Postgres Agent

Uses SmolAgents to interact with a Postgres DB. Incredibly fast with Groq's models, and quite intelligent at extracting information with Anthropics Claude models(Accessed here through Openrouter so it will cost ~10% more).

Prerequisites

• Python 3.8 or higher
• uv package manager
• PostgreSQL database

Running the Project

To install UV:

pip install uv

To run the agent:

uv run sfa_postgres.py

Command Line Arguments

The tool supports several command line arguments to customize the LLM model used for queries:

uv run sfa_postgres.py [query] [-a ALIAS] [-p PROVIDER] [-m MODEL]

Arguments:

• query: Natural language query to execute (optional, defaults to listing schemas and tables)
• -a, --alias: Use a predefined model alias. Available aliases:
  • groq/llama: Groq's Llama-3.3-70b
  • groq/deepseek: Groq's DeepSeek R1
  • openai/o1-mini: OpenAI's O1-mini model
  • openai/4o-mini: OpenAI's GPT-4o-mini
  • openrouter/4o-mini: OpenRouter's GPT-4o-mini
  • openrouter/sonnet: OpenRouter's Claude 3.5 Sonnet
• -p, --provider: Specify provider directly (groq, openai, or openrouter)
• -m, --model: Specify model ID directly when using provider

Examples:

# Use default model
uv run sfa_postgres.py "Show all users in the database"

# Use specific alias
uv run sfa_postgres.py -a groq/llama "List all tables"

# Use specific provider and model
uv run sfa_postgres.py -p openai -m gpt-4 "Count records in users table"