Skip to content

Configuration

Pumbaa can be configured through the interactive wizard, command-line flags, environment variables, or a configuration file.

pumbaa config init

This wizard guides you through:

  1. 🤖 Choosing your LLM provider (Ollama, Gemini API, or Vertex AI)
  2. 🔑 Provider-specific settings (API key, model, etc.)
  3. 🔗 Cromwell server URL
  4. 📁 Optional WDL directory for chat context

Config Location

Configuration is saved to ~/.pumbaa/config.yaml

Configuration Commands

# Interactive setup wizard
pumbaa config init

# Set individual values
pumbaa config set <key> <value>

# Get a value
pumbaa config get <key>

# List all configuration
pumbaa config list

# Show config file path
pumbaa config path

⚙ Available Keys

Key Description Example
llm_provider LLM backend ollama, gemini, vertex
cromwell_host Cromwell server URL http://localhost:8000
ollama_host Ollama server URL http://localhost:11434
ollama_model Ollama model name llama3.2:3b
gemini_api_key Gemini API key AIza...
gemini_model Gemini model gemini-2.5-flash
vertex_project GCP project ID my-project
vertex_location Vertex AI region us-central1
vertex_model Vertex AI model gemini-2.5-flash
wdl_directory WDL files for context /path/to/workflows

Environment Variables

Every setting can also be provided via environment variable. Env vars override the config file.

Variable Equivalent config key Default
CROMWELL_HOST cromwell_host http://localhost:8000
PUMBAA_LLM_PROVIDER llm_provider ollama
OLLAMA_HOST ollama_host http://localhost:11434
OLLAMA_MODEL ollama_model llama3.2:3b
GEMINI_API_KEY gemini_api_key —
GEMINI_MODEL gemini_model gemini-2.5-flash
VERTEX_PROJECT vertex_project —
VERTEX_LOCATION vertex_location us-central1
VERTEX_MODEL vertex_model gemini-2.5-flash
PUMBAA_WDL_DIR wdl_directory —
PUMBAA_WDL_INDEX — (WDL index cache path) ~/.pumbaa/wdl_index.json
PUMBAA_SESSION_DB — (chat sessions database) ~/.pumbaa/sessions.db
PUMBAA_TELEMETRY_ENABLED telemetry_enabled true

Prefix inconsistency

Provider variables follow their ecosystem's conventions and do not use the PUMBAA_ prefix: it is OLLAMA_HOST, GEMINI_API_KEY, VERTEX_PROJECT — not PUMBAA_OLLAMA_HOST.

🔄 Configuration Priority

Settings are applied in this order (later overrides earlier):

flowchart LR
    A[Default Values] --> B[Config File]
    B --> C[Environment Variables]
    C --> D[Command-line Flags]

🔗 Cromwell Server

pumbaa config set cromwell_host http://cromwell.example.com:8000

export CROMWELL_HOST=http://cromwell.example.com:8000

pumbaa --host http://cromwell.example.com:8000 dashboard

Default

The default host is http://localhost:8000

🤖 Chat LLM Providers

Get your API key at Google AI Studio.

pumbaa config set llm_provider gemini
pumbaa config set gemini_api_key <your-api-key>

pumbaa config set llm_provider ollama
pumbaa config set ollama_host http://localhost:11434
pumbaa config set ollama_model llama3.2:3b

pumbaa config set llm_provider vertex
pumbaa config set vertex_project <project-id>
pumbaa config set vertex_location us-central1

🔒 Authentication

Pumbaa assumes a direct connection to a reachable Cromwell server; it does not perform authentication itself.

Port-forwarding from Kubernetes 
kubectl -n <namespace> port-forward svc/cromwell 8000:8000
pumbaa config set cromwell_host http://localhost:8000

📊 Telemetry

Pumbaa collects anonymous usage statistics to help improve the tool.

What we collect

  • Command name (e.g., pumbaa chat)
  • Pumbaa version
  • OS / Architecture
  • Execution duration
  • Success/Failure status

What we DO NOT collect

  • Argument or flag values
  • File paths or API keys
  • File contents
  • Personal information

Opt-Out

pumbaa config set telemetry_enabled false

export PUMBAA_TELEMETRY_ENABLED=false

📚 Next Steps