Skip to main content

Baselinr Chat - Getting Started

The Baselinr Chat interface allows you to interactively query and investigate your data quality monitoring using natural language.

Prerequisites​

  1. LLM Provider Configuration: You need an LLM provider configured (OpenAI, Anthropic, Azure, or Ollama)
  2. Profiling Data: Run some profiling first to have data to query
  3. Python packages: rich and optionally prompt_toolkit for enhanced CLI experience

Quick Start​

1. Configure LLM in your config.yml​

# config.yml
source:
  type: postgres
  host: localhost
  database: mydb
  username: user
  password: password

storage:
  connection:
    type: postgres
    host: localhost
    database: baselinr_meta
    username: user
    password: password

# Add LLM configuration for chat
llm:
  enabled: true
  provider: openai      # or anthropic, azure, ollama
  model: gpt-4o-mini    # provider-specific model name
  # api_key: ${OPENAI_API_KEY}  # Optional, uses env var by default

2. Set your API key​

For OpenAI:

export OPENAI_API_KEY=your-api-key

For Anthropic:

export ANTHROPIC_API_KEY=your-api-key

3. Start the chat​

baselinr chat --config config.yml

4. Ask questions!​

🧑 You: what tables have been profiled recently?

🤖 Baselinr:
I found 5 tables profiled in the last 7 days:

1. **orders** - Last profiled 2 hours ago
   - 2.3M rows, 15 columns
   - No active alerts

2. **customers** - Last profiled 3 hours ago
   - 450K rows, 22 columns
   - 1 HIGH severity drift alert on email column

[...]

Available Commands​

Within the chat session, you can use these commands:

CommandDescription
/helpShow available commands
/clearClear conversation history
/historyShow conversation history
/statsShow session statistics (tokens used, cost estimate)
/toolsList available query tools
/verboseToggle verbose mode
/exitExit the chat

CLI Options​

baselinr chat --config config.yml [OPTIONS]

Options:
  --max-iterations N    Maximum tool-calling iterations (default: 5)
  --max-history N       Maximum messages in context (default: 20)
  --tool-timeout N      Tool execution timeout in seconds (default: 30)
  --show-tools          Show tool calls in output
  --verbose, -v         Enable verbose output

Next Steps​