- Large language model gives unstructured data - non deterministic output
- Agent - extra wrapper on the LLMs like openai, meta’s lama, tells how the data should be provided.
- some agents are like Langchain, chatgpt from openai etc.
Data Science Service (LLM service)
- Scrapes the message from the user and records the expense along with merchant.
- Useful to extract the data with permission of user and automatically add it into the database.
Tech Stack for the DS service-
- No need to make the service with low latency, since it works in background mode and has a work to just extract and add the expense. we can use any high level language.
- hence, Python can be used to implement the service. (good library support)
- LangChain (a cost is low for personal projects and can be swapped later with other agent).
- LangChain:
- LangChain is a framework for building applications using large language models (LLMs) like GPT. (It is not just a agent, but it includes agents as one of its core features).
- It helps connect LLMs with external data sources, APIs, and tools for richer functionality.
- Provides components for prompt management, conversation memory, and multi-step workflows.
- Commonly used for chatbots, AI assistants, and apps combining LLMs with real-world data.
- MistralAI (langchain_mistralAI):