BigPizzaV3/CodexPlusPlus · error · Error

API 输出达到长度上限,本批未提交;请使用输出容量更大的模型

Error message

API 输出达到长度上限,本批未提交;请使用输出容量更大的模型

What it means

completionText() inspects the Chat Completions response; finish_reason==='length' means the model hit its max output tokens, so the generated batch is incomplete. The script refuses to commit a truncated batch and asks the user to pick a model/output budget with larger capacity.

Solutions

  1. Switch to a model with a larger output token limit
  2. Reduce the batch size / number of messages per organize request so output fits
  3. If provider-side max_tokens is configurable, raise it
  4. Re-run the organize after changing the model; the batch was not committed

Example fix

// before
model: 'gpt-3.5-turbo' // 4k output, truncates
// after
model: 'gpt-4o' // larger output capacity
Defensive patterns

Strategy: try-catch

Validate before calling

// check after response: body?.choices?.[0]?.finish_reason !== 'length'

Try / catch

try{ await organizeLong(...) }catch(e){ if(e.message.includes('长度上限')){ switchToLargerOutputModel(); retry(); } }

Prevention

When it happens

Trigger: An organize batch request returns body.choices[0].finish_reason==='length' — the model stopped because max_tokens/output limit was reached before finishing the JSON batch.

Common situations: Large batch of long messages with a small-output model (e.g. 4k output limit); user set a low max_tokens in provider dashboard; model tends to verbosity and truncates the JSON payload.

Related errors


AI-assisted analysis of BigPizzaV3/CodexPlusPlus@b1ed92e5e4 (2026-09-19). Data as JSON: /api/errors/9597956b86b14401. Report an issue: GitHub.

Appendix: source

Thrown at tools/conversation-canvas/public/canvas.user.js:697

      const part=await waitForApi(reader.read(),signal);if(part.done)break;
      size+=part.value.byteLength;if(size>4*1024*1024)throw Object.assign(Error('API 响应超过大小限制'),{canvasApiLocal:true});
      const text=decoder.decode(part.value,{stream:true});
      if(!sse){raw+=text;if(/^\s*(data:|:)/.test(raw)){sse=true;buffer=raw;raw='';}}
      else buffer+=text;
      if(sse){let end;while((end=buffer.indexOf('\n'))>=0){consume(buffer.slice(0,end).replace(/\r$/,''));buffer=buffer.slice(end+1);}}
      progress({bytes:size,chars:content.length});if(done)break;
    }
    const tail=decoder.decode();if(sse){buffer+=tail;if(buffer.trim())consume(buffer.replace(/\r$/,''));
      if(!done&&!finish)throw Object.assign(Error('API 流式连接提前结束,本批未提交'),{canvasApiLocal:true,retryable:true});
      return {choices:[{finish_reason:finish,message:{content}}]};
    }
    try{return JSON.parse(raw+tail);}catch{throw Object.assign(Error('API 返回的不是 JSON,请检查 API 地址'),{canvasApiLocal:true});}
  }finally{void reader.cancel().catch(()=>{});try{reader.releaseLock();}catch{}}
}

function completionText(body){
  const choice=body?.choices?.[0];
  if(choice?.finish_reason==='length')throw Error('API 输出达到长度上限,本批未提交;请使用输出容量更大的模型');
  if(choice?.finish_reason==='content_filter'||choice?.message?.refusal)throw Error('API 未能生成本批整理结果');
  const content=choice?.message?.content;
  const value=typeof content==='string'?content:Array.isArray(content)?content.filter(p=>p?.type==='text').map(p=>p.text||'').join(''):'';
  if(!value.trim())throw Error('API 未返回有效的 choices[0].message.content,请确认兼容 Chat Completions');
  return value;
}

function waitForApi(promise,signal){
  signal?.throwIfAborted();
  if(!signal)return promise;
  return new Promise((resolve,reject)=>{
    const abort=()=>{signal.removeEventListener('abort',abort);reject(signal.reason||new DOMException('Aborted','AbortError'));};
    signal.addEventListener('abort',abort,{once:true});
    promise.then(value=>{signal.removeEventListener('abort',abort);resolve(value);},error=>{signal.removeEventListener('abort',abort);reject(error);});
  });
}

function createApiOrganizer({request,timeoutMs=300000,maxRetries=2,retryDelayMs=2000}){

View on GitHub (pinned to b1ed92e5e4)