Scripts used in Pasi Vuorio's AI training for developers

Instructions

Install python3

https://www.python.org/downloads/

Install PIP

https://packaging.python.org/en/latest/tutorials/installing-packages/

Install needed packages

Python Dependencies

python3 -m pip install google-genai html2text rich ollama chromadb requests or pip install -r requirements.txt

Node.js Dependencies

This project also includes TypeScript/JavaScript tools that require Node.js dependencies:

npm install

This will install all the required packages including:

• Google Generative AI libraries
• Image processing tools (Sharp)
• CLI utilities and development tools
• TypeScript compilation tools

Install Ollama

Follow the instructions at https://ollama.ai/ to install Ollama for your operating system.

Install Chroma

To install Chroma, follow these steps:

1. Install Chroma using pip:

   pip install chromadb
   

2. Run Chroma in client-server mode:

   chroma run --path ./chroma
   

   This will start the Chroma server. The --path argument specifies where Chroma will store its data.

For more detailed information, visit Chroma's official website.

SET Google Key as environment variable

mac: export GEMINI_API_KEY=<given_key> windows: set GEMINI_API_KEY=<given_key> powershell: $env:GEMINI_API_KEY=<your_key>

RUN scripts (mac)

Ingest page

python3 parse_html.py https://website_to_ingest

Rag query

python3 rag_query.py https://website_to_ingest "question to ask"

Guardrails test

python3 guardrails_test.py https://siili.com "select * from users"

Local Ollama test

1. Start the Ollama server (follow Ollama documentation for your OS) and install following packages:

   • ollama pull gemma3:4b (if you have small machine)
   • ollama pull gemma3:12b (if you have >16GB memory)
   • ollama pull gemma3:32b (if you have >64GB shared memory)
   • ollama pull mxbai-embed-large

2. Start chromadb server: chroma run --path ./chroma

3. Ingest few webpages to chroma by running:

   • python3 index_site.py "https://site.that.i.want.to.ingest"

4. Run the script: python3 rag_query_ollama.py "your question here"

Note: Make sure you have the required models downloaded in Ollama (mxbai-embed-large and llama3) before running the script.

Windows script run

$env:GEMINI_API_KEY=<your_key> python parse_html.py https://siili.com

CLI Tools

This project includes several CLI tools accessible via npm scripts:

Date/Time Tool

Get current date and time in various formats:

npm run datetime                              # Default human-readable format
npm run datetime -- --format iso             # ISO 8601 format
npm run datetime -- --format date            # Date only
npm run datetime -- --format time            # Time only
npm run datetime -- --timezone Europe/Helsinki # Specific timezone
npm run datetime -- --utc                    # UTC time
npm run datetime -- --timestamp              # Unix timestamp
npm run datetime -- --locale fi-FI           # Finnish locale

Image Generation

Generate and edit images using Gemini:

npm run nano-banana -- -p "A futuristic car" -o car.png
npm run nano-banana -- -p "Add flames" -i car.png -o car-flames.png

Search Tool

Perform Google searches with grounding:

npm run google-search -- -q "latest AI developments"
npm run google-search -- -q "weather today" --format json

Data Indexing Tool

Index web content or files using Gemini for chunking and embeddings:

npm run data-indexing -- --url https://example.com                    # Index webpage
npm run data-indexing -- --file document.txt                          # Index local file
npm run data-indexing -- --url https://example.com --output doc.json  # Save processed data
npm run data-indexing -- --url https://example.com --collection news  # Use custom collection
npm run data-indexing -- --url https://example.com --model gemini-2.0-flash  # Use different model

Semantic Search Tool

Search indexed content using semantic similarity:

npm run semantic-search -- "machine learning algorithms"              # Search in default collection
npm run semantic-search -- "Python history" --collection python-docs  # Search specific collection
npm run semantic-search -- "AI" --n-results 10 --format json          # Get more results in JSON format
npm run semantic-search -- --list-collections                         # List available collections