core: track retry attempts with detailed error context on assistant entries

users can now see when transient failures occur during assistant responses,
such as rate limits or provider overloads, giving visibility into what
issues were encountered and automatically resolved before the final response
This commit is contained in:
Dax Raad 2026-04-18 10:35:25 -04:00
parent 95edbc0ae6
commit 882b8e1e75
4 changed files with 130 additions and 4 deletions

View file

@ -1,4 +1,4 @@
import { castDraft, produce, type WritableDraft } from "immer"
import { produce, type WritableDraft } from "immer"
import { SessionEvent } from "./session-event"
import { SessionEntry } from "./session-entry"
@ -235,7 +235,15 @@ export function stepWith<Result>(adapter: Adapter<Result>, event: SessionEvent.E
)
}
},
retried: () => {},
retried: (event) => {
if (currentAssistant) {
adapter.updateAssistant(
produce(currentAssistant, (draft) => {
draft.retries = [...(draft.retries ?? []), SessionEntry.AssistantRetry.fromEvent(event)]
}),
)
}
},
compacted: (event) => {
adapter.appendEntry(SessionEntry.Compaction.fromEvent(event))
},

View file

@ -104,6 +104,24 @@ export class AssistantReasoning extends Schema.Class<AssistantReasoning>("Sessio
text: Schema.String,
}) {}
export class AssistantRetry extends Schema.Class<AssistantRetry>("Session.Entry.Assistant.Retry")({
attempt: Schema.Number,
error: SessionEvent.RetryError,
time: Schema.Struct({
created: Schema.DateTimeUtc,
}),
}) {
static fromEvent(event: SessionEvent.Retried) {
return new AssistantRetry({
attempt: event.attempt,
error: event.error,
time: {
created: event.timestamp,
},
})
}
}
export const AssistantContent = Schema.Union([AssistantText, AssistantReasoning, AssistantTool]).pipe(
Schema.toTaggedUnion("type"),
)
@ -113,6 +131,7 @@ export class Assistant extends Schema.Class<Assistant>("Session.Entry.Assistant"
...Base,
type: Schema.Literal("assistant"),
content: AssistantContent.pipe(Schema.Array),
retries: AssistantRetry.pipe(Schema.Array, Schema.optional),
cost: Schema.Number.pipe(Schema.optional),
tokens: Schema.Struct({
input: Schema.Number,
@ -137,6 +156,7 @@ export class Assistant extends Schema.Class<Assistant>("Session.Entry.Assistant"
created: event.timestamp,
},
content: [],
retries: [],
})
}
}

View file

@ -53,6 +53,15 @@ export namespace SessionEvent {
source: Source.pipe(Schema.optional),
}) {}
export class RetryError extends Schema.Class<RetryError>("Session.Event.Retry.Error")({
message: Schema.String,
statusCode: Schema.Number.pipe(Schema.optional),
isRetryable: Schema.Boolean,
responseHeaders: Schema.Record(Schema.String, Schema.String).pipe(Schema.optional),
responseBody: Schema.String.pipe(Schema.optional),
metadata: Schema.Record(Schema.String, Schema.String).pipe(Schema.optional),
}) {}
export class Prompt extends Schema.Class<Prompt>("Session.Event.Prompt")({
...Base,
type: Schema.Literal("prompt"),
@ -386,14 +395,16 @@ export namespace SessionEvent {
export class Retried extends Schema.Class<Retried>("Session.Event.Retried")({
...Base,
type: Schema.Literal("retried"),
error: Schema.String,
attempt: Schema.Number,
error: RetryError,
}) {
static create(input: BaseInput & { error: string }) {
static create(input: BaseInput & { attempt: number; error: RetryError }) {
return new Retried({
id: input.id ?? ID.create(),
type: "retried",
timestamp: input.timestamp ?? DateTime.makeUnsafe(Date.now()),
metadata: input.metadata,
attempt: input.attempt,
error: input.error,
})
}

View file

@ -27,6 +27,24 @@ function assistant() {
type: "assistant",
time: { created: time(0) },
content: [],
retries: [],
})
}
function retryError(message: string) {
return new SessionEvent.RetryError({
message,
isRetryable: true,
})
}
function retry(attempt: number, message: string, created: number) {
return new SessionEntry.AssistantRetry({
attempt,
error: retryError(message),
time: {
created: time(created),
},
})
}
@ -78,6 +96,12 @@ function tool(state: SessionEntryStepper.MemoryState, callID: string) {
return tools(state).find((x) => x.callID === callID)
}
function retriesOf(state: SessionEntryStepper.MemoryState) {
const entry = last(state)
if (!entry) return []
return entry.retries ?? []
}
function adapterStore() {
return {
committed: [] as SessionEntry.Entry[],
@ -168,6 +192,33 @@ describe("session-entry-stepper", () => {
])
expect(store.committed[0].time.completed).toEqual(time(7))
})
test("aggregates retry events onto the current assistant", () => {
const store = adapterStore()
store.committed.push(assistant())
SessionEntryStepper.stepWith(
adapterFor(store),
SessionEvent.Retried.create({
attempt: 1,
error: retryError("rate limited"),
timestamp: time(1),
}),
)
SessionEntryStepper.stepWith(
adapterFor(store),
SessionEvent.Retried.create({
attempt: 2,
error: retryError("provider overloaded"),
timestamp: time(2),
}),
)
expect(store.committed[0]?.type).toBe("assistant")
if (store.committed[0]?.type !== "assistant") return
expect(store.committed[0].retries).toEqual([retry(1, "rate limited", 1), retry(2, "provider overloaded", 2)])
})
})
describe("memory", () => {
@ -231,6 +282,21 @@ describe("session-entry-stepper", () => {
expect(reasons(state)).toEqual([{ type: "reasoning", text: "final" }])
})
test("stepWith through memory records retries", () => {
const state = active()
SessionEntryStepper.stepWith(
SessionEntryStepper.memory(state),
SessionEvent.Retried.create({
attempt: 1,
error: retryError("rate limited"),
timestamp: time(1),
}),
)
expect(retriesOf(state)).toEqual([retry(1, "rate limited", 1)])
})
})
describe("step", () => {
@ -481,6 +547,27 @@ describe("session-entry-stepper", () => {
})
})
test("records retries on the pending assistant", () => {
const next = run(
[
SessionEvent.Retried.create({
attempt: 1,
error: retryError("rate limited"),
timestamp: time(1),
}),
SessionEvent.Retried.create({
attempt: 2,
error: retryError("provider overloaded"),
timestamp: time(2),
}),
],
active(),
)
expect(retriesOf(next)).toEqual([retry(1, "rate limited", 1), retry(2, "provider overloaded", 2)])
})
})
describe("known reducer gaps", () => {
test("prompt appends immutably when no assistant is pending", () => {
FastCheck.assert(