LLM Basics

Simple, self-contained examples showing how to use an OpenAI-compatible API for various AI tasks. The scripts work with OpenAI directly or any other OpenAI-compatible endpoint (vLLM, LLM Factory, etc.) — just point config.ini at your server and pick your model IDs.

Setup Instructions

1. Install the dependencies:

   pip install -r requirements.txt

2. Copy config.ini.example to config.ini
3. Add your API key to the [API] section (LLM Factory keys are issued at https://llm-factory.ai.uky.edu/api-keys)
4. The [MODELS] section ships with the models currently served by the example LLM Factory endpoint (https://api-llm-factory.ai.uky.edu/v1). Using a different OpenAI-compatible server? Update the base URL and model IDs — run python test_models_endpoint.py to list what it offers (each entry includes a task field indicating what the model is for).

Example Scripts

• chat_completions_test.py – Chat with a language model
• stream_test.py – Stream chat responses token by token
• tools_test.py – Use tool-augmented language models (function calling)
• vision_test.py – Process images with multimodal models
• embeddings_test.py – Create vector embeddings from text
• simple_rag.py – Implement a basic RAG (Retrieval-Augmented Generation) system with FAISS
• transcriptions_test.py – Generate transcriptions from audio files (uses the included fitnessgram.mp3 sample)
• test_tts.py – Generate speech from text (writes test_<voice>.mp3 files)
• test_models_endpoint.py – List the models available on the configured server
• decisions_test.py – Probe the decisions endpoints (/v1/systemone Nimble / TypeSafe System One and /v1/decisions OpenAI Decisions format) and report whether they work on the configured server
• quota_test.py – Probe rate-limit/quota behavior (HTTP 429s, Retry-After and X-RateLimit headers, /quotas/me) if your server implements it

Run each script after configuring to test different capabilities of the API.

Imported from gh:Kentucky-Open-Science/llm-basics. Source last updated 2026-11-05. Synced 2026-11-05.