openai-completions: no min floor on max_completion_tokens, sends 1 token → 400 Bad Request

[sh1ftred]

What happened?

I'm using OpenAI models via a proxy and the context reported by this proxy is wrong. I've disabled auto compact for Pi so my context used is 400% of what's available haha. Becuase of this, the upstream provider is rejecting my requests with 400 Bad Request:

  Invalid 'max_output_tokens': integer below minimum value. Expected a value >= 16, but got 1 instead.

The Responses API path in Pi's code already floors to 16 (OPENAI_RESPONSES_MIN_OUTPUT_TOKENS = 16), but openai-completions.ts passes the raw clamped (1 here) value through with no floor.

clampMaxTokensToContext in simple-options.ts uses MIN_MAX_TOKENS = 1 as its floor, and openai-completions.ts (lines 570-574) sends that straight to the provider.

Steps to reproduce

1. Have a proxy configure context lower than actually served context. And have your session exceed 100% of the context window.
2. Send another message, context estimation + 4096 safety margin pushes available to ≤ 1.
3. Request fails with 400 from the provider.

Expected behavior

max_completion_tokens should never be sent below the provider's minimum (16). The completions path should floor it the same way the Responses API does.

Version: 80.6

[github-actions]

This issue was auto-closed. All issues from new contributors are auto-closed by default.

Maintainers review auto-closed issues daily and reopen worthwhile ones. Issues that do not meet the quality bar in CONTRIBUTING.md will not be reopened or receive a reply.

If a maintainer replies lgtmi on one of your issues, your future issues will stay open. If a maintainer replies lgtm, your future issues and PRs will stay open.

See CONTRIBUTING.md.

[sh1ftred]

My LLM suggested fixes:

1. Raise MIN_MAX_TOKENS in simple-options.ts from 1 to 16 (or higher) — this protects all providers.
2. Add a floor in openai-completions.ts at line 570–574, similar to the Responses API:

     const MIN_OUTPUT_TOKENS = 16;
     if (options?.maxTokens) {
         const clamped = Math.max(options.maxTokens, MIN_OUTPUT_TOKENS);
         if (compat.maxTokensField === "max_tokens") {
             (params as any).max_tokens = clamped;
         } else {
             params.max_completion_tokens = clamped;
         }
     }