CodexCliBackend

LlmBackend for the OpenAI codex CLI.

Usage

CodexCliBackend()

Invokes codex exec with an ephemeral session and a read-only sandbox.

Attributes

models: dict[TModel, str]: Mapping from abstract model size to an OpenAI model name.

Methods

Name	Description
build_command()	Build the `codex exec` argv for one stdin-prompted invocation.
env()	Return no extra environment variables; the `codex` CLI runs with the inherited environment.
parse_response()	Parse `codex` stdout into text or a validated model.

build_command()

Build the codex exec argv for one stdin-prompted invocation.

Usage

Source

build_command(model, schema_path, agent)

Every invocation runs an ephemeral session in a read-only sandbox. Non-agent invocations disable Codex hooks and MCP servers. A schema path adds --output-schema.

Parameters

model: str: OpenAI model name, e.g. gpt-5.5.
schema_path: str | None: Path to a JSON schema file passed to --output-schema, or None.
agent: bool: Whether the invocation may use tools / agent capabilities.

Returns

list[str]: The argv list to execute.

env()

Return no extra environment variables; the codex CLI runs with the inherited environment.

Usage

Source

env()

parse_response()

Parse codex stdout into text or a validated model.

Usage

Source

parse_response(raw, response_model)

Parameters

raw: str: Raw stdout from the codex CLI.
response_model: type[BaseModel] | None: Model to validate against, or None for raw text.

Returns

str | BaseModel: raw when response_model is None; otherwise raw validated as JSON against response_model.