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
fix: classify Zhipu "Prompt exceeds max length" 400 as context_too_large (#873) (#893)
Ben Younes15a73e2e4c2d
diff --git a/server/src/__tests__/lib/exhaustion-statuses.test.ts b/server/src/__tests__/lib/exhaustion-statuses.test.tsindex 8f621927..5dc45820 100644--- a/server/src/__tests__/lib/exhaustion-statuses.test.ts+++ b/server/src/__tests__/lib/exhaustion-statuses.test.ts@@ -96,6 +96,21 @@ describe('isContextTooLargeError', () => {expect(isContextTooLargeError(new Error('request exceeds the maximum number of tokens'))).toBe(true);});+ it('flags Zhipu AI prompt-length 400s (error 1261: "Prompt exceeds max length")', () => {+ // #873: Zhipu returns HTTP 400 with "Prompt exceeds max length" for an+ // over-long context. Its wording matches none of the OpenAI/Anthropic/Google+ // markers, so it was mis-bucketed as provider_bad_request instead of+ // context_too_large.+ expect(isContextTooLargeError(Object.assign(+ new Error('Zhipu API error 400: Prompt exceeds max length. Please shorten the prompt or set a longer context length.'),+ { status: 400 },+ ))).toBe(true);+ expect(classifyAttemptError(Object.assign(+ new Error('Zhipu API error 400: Prompt exceeds max length.'),+ { status: 400 },+ ))).toBe('context_too_large');+ });+it('flags literal HTTP 413s, including Groq TPM-ceiling 413s (Requested > Limit can never fit)', () => {expect(isContextTooLargeError(Object.assign(new Error('Groq API error 413: Request too large for model `llama-3.3-70b-versatile` in organization `org_x` on tokens per minute (TPM): Limit 30000, Requested 33476'),diff --git a/server/src/lib/error-classify.ts b/server/src/lib/error-classify.tsindex 2d2df22a..d1a9cdd6 100644--- a/server/src/lib/error-classify.ts+++ b/server/src/lib/error-classify.ts@@ -374,6 +374,10 @@ export function isContextTooLargeError(err: any): boolean {|| msg.includes('request entity too large')|| msg.includes('request body too large')|| msg.includes('content too large')+ // Zhipu AI (bigmodel.cn) 400, error code 1261: "Prompt exceeds max length"+ // (#873). Its wording matches none of the markers above, so without this it+ // was mis-bucketed as provider_bad_request instead of context_too_large.+ || msg.includes('exceeds max length')|| msg.includes('api error 413');}