feat: Add content filtering with convenience functions

[KavithaSiva]

Proposed state:

client.chatCompletion({
  deploymentConfiguration: { deploymentId: 'id' },
  llmConfig: {
    model_name: 'gpt-35-turbo-16k',
    model_params: { max_tokens: 50, temperature: 0.1 }
  },
  prompt: {
    template: [{ role: 'user', content: 'What is the capital of Germany?' }]
  },
  filterConfig: {
    input: azureContentFilter({ Hate: 4, SelfHarm: 2 }),
    output: azureContentFilter({ Sexual: 0, Violence: 6 })
  }
});

or you could assign filterConfig property with the content filtering JSON structure the orchestration service accepts:

client.chatCompletion({
  deploymentConfiguration: { deploymentId: 'id' },
  llmConfig: {
    model_name: 'gpt-35-turbo-16k',
    model_params: { max_tokens: 50, temperature: 0.1 }
  },
  prompt: {
    template: [{ role: 'user', content: 'What is the capital of Germany?' }]
  },
  filterConfig: {
    input: {
      filters: [
        {
          type: 'azure_content_safety',
          config: {
            Hate: 4,
            SelfHarm: 4,
            Sexual: 4,
            Violence: 4
          }
        }
      ]
    },
    output: {
      filters: [
        {
          type: 'azure_content_safety',
          config: {
            Hate: 0,
            SelfHarm: 2,
            Sexual: 2,
            Violence: 2
          }
        }
      ]
    }
  }
});

Note: I also re-generated the orchestration client.

[MatKuhr]

LGTM!
Nice that we have the test coverage and also the option for the default filter, where no config is given.
Maybe someone else wants to look more closely at the TS coding details, but semantically it makes sense to me 👍🏻

[marikaner]

I like the outcome. I think it makes sense to allow both variants and the convenient variant is easy to consume. One thing to consider in the future is, if discoverability can be improved when there are multiple services in orchestration.