env.sample

# KB environment variables
# Copy this file to .env and fill in real values.
# .env is gitignored and never committed.

# ---------------------------------------------------------------------------
# GitLab personal access token
# ---------------------------------------------------------------------------
# Required scope: read_api
#   read_api grants read-only access to the API, which is sufficient for
#   fetching repository archives via /api/v4/projects/:id/repository/archive.
#   Do NOT use a token with broader scopes (api, write_repository, etc.)
#   unless you have a specific reason.
#
# Generate one at: https://<your-gitlab>//-/user_settings/personal_access_tokens
KB_GITLAB_TOKEN=glpat-xxxxxxxxxxxxxxxxxxxx

# ---------------------------------------------------------------------------
# GitLab source repo URL
# ---------------------------------------------------------------------------
# The full URL of the GitLab repository to index.
# This is the repo's web URL (not the clone URL).
# Used by `kb sources add` and by integration tests when KB_SOURCE_URL is set.
#
# Examples:
#   https://gitlab.com/myorg/myrepo
#   https://gitlab.example.com/shared/docs
KB_SOURCE_URL=https://gitlab.example.com/org/repo

# ---------------------------------------------------------------------------
# OpenAI API key (for embedding generation)
# ---------------------------------------------------------------------------
# Used by the embed handler to call OpenAI's embeddings API.
# The key is sent as a Bearer token in the Authorization header.
#
# Required for the embedding pipeline — without it, `kb reindex` will enqueue
# embed jobs that fail when the embed worker processes them.
#
# Model: text-embedding-3-large (3072 dimensions, configured in kb.yaml)
#
# Generate one at: https://platform.openai.com/api-keys
# Use a service account key (sk-svcacct-...) for production, or a user key
# (sk-...) for local dev. The key only needs the "Embeddings" permission.
#
# Viper maps this to config.Embedding.APIKey (embedding.api_key in YAML).
KB_EMBEDDING_API_KEY=sk-xxxxxxxxxxxxxxxxxxxx

# ---------------------------------------------------------------------------
# LLM API key (for the knowledge agent)
# ---------------------------------------------------------------------------
# Used by the agent to call the LLM API at query time.
# Supports OpenAI and Anthropic — set provider in config (agent.provider).
#
# If using OpenAI for both embeddings and the agent, you can reuse the same key.
# Default model per provider (when agent.default_model is empty):
#   openai    → gpt-5.2
#   anthropic → claude-sonnet-4-5
#
# Viper maps this to config.Agent.APIKey (agent.api_key in YAML).
KB_AGENT_API_KEY=