Skip to content

Installation

All Candela components live in the candelahq/candela monorepo.

The fastest way to get started — a developer proxy that manages local and cloud models with built-in observability.

Terminal window
brew install candelahq/tap/candela

Create ~/.config/candela/config.yaml:

runtime_backend: ollama
port: 8181
lm_studio_port: 1234

To use cloud models (Gemini, Claude via Vertex AI) or connect to a shared Candela server, authenticate once:

Terminal window
candela auth login # opens browser — Google OAuth flow
candela auth status # verify credentials

Credentials are stored at ~/.config/gcloud/application_default_credentials.json (ADC-compatible). No gcloud CLI required.

To use Anthropic models via AWS Bedrock, authenticate with your AWS credentials:

Terminal window
candela auth login --provider aws # configures AWS credential chain
candela auth status # verify both GCP and AWS credentials

This configures the AWS credential chain for SigV4 signing. You can also set AWS_PROFILE, AWS_ACCESS_KEY_ID/AWS_SECRET_ACCESS_KEY, or use IAM roles — candela respects the standard AWS SDK credential resolution order.


The full backend with dashboard UI, deployed on Cloud Run.

Terminal window
git clone https://github.com/candelahq/candela.git
cd candela
nix develop -c go run ./cmd/candela-server

See candela-server docs for Cloud Run deployment with Terraform.


Lightweight production proxy for container environments.

Terminal window
docker pull ghcr.io/candelahq/candela-sidecar:latest

Desktop app for provider management and trace visualization.

Terminal window
brew install --cask candelahq/tap/candela-desktop

Terminal window
candela version