Config¶

The Config dataclass holds all oneprompt settings. It can be created from environment variables, a .env file, or direct initialization.

from oneprompt import Config

config = Config.from_env()

Constructor¶

Config(
    gemini_api_key: str = "",
    database_url: str = "",
    gemini_model: str = "gemini-3-flash-preview",
    schema_docs: str = "",
    schema_docs_path: str = None,
    data_dir: str = "./op_data",
    host: str = "0.0.0.0",
    port: int = 8000,
    artifact_store_port: int = 3336,
    postgres_mcp_port: int = 3333,
    chart_mcp_port: int = 3334,
    python_mcp_port: int = 3335,
    artifact_store_token: str = "",
    agent_max_recursion: int = 10,
)

Parameters¶

Parameter	Type	Default	Env Variable	Description
`gemini_api_key`	`str`	`""`	`GOOGLE_API_KEY`	Google Gemini API key
`database_url`	`str`	`""`	`DATABASE_URL`	PostgreSQL connection string
`gemini_model`	`str`	`gemini-3-flash-preview`	`GEMINI_MODEL`	Gemini model name
`schema_docs`	`str`	`""`	`OP_SCHEMA_DOCS`	Inline schema documentation
`schema_docs_path`	`str`	`None`	`OP_SCHEMA_DOCS_PATH`	Path to schema docs file
`data_dir`	`str`	`./op_data`	`OP_DATA_DIR`	Local data directory
`host`	`str`	`0.0.0.0`	`OP_HOST`	API server bind address
`port`	`int`	`8000`	`OP_PORT`	API server port
`artifact_store_port`	`int`	`3336`	`OP_ARTIFACT_PORT`	Artifact Store port
`postgres_mcp_port`	`int`	`3333`	`OP_POSTGRES_MCP_PORT`	PostgreSQL MCP port
`chart_mcp_port`	`int`	`3334`	`OP_CHART_MCP_PORT`	Chart MCP port
`python_mcp_port`	`int`	`3335`	`OP_PYTHON_MCP_PORT`	Python MCP port
`artifact_store_token`	`str`	`""`	`OP_ARTIFACT_TOKEN`	Shared auth token
`agent_max_recursion`	`int`	`10`	`OP_MAX_RECURSION`	Max agent iterations

Class Methods¶

`from_env()`¶

Create a Config from environment variables:

config = Config.from_env()

Reads all settings from the corresponding environment variables listed in the parameters table above.

Computed Properties¶

Property	Type	Description
`artifact_store_url`	`str`	`http://localhost:{artifact_store_port}`
`mcp_postgres_url`	`str`	`http://localhost:{postgres_mcp_port}/mcp`
`mcp_chart_url`	`str`	`http://localhost:{chart_mcp_port}/mcp`
`mcp_python_url`	`str`	`http://localhost:{python_mcp_port}/mcp`
`export_dir`	`Path`	`{data_dir}/exports` — used by Docker containers for artifact storage
`state_db_path`	`str`	`{data_dir}/state.db`

config = Config.from_env()

print(config.artifact_store_url)  # http://localhost:3336
print(config.mcp_postgres_url)    # http://localhost:3333/mcp
print(config.export_dir)          # /absolute/path/op_data/exports
print(config.state_db_path)       # /absolute/path/op_data/state.db

Methods¶

`validate()`¶

Validate the configuration and return a list of errors:

validate() -> list[str]

config = Config()
errors = config.validate()
# ["gemini_api_key is required (env: GOOGLE_API_KEY)",
#  "database_url is required (env: DATABASE_URL)"]

`to_env_dict()`¶

Export configuration as an environment variables dictionary (used internally for Docker/subprocess):

to_env_dict() -> dict[str, str]

config = Config.from_env()
env = config.to_env_dict()
# {"GOOGLE_API_KEY": "...", "DATABASE_URL": "...", ...}

Path Resolution¶

Relative paths in data_dir are resolved to absolute paths at initialization time, based on the current working directory. This ensures consistent behavior regardless of any later changes to the working directory.

# If cwd is /home/user/project
config = Config(data_dir="./op_data")
print(config.data_dir)  # /home/user/project/op_data

# Absolute paths are kept as-is
config = Config(data_dir="/var/data/tp")
print(config.data_dir)  # /var/data/tp

Auto-loading Schema Docs¶

If schema_docs_path is provided and the file exists, the content is automatically loaded into schema_docs during initialization:

config = Config(
    gemini_api_key="...",
    database_url="...",
    schema_docs_path="./DATABASE.md",
)
# config.schema_docs now contains the file content