AIMan API-Client by brandCompete

Preconditions

Python version: >=3.8.1

Installation

pip install aiman-client

Getting started

Instantiate the api client

IMPORTANT: The host_url needs to be the base address of the api you want to connect to. The format of the host_url should look like this example:

from aiman.client import AimanClient

client = AimanClient(
    host_url="https://aiman-api-test.brandcompete.com",
    user_name="john@doe.com",
    password="top_secret")

Autorefresh JWT-Token

The client takes care of updating the token during the client's runtime if it has expired.

Fetching available AI-Models

This method returns a list of available models of type AIModel.

models = client.get_models()
for model in models:
    print(f"[default tag:{model.default_model_tag_id:4}] {model.name.upper():25} - {model.short_description}")

Prompting a simple query to a specific model

In order to submit a query the model for the prompt must be determined. This is done via the model_tag_id as parameter and can be taken from the AIModel object by default_model_tag_id property. Alternatively, you can also view available models here

response:dict = client.prompt(
    model_tag_id=10,
    query="Enter a question to prompt for")
print(response["responseText"])

Prompting a query and attach one or more files

response:dict = client.prompt(
    model_tag_id=1, 
    query="Enter a question to ask something about the given file content...", 
    attachments=["file/path/file_1.pdf", "file/path/file_2.xlsx"])

Raging with datasources and documents

Datasource

Init a new datasource with minimum requirements: name and summary

datasource_id = client.init_new_datasource(
    name="Test datasource", 
    summary="New datasource for uploading some documents")

Or init a new datasource with a list of tags and categories

datasource_id = client.init_new_datasource(
    name="Test datasource", 
    summary="New datasource for uploading documents", 
    tags=["tagA","tagB", "etc"], 
    categories=["catA","catB","etc"])

Fetch all datasources (associated to your account). Possible status:

• 2 (ready)
• 1 (indexing)
• 0 (pending)

datasources = client.fetch_all_datasources()
for source in datasources:
    print(f"{source.id}")
    print(f"{source.name}")
    print(f"{source.status}")

Documents

Add multiple documents into a datasource (can be url or file)

client.add_documents(
    data_source_id=your_ds_id, 
    sources=["path/to_my_data/test.pdf", "https://www.brandcompete.com"] )

Prompt on datasource context

Prompt in conjunction with a datasource_id. You have to use the model_tag_id to specify the model to prompt. NOTE: The datasource requires status == 2 and should be checked before prompting.

client.prompt_on_datasource(
    datasource_id=datasource_id, 
    model_tag_id=200, 
    query="can you please summarize the content?", 
    prompt_options = None)

Prompt options

You can pass prompt options as a optional parameter to any kind of prompt.

New instance with current properties and their default values.

from aiman.core.classes import PromptOptions

options = PromptOptions()
options.keep_context = True,
options.num_ctx = 8128
options.raw = True
options.temperature = 0.8
options.mirostat = 0
options.mirostat_eta = 0.1
options.mirostat_tau = 5
options.num_gqa = 8
options.num_gpu = 0
options.num_thread = 0
options.repeat_last_n = 64
options.repeat_penalty = 1.1
options.seed = 0

Executing prompt with options:

options = PromptOptions()
options.keep_context = False

response:dict = client.prompt(
    model_tag_id=10,
    query="Enter a question to prompt for"
    prompt_options = options
    )

Prompt on datasource example:

options = PromptOptions()
options.temperature = 0.2

client.prompt_on_datasource(
    datasource_id=datasource_id, 
    model_tag_id=200, 
    query="can you please summarize the content?", 
    prompt_options = options)