Skip to content

Chat Agent

Interactive AI assistant for querying Cromwell workflows and reading files via natural language.

  • Natural Language

    Ask questions about workflows in plain English

  • Built-in Tools

    Query workflows, read GCS files, check status

  • Session Persistence

    Resume conversations with context retention

Quick Start

pumbaa config init

The wizard will guide you through setting up your preferred LLM provider.

pumbaa chat

LLM Providers

Choose your preferred AI backend:

pumbaa chat --provider gemini --gemini-api-key <API_KEY>

Get an API Key

Free tier available at Google AI Studio

pumbaa chat --provider vertex --vertex-project <PROJECT_ID>

GCP Project Required

Requires an active Google Cloud project with Vertex AI API enabled.

pumbaa chat --provider ollama

Default Option

Runs locally, no API key needed. Install from ollama.ai

Capabilities

  • Query Workflows

    Search by status, name, or labels

  • Get Status

    Check execution state in real-time

  • View Metadata

    Inspect workflow details and call info

  • Get Outputs

    List workflow output files

  • View Logs

    Access workflow log paths for debugging

  • Read GCS Files

    Fetch files from Google Cloud Storage

Controls

Key Action
Ctrl+D Send message
Up / Down Scroll messages (input mode)
Tab Toggle navigation mode
Y Copy selected message
Esc Exit chat

Session Management

Conversations are persisted in SQLite for context retention.

# List existing sessions
pumbaa chat --list

# Resume a session
pumbaa chat --session <SESSION_ID>

Configuration

Environment Variables

export PUMBAA_LLM_PROVIDER=gemini
export GEMINI_API_KEY=<your-api-key>
export GEMINI_MODEL=gemini-2.0-flash  # optional

export PUMBAA_LLM_PROVIDER=ollama
export OLLAMA_HOST=http://localhost:11434
export OLLAMA_MODEL=llama3.2:3b

export PUMBAA_LLM_PROVIDER=vertex
export VERTEX_PROJECT=<project-id>
export VERTEX_LOCATION=us-central1
export VERTEX_MODEL=gemini-2.0-flash

Session Storage

export PUMBAA_SESSION_DB=~/.pumbaa/sessions.db

Example Prompts

Try these prompts

  • "List my running workflows"
  • "What is the status of workflow abc-123?"
  • "Show me the outputs of workflow xyz-456"
  • "Read gs://bucket/path/to/file.txt"
  • "Query last 5 failed workflows"