Providers

Using Eden AI, you can make a call to one provider or multiple providers at once.

The providers field requires a string as input including the provider name:

• if only one provider, just input the provider name (eg: providers="amazon")
• if requesting multiple providers: enter them by separating them with commas (eg: providers="google,amazon,microsoft"

For some features, you can also choose between different models for each provider: providers="provider/model"(eg: providers="openai/gpt4-o")

Fallback

You can set a list of providers to use as fallback in case the primary provider fails.

The fallback_providers parameter must be comma-separated list, same as the providers parameter.

🚧

• If this parameter is set, providers must contain only one entry

• It doesn't work with asynchronous features

Example

{
  "providers": "openai",
  "fallback_providers": "microsoft, amazon",
  "settings": {
    "openai": "davinci
}

In this example, if openai fails, a call to microsoft will be made, and if microsoft fails as well, a amazon will be called.

Languages

Some features support languages inputs (Translation for example).

If you browse our API Reference on features supporting languages inputs, you will be presented a table listing the supported languages for this feature. On this table you might see values like: en, en-US, auto-detect.

Let's see the difference between theses values:

• auto-detect: self-explanatory, the provider will detect the language from your text/document.
• en(and other 2 char language codes): Language code of your text/document's language.
• en-US: if the value is longer than 2 characters, what comes after the "-" indicate the language's regional code, this is more detailed and might result in a more accurate response, but less providers supports it.

A more Detailed Article on how we standardized languages can be found here.

Formatting Response

Some parameters allow you to customize the response format to your needs:

response_as_dict

Default to true

Allows the response to be returned as a json dict containing providers as keys and their results as values.

Example:

{
  "emvista": {
    "status": "success",
    ...
  },
  "amazon": {
    "status": "success",
    ...
  },
  "openai": {
    "status": "success",
    ...
  }
}

if set to false the response will be a list of provider responses dicts as shown below:

[
  {
    "status": "success",
    "provider": "emvista"
    ...
  },
  {
    "status": "success",
    "provider": "amazon"
    ...
  },
  {
    "status": "success",
    "provider": "openai"
    ...
  }
]

attributes_as_list

Defaults to false

If parameter is set to true and the result contains a list of dict objects like:

{
  "microsoft": {
    "status": "success",
    "items": [
      {
        "keyword": "U.S. senator",
        "importance": null
      },
      {
        "keyword": "first African-American president",
        "importance": null
      },
      {
        "keyword": "Barack Hussein Obama",
        "importance": null
      },
			...
    ]
  }
}

Then the list of dict will be formatted into multiple lists (one per key present in the dict), here is the above response formatted using attributes_as_list:

{
  "microsoft": {
    "status": "success",
     "keyword": ["U.S. senator","first African-American president","Barack Hussein Obama"],
     "importance": [null, null, null]
  }
}

This can ease the response parsing workflow in some programming languages, or if you only need one particular field in the response.

show_original_response

Defaults to false

Whether or not to show the provider original response in addition to the standardized response we provide

if set to true and additional field original_response will be present in the providers result.