RobotFramework-AI

RobotFramework-AI is a library that adds AI functionality to the Robot Framework. It can generate test data for you using the RealTestDataGenerator but also reply to your messages with the Chatbot.

Setup

To setup the library:

1. Start by cloning this repository.
2. Open a terminal and locate the robotframework-ai folder.

3. Use the following command to create the package.

   python setup.py sdist bdist_wheel
   

   This requires the wheel package to be installed, which can be installed with:

   pip install wheel
   

4. To install the library using the package just created, use the following command.

   pip install .\dist\RobotFrameworkAI-0.0.2-py3-none-any.whl
   

5. Setup the API key for the AI you want to use. This can be done by creating a new environment variable called OPENAI_KEY with your key as the value. More information about this in the AI models header below.

6. The library is now properly setup, to confirm it works, run the following code in a robot file. This will generate 3 addresses and create a log.html file.

   *** Settings ***
   Library   RobotFrameworkAI
   
   *** Test Cases ***
   Exec Test
       [Documentation]    Test
       ${response}    Generate Test Data    openai    address
       Log    ${response}
   

   Opening this testlog file in a browser and opening the suite Test, the test Exec test and the keyword Log will reveal a list with the 3 addresses in the INFO section.

   

AI models

Each module in the RobotFramework-AI library can support multiple different AI models. Each AI model needs an API key for the generation of test data. This key gets read directly from your environment variables. Each AI model has their own API key. To define a key, create a new variable with the name of the AI model capitalized followed by "_KEY". Then set this variable to your key. At the moment only OpenAI is supported.

Example API keys - OPENAI_KEY=278bxw4m89monwxmu89wm98ufx8hwxfhqwifmxou09qwxp09jmx - GEMINI_KEY=cavhjbcZCJKnvmzxcnzkcjkczckzcskjnjn7h38nwd923hdnind

RealTestDataGenerator

RealTestDataGenerator can generate test data for the Robot Framework similar to the library Faker. The RealTestDataGenerator however, generates real existing data, using AI.

To generate test data simply import the package and use the keyword: Generate Test Data This keyword takes various arguments, some being specific for the generation of certain types of test data.

The following arguments can be used (arguments prefixed by an * are required):

• *ai_model: str: The AI model to be used, e.g. "openai", "gemini", "copilot", etc. Currently supporting: "openai"
• *type: str: The type of test data to create, e.g. "address", "user_data", etc. Currently supporting: "address"
• model: str: AI model specfic. The model of the AI model to be used. E.g. "gpt-3.5-turbo" when using the "openai" AI model. Default per AI model:
• "openai" = "gpt-3.5-turbo"
• amount: int = 3 The amount of rows of test data to generate.
• format: str = None The format in which the test data will be given. If None, will return a 2 dimensional list.
• max_tokens: int = 256 The token limit for a conversation. Both prompt and response tokens will count towards this limit.
• temperature: float = 1 This value determines the creativity of the AI model. Can be anything from 0-2.
• top_p: float = 1 Similar to temperature. Determines the selection of tokens before selecting one. The higher the value the more less likely tokens get added to the selection. Can be anything from 0-2. At 1, only the top 50% of tokens will be used when selecting a token at 0 all tokens will be taken into consideration
• frequency_penalty: float = 0 Penalizes more frequent token reducing the chance of it reappearing. Negative values encourage it to reuse tokens. Can be anything from -2 to 2.
• presence_penalty: float = 0 Exact same as frequency_penalty except its scope is reduced to the immediate context. The immediate context can be seen as one or more paragrahps about a singular subject. Can be anything from -2 to 2.
• kwargs: dict: Additional arguments can be supplied for specific types of test data. These will be explained in per type below

NOTE: Be careful with changing the temperature, top p, frequency- and presence penalty as it will likely deviate from the format we expect it to return.

Addresses

When generating addresses additional argument are available. These arguments are as follows:

• Country:str: The country from which to create addresses. If None, will generate an address from anywhere. Default = None

Examples

Generate 3 addresses from anywhere using OpenAI:

Generate Test Data    openai    address

Generate 10 addresses from Finland using the gpt-3.5-turbo from OpenAI in the default format with a token limit at 1024, temperature at 1, top p at .5 and frequency- and presence penalty at 0

Generate Test Data    openai    address    gpt-3.5-turbo    10    None    1024    1    .5    0    0    country=finland

Chatbot

The Chatbot is a simple response generating library for Robot Framework similar to ChatGPT on the web. You can ask it a question or give it a task to have it automatically reply to your emails.

The following arguments can be used (arguments prefixed by an * are required):

• *ai_model: str: The AI model to be used, e.g. "openai", "gemini", "copilot", etc. Currently supporting: "openai"
• *message: str: The message you want to send to the AI model, e.g. "What is the weather today?"
• model: str: AI model specfic. The model of the AI model to be used. E.g. "gpt-3.5-turbo" when using the "openai" AI model. Default depends on AI model:
• "openai" = "gpt-3.5-turbo"
• max_tokens: int = 256 The token limit for a conversation. Both prompt and response tokens will count towards this limit.
• temperature: float = 1 This value determines the creativity of the AI model. Can be anything from 0-2.
• top_p: float = 1 Similar to temperature. Determines the selection of tokens before selecting one. The higher the value the more less likely tokens get added to the selection. Can be anything from 0-2. At 1, only the top 50% of tokens will be used when selecting a token at 0 all tokens will be taken into consideration
• frequency_penalty: float = 0 Penalizes more frequent token reducing the chance of it reappearing. Negative values encourage it to reuse tokens. Can be anything from -2 to 2.
• presence_penalty: float = 0 Exact same as frequency_penalty except its scope is reduced to the immediate context. The immediate context can be seen as one or more paragrahps about a singular subject. Can be anything from -2 to 2.

• keep_history: bool = False A flag to keep the chat history of previous messages. When settings this to True, your previous prompt and the response by the AI will be saved for the next message. This feature will keep the previous message, so if you want to send two messages and refer to your first message from the second message, you need to set this flag to True in the second message. Leaving this on for the third message aswell will keep both the first and second message.

  NOTE: This works by incorporating the previous messages into the prompt, this will charge you again for both the prompt and response. So leaving this on, could quickly drain all your tokens.

• response_format: dict = None Can be used to make the response compile to JSON. Set this to { "type": "json_object" } to make the response compile to JSON or None if it shouldn't necessarily.

Examples

Generate a response to a question using OpenAI

Generate Response    openai    What is the wheather today?

Declare a rule in the first message and refer to it in the second message

Generate Response    openai    If I say water you say fire
Generate Response    openai    Water    keep_history=True

Generate the meaning of life in a json format using the gpt-3.5-turbo from OpenAI in the default format with a token limit at 1024, temperature at 2, top p at .5 and frequency- and presence penalty at 0 without using the previous messages in the response.

Generate Response    openai    What is the meaning of life? In json.    gpt-3.5-turbo    1024    2    .5    0    0    False    {"type": "json_object"}

Assistant

The Assistant is a module that allows talking to AI and adding files to the conversation.

It functions similar to the Chatbot module, but with some changes. With the Assistant module AI Assistants can be created, these will exists for aslong as they are not deleted, meaning that they are still available several weeks later or indefinitely. In the mean time new AI Assistants can be created and talked to.

Files can be uploaded to the AI for inspection or to add knowledge. Similarly files can also be send along side a message to the AI Assistant. The difference being that files send alongsides message only remain aslong as the thread (conversation). When a new thread is created, the AI Assistant wont remember previous message send to it, including files. It will only remember files uploaded to the AI assistant itself.

The Assistant can also be given instructions and parameters to influence its behaviour. These can be given at creation or later be changed.

Keywords

The Assistant has 9 different keywords that can be used to interact with the AI assistant. The way the Assistant module works is that there can be multiple AI assistants. But there can only be 1 active assistant. All keywords where it isn't explicitely mentioned which assistant will be affected (all keywords that don't require an Id as argument), will be performed on the active assistant.

Create Assistant

This keyword creates a new assistant and sets it as the active assistant.

Returns the Id of the newly created assistant.

The following arguments can be used (arguments prefixed by an * are required):

• *ai_model: str: The AI model to be used, e.g. "openai", "gemini", "copilot", etc. Currently supporting: "openai"
• *name: str: The name of the assistant. This name will show up in responses and in the logs, e.g. "Bob". Max 256 characters
• *instructions: str: The instructions given to the assistant. Here you can explain how the assistant should behave, e.g. "You are a software engineer that can debug my code and explain it in a easy to understand manner". Max 256,000 characters
• model: str: AI model specfic. The model of the AI model to be used. E.g. "gpt-3.5-turbo" when using the "openai" AI model. Default depends on AI model:
• "openai" = "gpt-3.5-turbo"
• temperature: float = 1 This value determines the creativity of the AI model. Can be anything from 0-2.
• top_p: float = 1 Similar to temperature. Determines the selection of tokens before selecting one. The higher the value the more less likely tokens get added to the selection. Can be anything from 0-2. At 1, only the top 50% of tokens will be used when selecting a token at 0 all tokens will be taken into consideration
• response_format: dict = None Can be used to make the response compile to JSON. Set this to { "type": "json_object" } to make the response compile to JSON or None if it shouldn't necessarily.

Examples

Create a new assistant Test Assistant:

Set AI Model    openai
Set Name    Test Assistant
Create Assistant

Create a new assistant with all changing all default arguments:

Set AI Model    openai
Set Name    Full Featured Assistant
Set Instructions    This assistant is capable of handling various technical queries.
Set Model    gpt-3.5-turbo
Set Temperature    1.5
Set Top P    0.8
Set Response Format    { "type": "json_object" }
Create Assistant

Update Assistant

This keyword updates parameters of the active assistant.

This keyword takes the same arguments as the Create Assistant keyword

Returns a success message explaining which parameters have been updated.

The following arguments can be used (arguments prefixed by an * are required):

• *ai_model: str: The AI model to be used, e.g. "openai", "gemini", "copilot", etc. Currently supporting: "openai"
• name: str: The name of the assistant. This name will show up in responses and in the logs, e.g. "Bob". Max 256 characters
• instructions: str: The instructions given to the assistant. Here you can explain how the assistant should behave, e.g. "You are a software engineer that can debug my code and explain it in a easy to understand manner". Max 256,000 characters
• model: str: AI model specfic. The model of the AI model to be used. E.g. "gpt-3.5-turbo" when using the "openai" AI model. Default depends on AI model:
• "openai" = "gpt-3.5-turbo"
• temperature: float = 1 This value determines the creativity of the AI model. Can be anything from 0-2.
• top_p: float = 1 Similar to temperature. Determines the selection of tokens before selecting one. The higher the value the more less likely tokens get added to the selection. Can be anything from 0-2. At 1, only the top 50% of tokens will be used when selecting a token at 0 all tokens will be taken into consideration
• response_format: dict = None Can be used to make the response compile to JSON. Set this to { "type": "json_object" } to make the response compile to JSON or None if it shouldn't necessarily.

Examples

Update the active assistant's name:

Set AI Model    openai
Set Name    Mark
Update Assistant

Update the all the parameters of the active assistant:

Set AI Model    openai
Set Name    Helpy
Set Instructions    You write documentation for code.
Set Model    gpt-3.5-turbo
Se