Key Terms

A

API (Application Programming Interface)

A set of functions and protocols for building software applications. Ollama provides an API for programmatic access.

C

Chat Model

A language model optimized for conversational interactions (e.g., Llama 3.2).

ChromaDB

An open-source vector database for storing and querying embeddings.

Context Window

The maximum amount of text (in tokens) a model can process at once. Llama 3.2 has a 128K token context window.

Cosine Similarity

A metric for measuring similarity between two vectors, commonly used with embeddings.

E

Embedding

A numerical representation of text that captures semantic meaning. Typically a high-dimensional vector.

Embedding Model

A specialized model for converting text into embeddings (e.g., nomic-embed-text).

F

Few-Shot Learning

Providing a model with a few examples to guide its behavior on a task.

Fine-Tuning

Training a pre-trained model on specific data to specialize it for a task.

G

GUI (Graphical User Interface)

Visual interface for interacting with software. Ollama now includes a native GUI.

H

Hallucination

When an LLM generates plausible-sounding but incorrect information.

I

Inference

The process of using a trained model to generate predictions or responses.

L

LLM (Large Language Model)

A neural network model trained on vast amounts of text data to understand and generate human language.

M

Model Parameters

Internal weights learned during training that determine model behavior. Llama 3.2 has 1B or 3B parameters.

N

nomic-embed-text

An embedding model with 1,024 dimensions, optimized for semantic search.

O

Ollama

An open-source tool for running LLMs locally on personal hardware.

P

Prompt

The input text provided to an LLM to generate a response.

Prompt Engineering

The practice of crafting effective prompts to get desired outputs from LLMs.

Q

Quantization

Reducing the numerical precision of model weights to decrease model size and memory usage.

R

RAG (Retrieval-Augmented Generation)

A technique that combines information retrieval with text generation to produce more accurate, grounded responses.

Retriever

The component in RAG that finds relevant documents based on a query.

S

Semantic Search

Searching for information based on meaning and context rather than exact keyword matches.

Streaming

Generating and displaying model output in real-time as it’s produced.

System Prompt

An instruction given to a model to define its role or behavior for a conversation.

T

Temperature

A parameter controlling randomness in model outputs. Lower = more deterministic, higher = more creative.

Token

The smallest unit of text processed by an LLM. Can be a word, part of a word, or punctuation.

Tokenization

The process of breaking text into tokens.

Top-K

Limiting the model to consider only the top K most likely next tokens.

Top-P (Nucleus Sampling)

Selecting from the smallest set of tokens whose cumulative probability exceeds P.

V

Vector

An array of numbers. Embeddings are represented as high-dimensional vectors.

Vector Database

A database optimized for storing and querying vector embeddings.

Z

Zero-Shot Learning

Asking a model to perform a task without providing examples, relying on its training.

---

Common Abbreviations

• AI: Artificial Intelligence
• ANN: Approximate Nearest Neighbor
• API: Application Programming Interface
• BERT: Bidirectional Encoder Representations from Transformers
• CLI: Command-Line Interface
• GPU: Graphics Processing Unit
• GUI: Graphical User Interface
• LLM: Large Language Model
• ML: Machine Learning
• NLP: Natural Language Processing
• RAG: Retrieval-Augmented Generation
• RAM: Random Access Memory

---

Units and Measurements

B (Billion): 1,000,000,000 (e.g., 3B parameters)

GB (Gigabyte): ~1 billion bytes of storage

K (Thousand): 1,000 (e.g., 128K tokens = 128,000 tokens)

ms (Millisecond): 1/1,000 of a second

Tokens/Second: Speed metric for text generation