Chat Stream

post

https://api.edenai.run/v2/text/chat/stream/

Streamed version of Chat feature, the raw text will be streamed chunk by chunk.

NOTE: For this feature, you an only request one provider at a time.

Recent Requests

Time	Status	User Agent
Retrieving recent requests…

Loading…

Body Params

settings

string

Defaults to [object Object]

A dictionnary or a json object to specify specific models to use for some providers.
It can be in the following format: {"google" : "google_model", "ibm": "ibm_model"...}.

providers

array of strings

required

It can be one (ex: 'amazon' or 'google') or multiple provider(s) (ex: 'amazon,microsoft,google') that the data will be redirected to in order to get the processed results.
Providers can also be invoked with specific models (ex: providers: 'amazon/model1, amazon/model2, google/model3')

providers*

fallback_providers

array of strings

length ≤ 5

Defaults to

Providers in this list will be used as fallback if the call to provider in providers parameter fails. To use this feature, you must input only one provider in the providers parameter. but you can put up to 5 fallbacks.

They will be tried in the same order they are input, and it will stop to the first provider who doesn't fail.

Doesn't work with async subfeatures.

fallback_providers

response_as_dict

boolean

Defaults to true

Optional : When set to true (default), the response is an object of responses with providers names as keys :
{"google" : { "status": "success", ... }, }
When set to false the response structure is a list of response objects :
[{"status": "success", "provider": "google" ... }, ].

attributes_as_list

boolean

Defaults to false

Optional : When set to false (default) the structure of the extracted items is list of objects having different attributes :
{'items': [{"attribute_1": "x1","attribute_2": "y2"}, ... ]}
When it is set to true, the response contains an object with each attribute as a list :
{ "attribute_1": ["x1","x2", ...], "attribute_2": [y1, y2, ...]}

show_base_64

boolean

Defaults to true

show_original_response

boolean

Defaults to false

Optional : Shows the original response of the provider.
When set to true, a new attribute original_response will appear in the response object.

text

string | null

Start your conversation here...

chatbot_global_action

string | null

A system message that helps set the behavior of the assistant. For example, 'You are a helpful assistant'.

previous_history

array of objects

A list containing all the previous conversations between the user and the chatbot AI. Each item in the list should be a dictionary with two keys: 'role' and 'message'. The 'role' key specifies the role of the speaker and can have the values 'user' or 'assistant'. The 'message' key contains the text of the conversation from the respective role. For example: [{'role': 'user', 'message': 'Hello'}, {'role': 'assistant', 'message': 'Hi, how can I help you?'}, ...]. This format allows easy identification of the speaker's role and their corresponding message.

previous_history

temperature

double

0 to 2

Defaults to 0

Higher values mean the model will take more risks and value 0 (argmax sampling) works better for scenarios with a well-defined answer.

max_tokens

integer

≥ 1

Defaults to 4096

The maximum number of tokens to generate in the completion. The token count of your prompt plus max_tokens cannot exceed the model's context length.

tool_choice

string

enum

Defaults to auto

auto - auto
required - required
none - none

Allowed:

available_tools

array of objects

A list of tools the model may generate the right arguments for.

available_tools

tool_results

array of objects

List of results obtained from applying the tool_call arguments to your own tool.

tool_results

reasoning_effort

string

enum

Choices:

'low': Minimal reasoning, quick responses
'medium': Balanced reasoning approach
'high': In-depth, comprehensive reasoning

Example: 'high' for complex problem-solving tasks

low - low
medium - medium
high - high

Allowed:

fallback_type

string

enum

Defaults to continue

rerun - Rerun
continue - Continue

Allowed:

Response

200