Struct TurnDriverConfig 

pub struct TurnDriverConfig {

    pub url: String,
    pub model: String,
    pub kind: BackendKind,
    pub api_key: Option<String>,
    pub workspace: String,
    pub caveats: Caveats,
    pub max_tool_rounds: usize,
    pub workflow_grace_rounds: usize,
    pub tool_output_lines: usize,
    pub num_ctx: Option<u32>,
    pub connect_timeout_secs: u64,
    pub inference_timeout_secs: u64,
    pub mid_loop_trim_threshold: usize,
    pub mid_loop_trim_tokens: Option<usize>,
    pub max_ok_input: Option<u32>,
    pub build_check_cmd: Option<String>,
    pub safe_context: Option<u32>,
}

Owned, 'static configuration for one TurnDriver — the cloneable counterpart of the borrow-heavy ChatCtx. The driver clones it into each spawned turn task so nothing borrows across the async boundary.

The fields mirror the inference-relevant subset of ChatCtx; the session-bound &mut handles (note sink, recall, permission gate, summarizer, compress state) are intentionally absent — a driven cowork turn is headless.

Fields

url: String

Inference endpoint base URL.

model: String

Model name.

kind: BackendKind

Wire protocol of the backend.

api_key: Option<String>

Bearer token for authenticated OpenAI-compatible endpoints.

workspace: String

Absolute workspace path the tool loop runs against.

caveats: Caveats

Permission caveats enforced for this turn’s tool calls.

max_tool_rounds: usize

Maximum tool-call rounds before a forced final completion.

workflow_grace_rounds: usize

Additional progress-aware rounds after max_tool_rounds; 0 makes the normal cap hard.

tool_output_lines: usize

Max lines of tool output shown inline.

num_ctx: Option<u32>

Ollama options.num_ctx (ignored on the OpenAI path).

connect_timeout_secs: u64

TCP connect timeout.

inference_timeout_secs: u64

Total inference timeout.

mid_loop_trim_threshold: usize

Message count at which the in-flight conversation is trimmed mid-turn.

mid_loop_trim_tokens: Option<usize>

Token threshold that also triggers a mid-loop trim. None disables it.

max_ok_input: Option<u32>

Highest input-token count the model has accepted (pre-send budget gate).

build_check_cmd: Option<String>

Shell command run after every successful write for ground-truth feedback.

safe_context: Option<u32>

Empirically safe context size used to detect likely overflow.

Implementations

impl TurnDriverConfig

pub fn new( url: impl Into<String>, model: impl Into<String>, kind: BackendKind, workspace: impl Into<String>, ) -> Self

Construct a config with sensible cowork defaults; only the endpoint, the model, the backend kind, and the workspace are required. Caveats default to Caveats::top — the downstream consumer is expected to narrow them.

Trait Implementations

impl Clone for TurnDriverConfig

fn clone(&self) -> TurnDriverConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for TurnDriverConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.