Re-enable Observability Features

Several observability features in CoreValue are gated behind COVA_FEATURE_* environment variables. This means they are not always available out of the box — you must explicitly opt in at deploy time by setting the appropriate environment variable.

Gated Features

How to Enable a Feature

Each feature is controlled by a boolean environment variable. Set the variable to "true" (or "on") at deploy time to enable the feature, or "false" (or "off") to disable it.

Cloud Run (staging / production)

Set the environment variable in your Cloud Run service configuration or via GCP Secret Manager:

gcloud run services update cova-core \
  --set-env-vars "COVA_FEATURE_EXPERIMENTS=true,COVA_FEATURE_EVALUATORS=true" \
  --region us-central1

For secrets managed via GCP Secret Manager, add the variable to your .env.staging file and push with the secrets script:

# .env.staging
COVA_FEATURE_EXPERIMENTS=true
COVA_FEATURE_EVALUATORS=true
COVA_FEATURE_DATASETS=true
COVA_FEATURE_PLAYGROUND=true
COVA_FEATURE_PROMPTS_UI=true

./infra/secrets/gcp-secrets.sh --env staging --env-file .env.staging

Local Development

Set the environment variable in your local .env file before starting the Core API:

# server/core/.env
COVA_FEATURE_EXPERIMENTS=true
COVA_FEATURE_EVALUATORS=true

Then restart the Core API server for the changes to take effect.

Feature Categories

Tracing

• Sessions (COVA_FEATURE_SESSIONS, default: ON) — Groups related requests into user sessions for tracing multi-step conversations and agent workflows. Enabled by default.

Eval

• Experiments (COVA_FEATURE_EXPERIMENTS, default: OFF) — Spreadsheet-like experience for tuning LLM prompts against production data with custom evaluators.
• Evaluators (COVA_FEATURE_EVALUATORS, default: OFF) — LLM-as-a-judge and custom evaluators for scoring request quality.
• Datasets (COVA_FEATURE_DATASETS, default: OFF) — Curate and export LLM request/response data for fine-tuning, evaluation, and analysis.

Dev Tooling

• Playground (COVA_FEATURE_PLAYGROUND, default: OFF) — Interactive LLM playground for testing prompts with different models and parameters.
• Prompts UI (COVA_FEATURE_PROMPTS_UI, default: OFF) — Prompt management interface for composing, versioning, and deploying prompts with dynamic variables.

Deprecation Notice

Related Documentation