Creating a simple assistant
The AI Assistant API feature is at the Preview stage.
AI Assistant API is a tool for creating AI assistants. It can be used to create personalized assistants, implement a generative response scenario adapted based on external information (known as retrieval augmented generation, or RAG), and save the model's request context.
Getting started
To use the examples:
-
Create a service account and assign the
ai.assistants.editor
andai.languageModels.user
roles to it. -
Get the service account API key and save it.
The following examples use API key authentication. Yandex Cloud ML SDK also supports IAM token and OAuth token authentication. For more information, see Authentication in Yandex Cloud ML SDK.
-
Use the pip
package manager to install the ML SDK library:pip install yandex-cloud-ml-sdk
Create an assistant
This example shows how to create an assistant to store your conversations with YandexGPT Pro RC.
-
Create a file named
simple-assistant.py
and paste the following code into it:#!/usr/bin/env python3 from __future__ import annotations from yandex_cloud_ml_sdk import YCloudML def main(): sdk = YCloudML( folder_id="<folder_ID>", auth="<API_key>", ) # Creating a thread to store the conversation # The thread will be stored for five days thread = sdk.threads.create( name="SimpleAssistant", ttl_days=5, expiration_policy="static" ) print(f"new {thread=}") # Specifying YandexGPT Pro RC for model and setting its maximum context length model = sdk.models.completions("yandexgpt", model_version="rc") # Creating an assistant for the model # The assistant will be deleted four days after its last activity assistant = sdk.assistants.create( model, ttl_days=4, expiration_policy="since_last_active", max_tokens=500 ) # Here you should make provision for reading user messages # For now, let's just write something to the thread input_text = "" while input_text != "exit": input_text = input("Enter your question to the assistant: ") if input_text != "exit": thread.write(input_text) # This way you can give the whole thread contents to the model run = assistant.run(thread) print(f"{run=}") # To get the result, wait until the run is complete result = run.wait() # You can view all the result fields print(f"run {result=}") # The `text` field stores a string which is convenient for later use print("Answer:", result.text) # You can view all the messages stored in the thread print("Outputting the whole message history when exiting the chat:") for message in thread: print(f" {message=}") print(f" {message.text=}\n") # Deleting everything you no longer need thread.delete() assistant.delete() if __name__ == "__main__": main()
Where:
-
<folder_ID>
: ID of the folder in which the service account was created. -
<API_key>
: Service account API key you got earlier required for authentication in the API.The following examples use API key authentication. Yandex Cloud ML SDK also supports IAM token and OAuth token authentication. For more information, see Authentication in Yandex Cloud ML SDK.
-
-
Run the created file:
python3 simple-assistant.py
The example implements the simplest chat possible: enter your requests to the assistant from your keyboard and get answers. To end the dialog, enter
exit
.Approximate result
new thread=Thread(id='fvt67i1ettn5********', expiration_config=ExpirationConfig(ttl_days=5, expiration_policy=<ExpirationPolicy.STATIC: 1>), name='SimpleAssistant', description=None, created_by='ajegtlf2q28a********', created_at=datetime.datetime(2024, 12, 15, 21, 5, 33, 8133), updated_by='ajegtlf2q28a********', updated_at=datetime.datetime(2024, 12, 15, 21, 5, 33, 8133), expires_at=datetime.datetime(2024, 12, 20, 21, 5, 33, 8133), labels=None) Enter your question to the assistant: Hi! run=Run(id='fvtm4n1o2hla********', assistant_id='fvthtngdnlkq********', thread_id='fvt67i1ettn5********', created_by='ajegtlf2q28a********', created_at=datetime.datetime(2024, 12, 15, 21, 5, 53, 312504), labels=None, custom_temperature=None, custom_max_tokens=None, custom_max_prompt_tokens=None) run result=RunResult(status=<RunStatus.COMPLETED: 4>, error=None, _message=Message(id='fvtburadn35c********', parts=('Hello! What can I do for you?',), thread_id='fvt67i1ettn5********', created_by='ajegtlf2q28a********', created_at=datetime.datetime(2024, 12, 15, 21, 5, 53, 728330), labels=None, author=Author(id='fvthtngdnlkq********', role='ASSISTANT')), usage=Usage(input_text_tokens=12, completion_tokens=6, total_tokens=18)) Answer: Hello! What can I do for you? Enter your question to the assistant: How much is 2 + 2? run=Run(id='fvt2eajhkqnb********', assistant_id='fvthtngdnlkq********', thread_id='fvt67i1ettn5********', created_by='ajegtlf2q28a********', created_at=datetime.datetime(2024, 12, 15, 21, 7, 36, 751185), labels=None, custom_temperature=None, custom_max_tokens=None, custom_max_prompt_tokens=None) run result=RunResult(status=<RunStatus.COMPLETED: 4>, error=None, _message=Message(id='fvtnmfeavsm1********', parts=('2 + 2 = 4.',), thread_id='fvt67i1ettn5********', created_by='ajegtlf2q28a********', created_at=datetime.datetime(2024, 12, 15, 21, 7, 37, 294861), labels=None, author=Author(id='fvthtngdnlkq********', role='ASSISTANT')), usage=Usage(input_text_tokens=36, completion_tokens=9, total_tokens=45)) Answer: 2 + 2 = 4. Enter your question to the assistant: How about 15,234 - 265? run=Run(id='fvt55e1qr4d0********', assistant_id='fvthtngdnlkq********', thread_id='fvt67i1ettn5********', created_by='ajegtlf2q28a********', created_at=datetime.datetime(2024, 12, 15, 21, 7, 59, 483141), labels=None, custom_temperature=None, custom_max_tokens=None, custom_max_prompt_tokens=None) run result=RunResult(status=<RunStatus.COMPLETED: 4>, error=None, _message=Message(id='fvtujctocrak********', parts=('15234 − 265 = 14969.',), thread_id='fvt67i1ettn5********', created_by='ajegtlf2q28a********', created_at=datetime.datetime(2024, 12, 15, 21, 8, 0, 244274), labels=None, author=Author(id='fvthtngdnlkq********', role='ASSISTANT')), usage=Usage(input_text_tokens=68, completion_tokens=19, total_tokens=87)) Answer: 15,234 − 265 = 14,969. Enter your question to the assistant: exit Outputting the whole message history when exiting the chat: message=Message(id='fvtujctocrak********', parts=('15234 − 265 = 14969.',), thread_id='fvt67i1ettn5********', created_by='ajegtlf2q28a********', created_at=datetime.datetime(2024, 12, 15, 21, 8, 0, 244274), labels=None, author=Author(id='fvthtngdnlkq********', role='ASSISTANT')) message.text='15234 − 265 = 14969.' message=Message(id='fvthbaqg1rep********', parts=('And 15234 - 265?',), thread_id='fvt67i1ettn5********', created_by='ajegtlf2q28a********', created_at=datetime.datetime(2024, 12, 15, 21, 7, 59, 375702), labels=None, author=Author(id='fvt82v4kgb7i********', role='USER')) message.text='How about 15,234 - 265?' message=Message(id='fvtnmfeavsm1********', parts=('2 + 2 = 4.',), thread_id='fvt67i1ettn5********', created_by='ajegtlf2q28a********', created_at=datetime.datetime(2024, 12, 15, 21, 7, 37, 294862), labels=None, author=Author(id='fvthtngdnlkq********', role='ASSISTANT')) message.text='2 + 2 = 4.' message=Message(id='fvtdd4fbb4n8********', parts=('How much is 2 + 2?',), thread_id='fvt67i1ettn5********', created_by='ajegtlf2q28a********', created_at=datetime.datetime(2024, 12, 15, 21, 7, 36, 625764), labels=None, author=Author(id='fvt82v4kgb7i********', role='USER')) message.text='How much is 2 + 2?' message=Message(id='fvtburadn35c********', parts=('Hello! What can I do for you?',), thread_id='fvt67i1ettn5********', created_by='ajegtlf2q28a********', created_at=datetime.datetime(2024, 12, 15, 21, 5, 53, 728331), labels=None, author=Author(id='fvthtngdnlkq********', role='ASSISTANT')) message.text='Hello! What can I do for you?' message=Message(id='fvt9apvs2au9********', parts=('Hi!',), thread_id='fvt67i1ettn5********', created_by='ajegtlf2q28a********', created_at=datetime.datetime(2024, 12, 15, 21, 5, 53, 136965), labels=None, author=Author(id='fvt82v4kgb7i********', role='USER')) message.text='Hi!'
See also
- Creating an assistant with a search index
- Examples of working with ML SDK on GitHub