Embedding models

2024-04-07 · Source: Ollama Blog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Intermediate, short

Summary

Ollama now supports embedding models, enabling the creation of retrieval augmented generation (RAG) applications that integrate text prompts with existing documents. Embedding models generate vector embeddings, which are numerical arrays representing the semantic meaning of text. These vectors are stored in databases for similarity-based data retrieval. Ollama provides access to models like `mxbai-embed-large` (334M parameters), `nomic-embed-text` (137M parameters), and `all-minilm` (23M parameters). Users can generate embeddings via Ollama's REST API, Python, or JavaScript libraries, and the platform integrates with tools such as LangChain and LlamaIndex. An example RAG application demonstrates generating embeddings, retrieving relevant documents from a ChromaDB collection, and using a large language model like Llama 2 to generate a response based on the retrieved data and a user prompt.

Key takeaway

For AI Engineers building RAG applications, Ollama's new embedding model support simplifies the workflow. You can now easily generate and manage vector embeddings for your documents using Ollama's API or libraries, integrating with tools like LangChain or LlamaIndex. This enables you to build more context-aware LLM applications by retrieving semantically relevant information from your data stores, enhancing response accuracy and relevance.

Key insights

Ollama now supports embedding models for RAG applications, converting text into vector embeddings for semantic search.

Principles

• Vector embeddings represent text meaning numerically.
• Semantic search relies on comparing vector embeddings.

Method

Generate embeddings for documents, store them in a vector database, retrieve relevant documents based on a query's embedding, and then use an LLM to generate a response combining the query and retrieved data.

In practice

• Use `ollama pull [model_name]` to get embedding models.
• Integrate with LangChain or LlamaIndex for RAG workflows.
• Store embeddings in ChromaDB for efficient retrieval.

Topics

• Embedding Models
• Retrieval-Augmented Generation
• Vector Embeddings
• Ollama Platform
• LLM Integration

Code references

• ollama/ollama
• ollama/ollama-python
• ollama/ollama-js

Best for: Machine Learning Engineer, AI Engineer, Data Scientist

Related on AIssential

• Vector Search Using Ollama for Retrieval-Augmented Generation (RAG) — RAG combines vector search with LLMs to provide context-aware, fact-grounded responses from external data.
• Ollama: Unleash the Power of Local LLMs on Your Machine — Ollama enables private, cost-effective, and low-latency local LLM inference and customization via a Python library.
• Top 5 Embedding Models for Your RAG Pipeline — Effective RAG pipelines rely on embedding models that balance retrieval performance, real-world adoption, and deployment practicality.
• Building a PDF Question-Answering Chatbot with Spring AI: From PDF Upload to RAG-Powered Answers — RAG systems enhance LLM accuracy by grounding answers in retrieved, relevant private data, outperforming larger models without context.
• LangChain Explained: Understanding Models, Prompts, Chains, Memory, Indexes, and Agents — LangChain abstracts LLM complexities, enabling modular, context-aware, and tool-integrated AI application development.
• Web search — Ollama's new web search API enhances LLMs by providing real-time web data to improve accuracy and reduce hallucinations.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Ollama Blog.