ehtick / teams-ai / 17110154552 / 2

80%

main: 80%

push

github

[JS] feat: #2560 Add max_completion_tokens handling for gpt-5 (#2561)

OpenAi released gpt-5 in Aug 7, 2025. gpt-5 has deprecated the
max_tokens parameter and old variable temperatures, having an api
similar to the o* models. This change fixes gpt-5 usage and allows users
to choose cheaper/better gpt-5 models as their base model.

Since gpt-5 models allows for tools usage and system messages but do not
allow for temperature, top_p and presence_penalty parameters, a new flag
has been added to toggle on just the parameters handling. This is going
to fix the bad request error that happens when trying to use gpt-5
models using teams-ai

## Linked issues

closes: #2560

## Details

Added a new isGpt5Model flag to toggle off deprecated api parameters,
but keeping tool usage as system prompts.

#### Change details

Create a new flag variable named isGpt5Model to toggle off new
max_completion_tokens and other parameters when detected (just like
reasoning models)

**code snippets**:


```typescript       
const isGpt5Model = model.startsWith('gpt-5')
// --- ... --  
if (isThinkingModel || isGpt5Model) {
          if (params.max_tokens) {
              params.max_completion_tokens = params.max_tokens;
              delete params.max_tokens;
          }
          params.temperature = 1;
          params.top_p = 1;
          params.presence_penalty = 0;
      }
          
```

## Attestation Checklist

- [X] My code follows the style guidelines of this project

- I have checked for/fixed spelling, linting, and other errors
- I have commented my code for clarity
- I have made corresponding changes to the documentation (updating the
doc strings in the code is sufficient)
- My changes generate no new warnings
- I have added tests that validates my changes, and provides sufficient
test coverage. I have tested with:
  - Local testing
  - E2E testing in Teams
- New and existing unit tests pass locally with my changes

Co-authored-by: Lily Du <lilydu@microsoft.com>

1700 of 2618 branches covered (64.94%)

Branch coverage included in aggregate %.

4818 of 5994 relevant lines covered (80.38%)

27.27 hits per line