Configuration (Reference)

`~/.vectora/config.toml`

Runtime configuration, generated/edited by vectora config or the Settings UI. Minimal example:

[global]
mode = "desktop"   # or "cli", "headless", "mcp", "web"

[runtime]
auto_update = true

[server]
host = "0.0.0.0"
port = 8080

Variable	Required?	Description
`LLM_PROVIDER`	Yes	`google-genai`, `openai`, `anthropic`, `cohere`, or `ollama`
`GOOGLE_API_KEY` / `OPENAI_API_KEY` / `ANTHROPIC_API_KEY`	Depends on provider	the chosen LLM provider’s key
`COHERE_API_KEY`	Yes (for RAG)	embeddings + reranking
`TAVILY_API_KEY`	Optional	web search
`STORAGE_MODE`	No (default `lite`)	`lite` or `complete` — see Storage
`POSTGRES_DSN`	Only in `complete`	e.g. `postgresql+asyncpg://user:pass@host:5432/vectora`
`QDRANT_URL` / `QDRANT_API_KEY`	Only in `complete`	Qdrant endpoint
`REDIS_URL`	Only in `complete`	Redis endpoint
`TELEMETRY_ENABLED`	No (default `true`)	Enables native agent telemetry (turn start/end, tool calls, unhandled errors) via structured logging
`TELEMETRY_OUTPUT_PATH`	Optional	Path to a dedicated JSONL file for telemetry events. If unset, events go to the standard JSON log (`~/.vectora/logs/backend.jsonl`)

defaults.env (built-in)  →  .env (project)  →  ~/.vectora/.env (user)  →  per-user overrides (database)

See Configuration for the full usage guide.