SocketError: other side closed

[cloudycotton]

Description

I'm facing a problem where if my agent streams for long (we have agent that runs for 100+ steps and write big files), I get error socket connection closed error. It always happens during streaming a big write to a file using a tool.

This seems to happen across providers (anthropic, openai).

AI SDK Version

latest verison of anthropic provider and ai sdk.

Code of Conduct

• I agree to follow this project's Code of Conduct

[cloudycotton]

@gr2m any pointers for me to create a reproducible example?

The only thing I can do to reproduce the problem consistently is to ask it to write 50 page long research.txt file in a loop. That will run into this.

(formatted for readability)

start: {
  "level": 50,
  "time": 1761315667228,
  "pid": 14,
  "hostname": "srv-d2j1d7mmcj7s73eh3kmg-5c9bb5c585-cb5rn",
  "name": "@runable/api",
  "err": {
    "type": "TypeError",
    "message": "terminated: other side closed",
    "stack": "TypeError: terminated ... error stack trace ..."
  },
  "msg": "Got error in agent stream"
}
start: {
  "level": 50,
  "time": 1761315667228,
  "pid": 14,
  "hostname": "srv-d2j1d7mmcj7s73eh3kmg-5c9bb5c585-cb5rn",
  "name": "@runable/api",
  "err": {
    "type": "Error",
    "message": "",
    "stack": "Error ... error stack trace ..."
  },
  "msg": "Got error in agent stream"
}

[cloudycotton]

So I think the reason why it kills the stream and throw error beyond max-retry is because we don't handle the unidici error separately because those error structures contain circular dependencies causing json serialization to fail and not re-try. @gr2m

[cloudycotton]

@lgrammel this is a very frequent issue on openai side. Any time, I use codex model to write a long file. It times out especially in non streaming mode eg - generateText

[gr2m]

It's hard for us to investigate until we have a way to replicate the problem ourselves, unfortunately

[ItayElgazar]

I'm also getting it pretty frequently. For us, it happens as long as there very randomly, when the edits are pretty large.

at AISDKV5LanguageModel.doStream 
cause: SocketError: other side closed
at TLSSocket.<anonymous> (node:internal/deps/undici/undici:6294:28)
 at TLSSocket.emit (node:events:531:35)

[armordog]

I just hit this, in case anyone thought it fixed itself.

[songkeys]

Having the same error as well:

GatewayResponseError: Invalid error response format: Gateway request failed
    at createGatewayErrorFromResponse (//node_modules/.pnpm/@ai-sdk+gateway@3.0.55_zod@4.3.5/node_modules/@ai-sdk/gateway/src/errors/create-gateway-error.ts:47:12)
    at async asGatewayError (/node_modules/.pnpm/@ai-sdk+gateway@3.0.55_zod@4.3.5/node_modules/@ai-sdk/gateway/src/errors/as-gateway-error.ts:55:12)
    at async GatewayLanguageModel.doGenerate (/node_modules/.pnpm/@ai-sdk+gateway@3.0.55_zod@4.3.5/node_modules/@ai-sdk/gateway/src/gateway-language-model.ts:91:13)


[cause]: {
  "__serialized": true,
  "cause": {
    "__serialized": true,
    "code": "UND_ERR_SOCKET",
    "message": "other side closed",
    "name": "Error",
    "stack": "SocketError: other side closed
    at TLSSocket.onHttpSocketEnd (/node_modules/.pnpm/undici@7.24.4/node_modules/undici/lib/dispatcher/client-h1.js:896:22)
    at TLSSocket.emit (node:events:520:35)
    at endReadableNT (node:internal/streams/readable:1701:12)
    at process.processTicksAndRejections (node:internal/process/task_queues:89:21)"
  },
  "message": "Cannot connect to API: other side closed",
  "name": "Error",
  "stack": "AI_APICallError: Cannot connect to API: other side closed
    at handleFetchError (/node_modules/.pnpm/@ai-sdk+provider-utils@4.0.15_zod@4.3.5/node_modules/@ai-sdk/provider-utils/src/handle-fetch-error.ts:51:14)
    at postToApi (/node_modules/.pnpm/@ai-sdk+provider-utils@4.0.15_zod@4.3.5/node_modules/@ai-sdk/provider-utils/src/post-to-api.ts:164:11)
    at async GatewayLanguageModel.doGenerate (/node_modules/.pnpm/@ai-sdk+gateway@3.0.55_zod@4.3.5/node_modules/@ai-sdk/gateway/src/gateway-language-model.ts:66:11)
    at async fn (/node_modules/.pnpm/ai@6.0.99_zod@4.3.5/node_modules/ai/src/generate-text/generate-text.ts:787:34)
    at async <anonymous> (/node_modules/.pnpm/ai@6.0.99_zod@4.3.5/node_modules/ai/src/telemetry/record-span.ts:32:24)
    at async _retryWithExponentialBackoff (/node_modules/.pnpm/ai@6.0.99_zod@4.3.5/node_modules/ai/src/util/retry-with-exponential-backoff.ts:96:12)
    at async fn (/node_modules/.pnpm/ai@6.0.99_zod@4.3.5/node_modules/ai/src/generate-text/generate-text.ts:742:36)
    at async <anonymous> (/node_modules/.pnpm/ai@6.0.99_zod@4.3.5/node_modules/ai/src/telemetry/record-span.ts:32:24)
    at async generateText (/node_modules/.pnpm/ai@6.0.99_zod@4.3.5/node_modules/ai/src/generate-text/generate-text.ts:518:12)
    at async repairInsightOutput (main.ts:53:20)"
}

I'm using vercel ai gateway. When streaming for ~5 minutes, it will throw this error.