| 1 | //! Error types for the dynamic Workflow runtime. |
| 2 | |
| 3 | use thiserror::Error; |
| 4 | |
| 5 | /// Errors surfaced by [`crate::WorkflowVm::run_script`]. |
| 6 | /// |
| 7 | /// Script-visible failures (thrown JS exceptions, rejected promises, host |
| 8 | /// function errors that were not caught inside the script) all collapse into |
| 9 | /// [`WorkflowJsError::Script`] with the exception message and stack. The |
| 10 | /// remaining variants describe runtime-level failures that never reached the |
| 11 | /// script. |
| 12 | #[derive(Debug, Error)] |
| 13 | pub enum WorkflowJsError { |
| 14 | /// The QuickJS runtime or context could not be created. |
| 15 | #[error("failed to initialize the Workflow JS VM: {0}")] |
| 16 | VmInit(String), |
| 17 | /// The script threw (or a promise rejected) and nothing caught it. |
| 18 | /// Carries the exception message plus stack when available. |
| 19 | #[error("script error: {0}")] |
| 20 | Script(String), |
| 21 | /// The run was cancelled — either the caller dropped the run future or |
| 22 | /// the cooperative cancel signal fired mid-script. |
| 23 | #[error("workflow run cancelled")] |
| 24 | Cancelled, |
| 25 | /// The script completed but its return value could not be encoded as |
| 26 | /// JSON (e.g. it returned a function or a cyclic object). |
| 27 | #[error("script result is not JSON-encodable: {0}")] |
| 28 | ResultEncoding(String), |
| 29 | /// The invocation arguments could not be injected into the VM. |
| 30 | #[error("invalid workflow arguments: {0}")] |
| 31 | InvalidArgs(String), |
| 32 | /// The dedicated VM thread exited without reporting a result (panic or |
| 33 | /// spawn failure). Outstanding driver tasks are cancelled when this is |
| 34 | /// observed. |
| 35 | #[error("Workflow VM thread terminated unexpectedly: {0}")] |
| 36 | VmTerminated(String), |
| 37 | } |
| 38 | |
| 39 | /// Errors a [`crate::WorkflowDriver`] can return from `spawn_task`. |
| 40 | /// |
| 41 | /// Both variants surface inside the script as a thrown exception on the |
| 42 | /// corresponding `task()` call, so a script can `try`/`catch` an individual |
| 43 | /// rejection (admission, depth, budget) without the whole run failing. |
| 44 | #[derive(Debug, Clone, Error)] |
| 45 | pub enum DriverError { |
| 46 | /// The driver refused to spawn this task (admission cap, depth ceiling, |
| 47 | /// budget reservation failure, invalid subagent type, ...). |
| 48 | #[error("spawn rejected: {0}")] |
| 49 | Rejected(String), |
| 50 | /// The driver is gone or its channel closed; no more spawns will work. |
| 51 | #[error("driver unavailable: {0}")] |
| 52 | Unavailable(String), |
| 53 | } |
| 54 | |
| 55 | /// Why a `task()` call failed, as a stable machine kind (R9). |
| 56 | /// |
| 57 | /// The host assigns the kind at the point the failure actually happens and |
| 58 | /// ships it on the task envelope as `error_kind`; the JS prelude copies it |
| 59 | /// onto the thrown `Error` as `.kind`, and `parallel()` / `pipeline()` |
| 60 | /// classify slots from that field alone. |
| 61 | /// |
| 62 | /// This exists because the classification used to be a substring match on the |
| 63 | /// operator-facing message. A child whose own reply text contained the words |
| 64 | /// "budget exhausted" — or a script that threw `new Error("run cancelled")` — |
| 65 | /// could forge a fatal classification and abort a healthy run, and a genuine |
| 66 | /// subagent failure was indistinguishable from a plain script throw. The kind |
| 67 | /// is now data, not prose. |
| 68 | #[derive(Debug, Clone, Copy, PartialEq, Eq)] |
| 69 | pub enum TaskErrorKind { |
| 70 | /// The call never became a child: malformed options, an invalid |
| 71 | /// `responseSchema`, the workflow lifetime cap, or a driver admission |
| 72 | /// refusal. Nothing ran, so nothing was spent. |
| 73 | Admission, |
| 74 | /// The run's shared token pool is exhausted — either the pre-spawn gate |
| 75 | /// or a child that reported [`crate::TaskCompletion::BudgetExhausted`]. |
| 76 | Budget, |
| 77 | /// The run or the child was cancelled. Always fatal to the fan-out: |
| 78 | /// cancellation is the run's own deadline, never a per-slot outcome. |
| 79 | Cancelled, |
| 80 | /// The child ran and failed (error result, timeout, tool refusal, ...). |
| 81 | Agent, |
| 82 | /// The reply did not satisfy `responseSchema` and no bounded repair |
| 83 | /// remained. |
| 84 | Schema, |
| 85 | /// The driver seam broke: unavailable, or the completion channel dropped |
| 86 | /// before a terminal outcome arrived. |
| 87 | Driver, |
| 88 | } |
| 89 | |
| 90 | impl TaskErrorKind { |
| 91 | /// The wire spelling carried on the envelope and on JS `Error.kind`. |
| 92 | pub fn as_str(self) -> &'static str { |
| 93 | match self { |
| 94 | Self::Admission => "admission", |
| 95 | Self::Budget => "budget", |
| 96 | Self::Cancelled => "cancelled", |
| 97 | Self::Agent => "agent", |
| 98 | Self::Schema => "schema", |
| 99 | Self::Driver => "driver", |
| 100 | } |
| 101 | } |
| 102 | } |
| 103 | |
| 104 | /// One `task()` failure: the operator-facing message plus its typed kind. |
| 105 | #[derive(Debug, Clone, PartialEq, Eq)] |
| 106 | pub(crate) struct TaskError { |
| 107 | pub(crate) kind: TaskErrorKind, |
| 108 | pub(crate) message: String, |
| 109 | } |
| 110 | |
| 111 | impl TaskError { |
| 112 | pub(crate) fn new(kind: TaskErrorKind, message: impl Into<String>) -> Self { |
| 113 | Self { |
| 114 | kind, |
| 115 | message: message.into(), |
| 116 | } |
| 117 | } |
| 118 | } |
| 119 | |
| 120 | impl From<&DriverError> for TaskErrorKind { |
| 121 | fn from(err: &DriverError) -> Self { |
| 122 | match err { |
| 123 | // A refused spawn never produced a child; it is admission, not a |
| 124 | // failure of work that ran. |
| 125 | DriverError::Rejected(_) => Self::Admission, |
| 126 | DriverError::Unavailable(_) => Self::Driver, |
| 127 | } |
| 128 | } |
| 129 | } |
| 130 |