Skip to content

Allow all recommended model for Vercel routing #417

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
chrarnoldus wants to merge 2 commits into
base: main
Choose a base branch
from
+17 −28
Draft

Allow all recommended model for Vercel routing #417

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
658df6a
Allow all recommended model for Vercel routing
chrarnoldus Feb 20, 2026
1031c1d
Merge branch 'main' into christiaan/vercel
chrarnoldus Feb 20, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 8 additions & 4 deletions src/lib/providers/openrouter/inference-provider-id.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,7 @@
import * as z from 'zod';

export const OpenRouterInferenceProviderIdSchema = z.enum([
'alibaba',
'amazon-bedrock',
'anthropic',
'arcee-ai',
Expand Down Expand Up @@ -46,7 +47,7 @@ export const UserByokProviderIdSchema = VercelUserByokInferenceProviderIdSchema.

export type UserByokProviderId = z.infer<typeof UserByokProviderIdSchema>;

export const VercelNonUserByokInferenceProviderIdSchema = z.enum(['bedrock', 'vertex']);
export const VercelNonUserByokInferenceProviderIdSchema = z.enum(['alibaba', 'bedrock', 'vertex']);

export const VercelInferenceProviderIdSchema = VercelUserByokInferenceProviderIdSchema.or(
VercelNonUserByokInferenceProviderIdSchema
Expand Down Expand Up @@ -81,19 +82,22 @@ const modelPrefixToVercelInferenceProviderMapping = {
openai: VercelUserByokInferenceProviderIdSchema.enum.openai,
minimax: VercelUserByokInferenceProviderIdSchema.enum.minimax,
mistralai: VercelUserByokInferenceProviderIdSchema.enum.mistral,
qwen: VercelNonUserByokInferenceProviderIdSchema.enum.alibaba,
'x-ai': VercelUserByokInferenceProviderIdSchema.enum.xai,
'z-ai': VercelUserByokInferenceProviderIdSchema.enum.zai,
} as Record<string, VercelUserByokInferenceProviderId | undefined>;
} as Record<string, VercelInferenceProviderId | undefined>;

export function inferUserByokProviderForModel(model: string): UserByokProviderId | null {
return model.startsWith('mistralai/codestral')
? AutocompleteUserByokProviderIdSchema.enum.codestral
: inferVercelFirstPartyInferenceProviderForModel(model);
: (VercelUserByokInferenceProviderIdSchema.safeParse(
inferVercelFirstPartyInferenceProviderForModel(model)
).data ?? null);
}

export function inferVercelFirstPartyInferenceProviderForModel(
model: string
): VercelUserByokInferenceProviderId | null {
): VercelInferenceProviderId | null {
return model.startsWith('openai/gpt-oss')
? null
: (modelPrefixToVercelInferenceProviderMapping[model.split('/')[0]] ?? null);
Expand Down
33 changes: 9 additions & 24 deletions src/lib/providers/vercel.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,8 +1,7 @@
import type { BYOKResult } from '@/lib/byok';
import { kiloFreeModels } from '@/lib/models';
import { kiloFreeModels, preferredModels } from '@/lib/models';
import { isAnthropicModel } from '@/lib/providers/anthropic';
import { getGatewayErrorRate } from '@/lib/providers/gateway-error-rate';
import { minimax_m25_free_model } from '@/lib/providers/minimax';
import {
AutocompleteUserByokProviderIdSchema,
inferVercelFirstPartyInferenceProviderForModel,
Expand All @@ -15,7 +14,6 @@ import type {
VercelInferenceProviderConfig,
VercelProviderConfig,
} from '@/lib/providers/openrouter/types';
import { zai_glm5_free_model } from '@/lib/providers/zai';
import * as crypto from 'crypto';

// EMERGENCY SWITCH
Expand All @@ -24,22 +22,6 @@ import * as crypto from 'crypto';
// Only use when OpenRouter is down and automatic failover is not working adequately.
const ENABLE_UNIVERSAL_VERCEL_ROUTING = false;

const VERCEL_ROUTING_ALLOW_LIST = [
'arcee-ai/trinity-large-preview:free',
'google/gemini-3-pro-preview',
'google/gemini-3-flash-preview',
'minimax/minimax-m2.1',
minimax_m25_free_model.public_id,
'minimax/minimax-m2.5',
'openai/gpt-5.2',
'openai/gpt-5.2-codex',
'x-ai/grok-code-fast-1',
'z-ai/glm-4.7',
zai_glm5_free_model.public_id,
'z-ai/glm-5',
// TODO: test and add anthropic, kat-coder, kimi, mistral, qwen models
];

const ERROR_RATE_THRESHOLD = 0.5;

function getRandomNumberLessThan100(randomSeed: string) {
Expand Down Expand Up @@ -77,13 +59,18 @@ export async function shouldRouteToVercel(
return false;
}

if (ENABLE_UNIVERSAL_VERCEL_ROUTING && isOpenRouterModel(requestedModel)) {
if (!isOpenRouterModel(requestedModel)) {
console.debug(`[shouldRouteToVercel] model not available on the gateways`);
return false;
}

if (ENABLE_UNIVERSAL_VERCEL_ROUTING) {
console.debug(`[shouldRouteToVercel] universal Vercel routing is enabled`);
return true;
}

if (!VERCEL_ROUTING_ALLOW_LIST.includes(requestedModel)) {
console.debug(`[shouldRouteToVercel] model not on the allow list for Vercel routing`);
if (!preferredModels.includes(requestedModel)) {
console.debug(`[shouldRouteToVercel] only recommended models are tested for Vercel routing`);
return false;
}

Expand All @@ -108,10 +95,8 @@ function convertProviderOptions(

const vercelModelIdMapping = {
'arcee-ai/trinity-large-preview:free': 'arcee-ai/trinity-large-preview',
'google/gemini-3-flash-preview': 'google/gemini-3-flash',
'mistralai/codestral-2508': 'mistral/codestral',
'mistralai/devstral-2512': 'mistral/devstral-2',
'mistralai/devstral-2512:free': 'mistral/devstral-2',
} as Record<string, string>;

export function applyVercelSettings(
Expand Down