This is a modification of the LM Studio's documentation's example for interacting with the LLM using the OpenAI API and allowing it to make calls to "tools". In this case, tools are Python functions, and are exposed to the LLM using handcrafted schemas.
This was an attempt to unblock the AI from answering a question completely incorrectly due to its training on what the current date and current state of the world is (such as the currently elected political officials), where a particular model was very much locked into believing the situation from 2023-2024 is the current one.
Additionally, it was an attempt to get the model to not trust its own mathematical prowess and to instead diff the dates using a specialized tool.
The tool, as provided and modified, does not allow a chain of tool calls to be made. Therefore, a request to fetch some dates (such as birthdays) from Wikipedia followed by using a subtraction tool has not worked.
Set up:
python3 -m venv env
. env/bin/activate
python3 -m pip install -r requirements.txtRun with ollama:
# possibly needed: 'ollama serve' before:
ollama run llama3.2:latestthen in another tab:
OPENAI_API=http://127.0.0.1:11434/v1 OPENAI_MODEL=llama3.2:latest python3 python-use-example.pyOr, run LM Studio and install mlx-community/llama-3.2-3b-instruct from the GUI, then:
OPENAI_API=http://127.0.0.1:5001/v1 python3 python-use-example.pyIf you mkdir cache, then fetched articles will be stored locally.
Another model that seems to work reasonably reliably is
lmstudio-community/mistral-nemo-instruct-2407.
To run the existing tests in python_use_example_test.py, follow these steps:
-
Ensure you have the necessary dependencies installed. You can do this by running:
python3 -m venv env . env/bin/activate python3 -m pip install -r requirements.txt -
Run the tests using
pytest. You can do this by executing the following command in your terminal:pytest python_use_example_test.py
The pytest command will discover and run all the tests defined in python_use_example_test.py. The test results will be displayed in the terminal.
The repository includes a GitHub Actions workflow file python-app.yml to automate the testing process. The workflow is triggered on every push and pull request to the main branch. It sets up Python 3.8, installs dependencies, and runs pytest on python_use_example_test.py.
You can view the results of the GitHub Actions workflow by navigating to the "Actions" tab of your GitHub repository. The workflow will show the status of each test run, including any errors or failures.
-
compare pikachu and bulbasaur after looking them up and put it in a markdown table. tell me about what franchise they come from, and tell me history of companies that made that franchise
-
how long passed between birthday of Nikola Tesla and start of WW1?