
+					description: 'Best performance per energy',
+				},
+				{
+					key: 'openai/gpt-oss-120b',
+					name: 'GPT-OSS 120B',
+					provider: 'OpenAI',
+					description: 'Best performance per energy',
+				},
+				// {
+				// 	key: 'openai/gpt-oss-20b',
+				// 	name: 'GPT-OSS 20B',
+				// 	provider: 'OpenAI',
+				// 	description: 'Best performance per energy',
+				// },
+				{
+					key: 'anthropic/claude-4-sonnet',
+					name: 'Claude 4 Sonnet',
+					provider: 'Anthropic',
+					description: 'Best overall',
+				},
 				{
 					key: 'moonshot/kimi-2',
 					name: 'Kimi 2',
@@ -151,6 +187,26 @@ export const availableIntelligences = query({
 					provider: 'Groq',
 					description: 'Insanely faaaast, but not very smart',
 				},
+				{
+					key: 'openrouter/qwen-3-coder',
+					name: 'Qwen 3 Coder',
+					provider: 'OpenRouter',
+				},
+				{
+					key: 'openrouter/GLM-4.5-Air',
+					name: 'GLM 4.5 Air',
+					provider: 'OpenRouter',
+				},
+				{
+					key: 'openrouter/GLM-4.5',
+					name: 'GLM 4.5',
+					provider: 'OpenRouter',
+				},
+				{
+					key: 'deepinfra/glm-4.5',
+					name: 'GLM 4.5',
+					provider: 'DeepInfra',
+				},
 				{
 					key: 'cerebras/qwen3-235b',
 					name: 'Qwen 235B',

convex/tasks/private.ts

@@ -131,12 +131,14 @@ export const _add = internalMutation({
 						{
 							skillKey: 'increaseBudget',
 							args: { amount: initialFunds, shouldIterate: false },
+							status: 'enqueued' as const,
 						},
 					]
 				: []),
 			{
 				skillKey: 'say',
 				args: { message },
+				status: 'enqueued' as const,
 			},
 		];
 
@@ -164,6 +166,8 @@ export const _addWithActions = internalMutation({
 			z.object({
 				skillKey: z.string().describe('The key of the skill to use'),
 				args: z.record(z.any()),
+				status: z.enum(['enqueued', 'succeeded']),
+				result: z.string().optional(),
 			}),
 		),
 	},