farooquiowais commited on
Commit
2be6e7d
·
1 Parent(s): 64a3efa

feat: raise default max_tokens fallback from 4096 to 16384

Browse files

OpenAI-compatible (/v1/chat/completions) and Anthropic-compatible
(/v1/messages) endpoints now default to 16384 output tokens when the
client omits max_tokens. This matches the existing Gemini hard cap and
mitigates truncation for streaming clients (e.g. Mike tabular batch
extraction) until they send explicit per-request budgets.

Files changed (2) hide show
  1. src/format/openai-compat.js +1 -1
  2. src/server.js +1 -1
src/format/openai-compat.js CHANGED
@@ -390,7 +390,7 @@ export function convertOpenAIRequestToAnthropic(openAIRequest, configuredMapping
390
  const model = resolveModelAlias(openAIRequest.model, configuredMappings);
391
  const maxTokens = openAIRequest.max_completion_tokens
392
  ?? openAIRequest.max_tokens
393
- ?? 4096;
394
 
395
  if (!Number.isInteger(maxTokens) || maxTokens <= 0) {
396
  throw new OpenAICompatError('max_tokens must be a positive integer', {
 
390
  const model = resolveModelAlias(openAIRequest.model, configuredMappings);
391
  const maxTokens = openAIRequest.max_completion_tokens
392
  ?? openAIRequest.max_tokens
393
+ ?? 16384;
394
 
395
  if (!Number.isInteger(maxTokens) || maxTokens <= 0) {
396
  throw new OpenAICompatError('max_tokens must be a positive integer', {
src/server.js CHANGED
@@ -773,7 +773,7 @@ app.post('/v1/messages', async (req, res) => {
773
  const request = {
774
  model: modelId,
775
  messages,
776
- max_tokens: max_tokens || 4096,
777
  stream,
778
  system,
779
  tools,
 
773
  const request = {
774
  model: modelId,
775
  messages,
776
+ max_tokens: max_tokens || 16384,
777
  stream,
778
  system,
779
  tools,