Skip to content

7.4 billion tokens per month. 34 free LLM providers. 635 free model endpoints. All behind one /v1 endpoint, plus any custom OpenAI-compatible endpoint. Smart routing, automatic failover, encrypted keys. Personal experimentation only.

TypeScript

chore(catalog): V14 — disable Cerebras qwen-3-235b + llama3.1-8b ahead of 2026-05-27 deprecation (#79)

Tashfeendaf7fdc65bbd
diff --git a/server/src/__tests__/db/idempotency.test.ts b/server/src/__tests__/db/idempotency.test.ts
index bad02673..d2d0b260 100644
--- a/server/src/__tests__/db/idempotency.test.ts
+++ b/server/src/__tests__/db/idempotency.test.ts
@@ -190,6 +190,24 @@ describe('Migration idempotency', () => {
]);
});
+ it('V14: cerebras deprecation disables qwen-3-235b and llama3.1-8b but keeps gpt-oss-120b enabled', () => {
+ process.env.ENCRYPTION_KEY = '0'.repeat(64);
+ const db = initDb(':memory:');
+
+ const rows = db.prepare(`
+ SELECT model_id, enabled FROM models
+ WHERE platform = 'cerebras'
+ AND model_id IN ('qwen-3-235b-a22b-instruct-2507', 'llama3.1-8b', 'gpt-oss-120b')
+ ORDER BY model_id
+ `).all() as { model_id: string; enabled: number }[];
+
+ expect(rows).toEqual([
+ { model_id: 'gpt-oss-120b', enabled: 1 },
+ { model_id: 'llama3.1-8b', enabled: 0 },
+ { model_id: 'qwen-3-235b-a22b-instruct-2507', enabled: 0 },
+ ]);
+ });
+
it('all enabled catalog platforms have a registered provider', async () => {
process.env.ENCRYPTION_KEY = '0'.repeat(64);
const db = initDb(':memory:');
diff --git a/server/src/db/index.ts b/server/src/db/index.ts
index 513e504f..3e8e708f 100644
--- a/server/src/db/index.ts
+++ b/server/src/db/index.ts
@@ -48,6 +48,7 @@ export function initDb(dbPath?: string): Database.Database {
migrateModelsV11(db);
migrateModelsV12(db);
migrateModelsV13(db);
+ migrateModelsV14(db);
ensureUnifiedKey(db);
console.log(`Database initialized at ${resolvedPath}`);
@@ -1207,6 +1208,28 @@ function migrateModelsV13(db: Database.Database) {
apply();
}
+/**
+ * V14 (May 2026): Cerebras hard-deprecation 2026-05-27.
+ *
+ * Per inference-docs.cerebras.ai/models/overview, both
+ * `qwen-3-235b-a22b-instruct-2507` and `llama3.1-8b` hit a hard deprecation
+ * on 2026-05-27 with no announced free-tier replacement at the same
+ * parameter class. Disable both ahead of the cutover so the router stops
+ * sending traffic to them. Row kept (not deleted) so it can be re-enabled
+ * if Cerebras restores or renames either model — same pattern as V9's
+ * disable of `zai-glm-4.7`.
+ *
+ * Cerebras `gpt-oss-120b` is NOT in the deprecation list and stays enabled
+ * as the sole free-tier Cerebras route.
+ */
+function migrateModelsV14(db: Database.Database) {
+ db.prepare(`
+ UPDATE models SET enabled = 0
+ WHERE platform = 'cerebras'
+ AND model_id IN ('qwen-3-235b-a22b-instruct-2507', 'llama3.1-8b')
+ `).run();
+}
+
function ensureUnifiedKey(db: Database.Database) {
const existing = db.prepare("SELECT value FROM settings WHERE key = 'unified_api_key'").get() as { value: string } | undefined;
if (!existing) {