mirror of
https://github.com/MBombeck/HealthLog.git
synced 2026-11-18 08:02:12 +00:00
fix(documents): bound an extraction that runs inside the request
Without Prefer: respond-async the extract route ran the read with no schedule, so a twenty-page report read a page at a time, with a retry each, could make forty model calls in one HTTP request. The iPhone app sends no Prefer header and reads the document detail from the answer, so the request keeps its 200 and its body. An extraction without a run deadline now plans at most three reads (six calls): a longer document has its pages or text shared out between them, and what does not fit is reported as extractionIncomplete on the document. The background run keeps reading it a page at a time. Refs #1257
This commit is contained in:
1 parent
958d230b9f
commit
c9481b672c
5 files changed
+100
-18
No files matched your search
@@ -924,3 +924,55 @@ describe("an extraction that did not read the whole document", () => {
|
||||
});
|
||||
});
|
||||
});
|
||||
|
||||
describe("a synchronous extraction stays inside one request", () => {
|
||||
/** A stored text of many reads' worth: forty lab lines per part. */
|
||||
async function longIndexedDocument(userId: string) {
|
||||
const document = await storeDocument(userId);
|
||||
const { upsertContentIndex } =
|
||||
await import("@/lib/documents/content-index");
|
||||
const lines = Array.from(
|
||||
{ length: 1200 },
|
||||
(_, i) => `Analyte ${i} ${(i % 90) + 10}.0 mg/dL reference 10-100`,
|
||||
);
|
||||
await upsertContentIndex({
|
||||
userId,
|
||||
documentId: document.id,
|
||||
text: lines.join("\n"),
|
||||
source: "text-ocr",
|
||||
providerType: null,
|
||||
});
|
||||
return document;
|
||||
}
|
||||
|
||||
it("reads a long document in a few reads and says what it left out", async () => {
|
||||
const user = await makeUser();
|
||||
await grantConsent(user.id);
|
||||
const document = await longIndexedDocument(user.id);
|
||||
await signIn(user.id);
|
||||
|
||||
const res = await postStoredExtract(document.id, false);
|
||||
expect(res.status).toBe(200);
|
||||
const { SYNC_EXTRACT_MAX_READS } =
|
||||
await import("@/lib/documents/ai-runs/extract-run");
|
||||
const args = vi.mocked(runInboundExtraction).mock.calls[0][0];
|
||||
expect(args.chunks.length).toBeGreaterThan(0);
|
||||
expect(args.chunks.length).toBeLessThanOrEqual(SYNC_EXTRACT_MAX_READS);
|
||||
expect(res.body.data).toMatchObject({ extractionIncomplete: true });
|
||||
});
|
||||
|
||||
it("reads it in full in the background", async () => {
|
||||
const user = await makeUser();
|
||||
await grantConsent(user.id);
|
||||
const document = await longIndexedDocument(user.id);
|
||||
await signIn(user.id);
|
||||
|
||||
const queued = await postStoredExtract(document.id, true);
|
||||
const { runId } = queued.body.data as { runId: string };
|
||||
expect(await work(runId)).toBe("succeeded");
|
||||
const { SYNC_EXTRACT_MAX_READS } =
|
||||
await import("@/lib/documents/ai-runs/extract-run");
|
||||
const args = vi.mocked(runInboundExtraction).mock.calls[0][0];
|
||||
expect(args.chunks.length).toBeGreaterThan(SYNC_EXTRACT_MAX_READS);
|
||||
});
|
||||
});
|
||||
Reference in new issue
Block a user