Configuration Schema
This page documents the complete configuration schema for ~/.mooshik/config.toml and every key supported by mooshik config set.
Settable Keys Reference
The following 19 keys can be updated using mooshik config set <key> <value>:
| Key | Expected Type | Description |
|---|---|---|
store.kind | sqlite | postgres | Graph storage backend. |
store.dsn_secret | Secret Name | Vault secret name holding PostgreSQL connection DSN. |
store.path | File Path | SQLite database file path. |
embedder.kind | fixture | bge_m3 | gemini | Embedding provider. |
embedder.dim | Positive Integer | Vector embedding dimension (e.g. 1024 or 1536). |
embedder.gemini_project | String | Google Cloud project ID for embeddings. |
embedder.gemini_location | String | Google Cloud region for embeddings (us-central1). |
embedder.gemini_model | String | Embedding model name (gemini-embedding-001). |
embedder.gemini_credentials | File Path | Path to service account JSON for embeddings. |
daemon.flush_interval_ms | Positive Integer | Background flush interval in milliseconds. |
companion.base_url | URL | Endpoint URL for OpenAI-compatible /v1 companion. |
companion.model | String | Language model identifier. On the Google posture this needs the publisher prefix: google/gemini-3.7-flash. |
companion.api_key_secret | Secret Name | Vault secret name holding companion API bearer key. |
companion.auth | none | bearer | google | Authentication method for the companion endpoint. |
companion.google_project | String | Google Cloud project ID for Vertex AI companion. |
companion.google_location | String | Google Cloud location for Vertex AI (global). |
companion.google_credentials | File Path | Path to service account JSON for Vertex AI companion. |
companion.context_window | Positive Integer | Maximum context window tokens. |
companion.temperature | Number | Sampling temperature for model completions. |
TOML Tables Structure
[vault]
Configures the local secret vault encryption provider.
[vault]provider = "keyring" # "keyring" or "passphrase"[store]
Defines graph storage persistence.
[store]kind = "postgres"dsn_secret = "store-dsn"[embedder]
Defines vector generation parameters.
[embedder]kind = "gemini"dim = 1536gemini_location = "us-central1"gemini_model = "gemini-embedding-001"[companion]
Configures the language model for conversational turns and tool execution.
[companion]auth = "google"google_project = "my-project"google_location = "global"model = "google/gemini-3.7-flash"context_window = 32768temperature = 0.2[permissions]
Controls tool execution policies.
[permissions]memory = ["recall", "derive"]scratch = "prompt""mcp.news.*" = "allow""mcp.coder.*" = "prompt"[mcp_servers.<name>]
Configures external stdio Model Context Protocol servers.
[mcp_servers.news]command = "/home/you/.local/share/mooshik/venv/bin/mooshik-news-mcp"expose = ["search_news", "fetch_article"]
[mcp_servers.news.env]MOOSHIK_GEMINI_PROJECT = "gemini-project"