LLM4ELN llm4eln-assessment - Basic LLM Capabilities with LangChain

This tutorial will guide you through setting up a Python environment and working with Large Language Models (LLMs) using LangChain.

Table of Contents

• 1. Install uv (Fast Python Package Manager)
• 2. Create Project and Virtual Environment
• 3. Create a .env File
• 4. Initialize a LLM Object
• 5. Use the model with a simple prompt
• 6. Create a proof file
• 7. Push your proof

1. Install uv (Fast Python Package Manager)

Install uv on Linux

curl -LsSf https://astral.sh/uv/install.sh | sh

or on Windows with PowerShell

powershell -c "irm https://astral.sh/uv/install.ps1 | iex"

See uv docs for more information

2. Create Project and Virtual Environment

1. First create your own fork: https://github.com/LLM4ELN/llm4eln-assessment/fork

2. Then clone your fork and set up the project:

   git clone https://github.com/<YOUR_GITHUB_USERNAME>/llm4eln-assessment
   # or use ssh if you have configured ssh keys
   # git clone git@github.com:<YOUR_GITHUB_USERNAME>/llm4eln-assessment.git
   cd llm4eln-assessment

3. Ensure Python 3.12 is installed

   uv python install 3.12

4. Initialize Python project with uv

   uv init --python 3.12

5. Create local virutal environment

   uv venv

6. Activate Virtual Environment

   On Windows

   .venv\Scripts\activate

   Or on Linux

   source .venv/bin/activate

7. Install required packages

   uv add langchain langchain-openai langchain-ollama langchain-anthropic langchain-google-genai langchain-ollama python-dotenv

3. Create a .env File

See .env.example

# use one (!) of the following configurations

# Azure OpenAI
API_PROVIDER=azure
API_KEY=<your-api-key>
API_ENDPOINT=https://your-deplyoment.openai.azure.com/ # example
API_VERSION=2024-10-21 # example
API_MODEL=gpt-5-nano-2025-08-07 # example

# ollama
API_PROVIDER=ollama
API_MODEL=deepseek-r1:7b # example

# Blablador
API_PROVIDER=blablador
API_ENDPOINT=https://api.helmholtz-blablador.fz-juelich.de/v1/
API_MODEL=alias-code
API_KEY=<your-api-key>

4. Initialize a LLM Object

See tutorial.py

from dotenv import load_dotenv
from os import environ
load_dotenv()

# Initialize the language model

if environ.get("API_PROVIDER") == "azure":
    # https://docs.langchain.com/oss/python/integrations/providers/microsoft
    from langchain_openai import AzureChatOpenAI
    llm = AzureChatOpenAI(
        azure_deployment=environ.get("API_MODEL"),  # or your deployment
        api_version=environ.get("API_VERSION"),  # or your api version
        api_key=environ.get("API_KEY"),  # or your api key
        azure_endpoint=environ.get("API_ENDPOINT")
    )

if environ.get("API_PROVIDER") == "ollama":
    # https://docs.langchain.com/oss/python/integrations/chat/ollama
    from langchain_ollama import ChatOllama
    llm = ChatOllama(model=environ.get("API_MODEL"))

if environ.get("API_PROVIDER") == "blablador":
    # https://sdlaml.pages.jsc.fz-juelich.de/ai/guides/blablador_api_access/
    from langchain_openai import ChatOpenAI
    llm = ChatOpenAI(
        model=environ.get("API_MODEL"),
        api_key=environ.get("API_KEY"),
        base_url=environ.get("API_ENDPOINT")
    )
# ToDo: Gemini, Anthropic etc.

5. Use the model with a simple prompt

messages = [
    (
        "system",
        "You are a helpful assistant called 'LLM4ELN'."
        "Your purpose is to help users of electronic lab notebook."
    ),
    ("human", "What is your purpose?"),
]
ai_msg = llm.invoke(messages)
print(ai_msg.content)

6. Create a proof file

import subprocess  # noqa: E402

username = subprocess.run(
    ["git", "config", "user.name"],
    capture_output=True,
    text=True,
    check=True
).stdout.strip()

import hashlib  # noqa: E402

hash_value = hashlib.sha256(
    (username + str(messages)).encode()
).hexdigest()

file_name = f"{username}_{environ.get('API_PROVIDER')}_proof_{hash_value}.txt"
with open(f"./proof/{file_name}", "w") as f:
    f.write(environ.get('API_MODEL', '') + ":\n" + ai_msg.content)

7. Push your proof

git add proof/<your_proof_file>.txt
git commit -m "Add proof file for <user_name>"
git push origin main

Create a pull request by visiting

https://github.com/LLM4ELN/llm4eln-assessment/compare/main...<your_username>:llm4eln-assessment:main