Alan VoandN0zoM1z0
5a67c7fc54
* fix(hopper): diagnose MCP provider failures with stage, state, and request context * fix(hopper): retain health at the actual provider lifecycle boundary * test(hopper): distinguish socket loss from later observed process exit * test(hopper): cover observed exit separately from socket-first replies --------- Co-authored-by: N0zoM1z0 <[email protected]>
89 lines
2.6 KiB
TypeScript
89 lines
2.6 KiB
TypeScript
/** Caller-visible operational state for one provider bridge. */
|
|
export const PROVIDER_OPERATION_STATES = [
|
|
"idle",
|
|
"busy",
|
|
"exited",
|
|
"unreachable",
|
|
"not_started",
|
|
"unknown",
|
|
] as const;
|
|
|
|
/** Operational state REA can support with evidence, or unknown. */
|
|
export type ProviderOperationState = (typeof PROVIDER_OPERATION_STATES)[number];
|
|
|
|
/** States a Hopper process failure can establish. */
|
|
export type HopperProcessProviderState = "exited" | "unreachable" | "unknown";
|
|
|
|
/** What the caller should do next for this provider state. */
|
|
export const PROVIDER_RETRY_ACTIONS = [
|
|
"retry",
|
|
"wait",
|
|
"restart_provider",
|
|
"unknown",
|
|
] as const;
|
|
|
|
/** Recovery choice for a provider operation failure. */
|
|
export type ProviderRetryAction = (typeof PROVIDER_RETRY_ACTIONS)[number];
|
|
|
|
/** Stage that was in progress when the provider operation failed. */
|
|
export const PROVIDER_FAILURE_STAGES = [
|
|
"launch",
|
|
"connection",
|
|
"analysis",
|
|
"decompilation",
|
|
"unknown",
|
|
] as const;
|
|
|
|
/** Known provider-operation stage, including an explicit unknown. */
|
|
export type ProviderFailureStage = (typeof PROVIDER_FAILURE_STAGES)[number];
|
|
|
|
/** One provider request retained with a failure or an in-flight operation. */
|
|
export interface ProviderOperationRequest {
|
|
readonly requestId: number;
|
|
readonly operation: string;
|
|
readonly stage: ProviderFailureStage;
|
|
}
|
|
|
|
/** Sanitized provider health a session can show after the request returns. */
|
|
export interface ProviderOperationHealth {
|
|
readonly state: ProviderOperationState;
|
|
readonly stage: ProviderFailureStage | null;
|
|
readonly retryAction: ProviderRetryAction | null;
|
|
readonly exitCode: number | null;
|
|
readonly requests: readonly ProviderOperationRequest[];
|
|
}
|
|
|
|
/**
|
|
* Choose a recovery action from the observed state.
|
|
* Startup failures keep their existing install or configuration guidance.
|
|
*/
|
|
export const providerRetryAction = (
|
|
state: ProviderOperationState,
|
|
startupFailure = false,
|
|
): ProviderRetryAction | null => {
|
|
if (state === "idle") return null;
|
|
if (startupFailure && state === "exited") return "retry";
|
|
switch (state) {
|
|
case "busy":
|
|
return "wait";
|
|
case "exited":
|
|
case "unreachable":
|
|
return "restart_provider";
|
|
case "not_started":
|
|
return "retry";
|
|
case "unknown":
|
|
return "unknown";
|
|
}
|
|
};
|
|
|
|
/** Use one stage when every retained request agrees, otherwise unknown. */
|
|
export const sharedProviderFailureStage = (
|
|
requests: readonly ProviderOperationRequest[],
|
|
): ProviderFailureStage | null => {
|
|
const first = requests[0]?.stage;
|
|
if (first === undefined) return null;
|
|
return requests.every((request) => request.stage === first)
|
|
? first
|
|
: "unknown";
|
|
};
|