Langchain
- Langchain is a framework which abstracts
- models
- tools
- prompts
- all the other necessary components
-
Langchain supports two languages
- python
- javascript
-
In Langchain, most of the classes are dervied from Runnable
-
To interact with llm we need prompts
- For llms (models) the base class is BaseChatModel
- When we are prompting we have different types of messsage
- System Prompt
- Message
- Response
- Tool Call Responses
- Langchain uses a class Called as BaseMessage and from this BaseMessage the other messages are derived
- SystemMessage -> To set system prompt (define a role or set a persona)
- HumanMessage -> This is the message to llm
- AIMessage -> This is response from llm
- ToolMessage -> This response includes tool calling structure
-
Refer Here for messages documentation
-
Base Prompt Template: is the base for all prompt templates
-
Understanding langchain packages
- langchain-core: This package contains all the base implementations
- langchain-community: Here we have additional packages to connect to vector database, documents, etc
- langchain: This is major package
- Vendor specific packages
- GCP: langchain-google-genai
- AWS: langchain-aws
- Azure: langchain-azure-ai
- openai: langchain-open-ai
Exercise: Lets create a project to get response from gcp vertex models
- create a new directory
hello_langchain - Lets initialize this folder
uv init - Lets add some packages
uv add- langchain-core
- langchain
- langchain-google-genai
-
Open Vs code and select interpretor
-
note: references
-
Refer Here for the work done in todays session
