Struct Context 

pub struct Context {
    pub config: Config,
    pub filesystem: Client,
    pub plugin: Option<PluginPath>,
    /* private fields */
}

Fields

config: Config

filesystem: Client

plugin: Option<PluginPath>

The plugin a command is running on behalf of, when it was invoked through a plugin’s nested-command protocol. Assembled from the OBJECTIVEAI_PLUGIN_* env vars (via Config).

Implementations

impl Context

pub fn new(config: Config) -> Self

pub async fn python(&self) -> Result<&Python, Error>

The WASI python runtime, initialized on first use and memoized. First use machine-wide JIT-compiles the embedded interpreter and publishes <bin>/cache/rustpython-<hash>.cwasm under the bin lock; later uses deserialize that artifact in milliseconds. Commands that never execute python never pay either cost.

pub async fn api_client(&self) -> Result<&HttpClient, Error>

The API HttpClient, built on first use and memoized.

Address resolution: api.address from the merged (--final) config view when set, else the api spawn flow — which returns the lock-published URL of an already-running api without spawning, or starts one and waits for its lock.

pub async fn github_http_client(&self) -> Result<HttpClient, Error>

Builds an HttpClient for direct GitHub reads (agents / swarms / functions / profiles) — without spawning or contacting the ObjectiveAI API. The github_* methods talk to github.com directly and never use the base address, so we hand it a deliberately invalid one: any accidental API call through this client fails loudly instead of silently hitting a real server. x_github_authorization (from config) authenticates the GitHub requests. Built fresh per call — these commands are short-lived and don’t memoize a client.

pub async fn resolve_mcp_timeout_ms(&self) -> Result<u64, Error>

Effective MCP timeout (ms), used as BOTH the connect and per-call timeout for every MCP client this CLI drives (its streaming conduit). The merged (--final) api.mcp_timeout_ms config value, or the canonical default (60000ms) when unset.

pub async fn resolve_backoff_max_elapsed_time_ms(&self) -> Result<u64, Error>

Effective backoff max-elapsed-time (ms) — the retry budget for the CLI’s own MCP client. The merged api.backoff_max_elapsed_time_ms config value, or the canonical default (60000ms) when unset. The other exponential-backoff knobs keep their built-in defaults.

pub async fn resolve_backoff_max_elapsed_time_ms_opt( &self, ) -> Result<Option<u64>, Error>

The configured api.backoff_max_elapsed_time_ms, or None when unset — the api spawn uses this to project the backoff env onto the spawned server only when the user explicitly set it.

pub async fn resolve_mcp_timeout_ms_opt(&self) -> Result<Option<u64>, Error>

The configured api.mcp_timeout_ms, or None when unset.

The api spawn uses this to project the MCP_* timeout env onto the spawned (machine-wide, shared) server ONLY when the user explicitly set a timeout. Leaving it unset lets the api resolve it itself — <OBJECTIVEAI_DIR>/.env, then its built-in default. That fallback is behaviorally identical to passing the canonical default in production, and it’s what the test .env (which sets MCP_BACKOFF_*=0 to fast-fail real MCP errors instead of masking them with retry storms) relies on: unconditionally setting the default here would clobber that 0 on the shared api.

pub async fn viewer_client(&self) -> Result<&ViewerClient, Error>

The synchronous-response viewer client, built on first use and memoized.

Address resolution mirrors Self::api_client: viewer.address from the merged config view when set, else the viewer spawn flow. Signature: env VIEWER_SIGNATURE, else viewer.signature from the same view.

pub async fn db_client(&self) -> Result<&Pool, Error>

The db pool, connected on first use and memoized — the pool-only view of Self::db_handle.

pub async fn db_handle(&self) -> Result<&DbHandle, Error>

The db handle — pool plus the admin coordinates it was built from (address, admin user/password, database name), which crate::db::compartment needs to mint derived per-plugin connection strings. Connected on first use (ensuring the application database + schema exist) and memoized. Must NOT connect eagerly: commands like db config ... and db spawn have to work before any database exists — they’re how you bring one up in the first place.

URL resolution: when db.address is set in the merged config view, a remote-postgres URL is composed from the config db parts; otherwise the db spawn flow returns the local cluster’s lock-published postgresql:// URL (starting the objectiveai-db supervisor if needed), whose admin coordinates are parsed back out of that URL (our own postgresql://postgres:{password}@{host}:{port} shape).

pub fn reset_api_client(&mut self)

Detach this clone’s API-client cell. MUST be called after mutating config.agent_instance_hierarchy or config.mcp_session_id on a cloned Context: the memoized HttpClient snapshots those into headers, so a shared cell would either serve the base identity to this clone or poison the base context with this clone’s identity. The viewer/db cells stay shared — neither client carries identity.

Trait Implementations

impl Clone for Context

fn clone(&self) -> Context

Returns a duplicate of the value.

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

Performs copy-assignment from source.