Files
OmniRoute/tests/unit/responses-continuation-store.test.ts
Diego Rodrigues de Sa e Souza 3d4f3e4960 test(infra): retry recursive temp-dir removal instead of failing a shard on ENOTEMPTY (#11966) (#11968)
* test(infra): retry recursive temp-dir removal instead of failing a shard on ENOTEMPTY (#11966)

Two shards on release/v3.8.51 went red in one day with the same signature —
"ENOTEMPTY, Directory not empty: /tmp/omniroute-<test>-XXXXXX" — from
combo-same-provider-cascade (Unit Tests fast-path 4/4, on a PR that touches only
.github/) and auth-policy-embeddings-webfetch-7785 (the 20k-test TIA step). Both pass
alone and on re-run: the cleanup races something still writing into the directory
(SQLite WAL/-shm checkpoint, a worker, the backup) and under a loaded hosted runner
the window opens. 1154 test files do their own cleanup with
fs.rmSync(dir, { recursive: true, force: true }); 57 already asked for retries.

One-shot codemod (scripts/ad-hoc/codemod-rm-maxretries.mjs, kept for the record):
every rm / rmSync / rmdirSync option object with `recursive: true` and no
`maxRetries` gains `maxRetries: 5, retryDelay: 100` — Node itself then retries
ENOTEMPTY/EBUSY/EPERM for up to ~0.5 s before giving up. 2243 call sites in 1292
files under tests/, the shared tests/_setup/isolateDataDir.ts exit hook included.
Only the option object changes: no call site, assertion or import is touched.

Validation: prettier and ESLint (with the frozen suppressions) clean on all 1292
files; a random 20-file sample runs green (quota-redis-store hangs identically on
the untouched tree — it needs a Redis on localhost, an environment matter). The
four unit shards on this PR are the full run.

* fix(quality): let check-forgotten-sibling-tests read a 1,000-file diff

The gate shells out to `git diff` through execFileSync with Node's default 1 MB
maxBuffer; the 1,292-file codemod in this PR is the first diff large enough to
overflow it, and the gate died with `spawnSync git ENOBUFS` before comparing
anything. 64 MB is far above any real PR and costs nothing when unused.
2026-08-29 01:17:40 -03:00

275 lines
10 KiB
TypeScript

import test from "node:test";
import assert from "node:assert/strict";
import fs from "node:fs";
import os from "node:os";
import path from "node:path";
// OmniRoute-native `previous_response_id` virtualization: resolvePreviousResponseState
// resolves a response id back to the full input/output a prior call produced by
// reading the already-persisted call-log artifact, so a later request can be
// reconstructed to full history server-side without duplicating conversation
// content into a second store.
const TEST_DATA_DIR = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-responses-continuation-"));
process.env.DATA_DIR = TEST_DATA_DIR;
const core = await import("../../src/lib/db/core.ts");
const store = await import("../../src/lib/db/responsesContinuationStore.ts");
test.after(() => {
core.resetDbInstance();
fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true, maxRetries: 5, retryDelay: 100 });
});
function insertCallLog(row: {
id: string;
responseId: string | null;
apiKeyId: string | null;
detailState: string;
artifactRelPath: string | null;
}) {
const db = core.getDbInstance();
db.prepare(
`INSERT INTO call_logs
(id, timestamp, method, path, status, model, provider, account, duration,
tokens_in, tokens_out, api_key_id, detail_state, artifact_relpath, response_id)
VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`
).run(
row.id,
new Date().toISOString(),
"POST",
"/v1/responses",
200,
"gpt-5.4-pro",
"openai",
"acc1",
100,
10,
20,
row.apiKeyId,
row.detailState,
row.artifactRelPath,
row.responseId
);
}
function writeArtifact(relPath: string, pipeline: Record<string, unknown>) {
const absPath = path.join(TEST_DATA_DIR, "call_logs", relPath);
fs.mkdirSync(path.dirname(absPath), { recursive: true });
fs.writeFileSync(
absPath,
JSON.stringify({
schemaVersion: 5,
summary: {},
requestBody: null,
responseBody: null,
error: null,
pipeline,
})
);
}
test("resolvePreviousResponseState reconstructs input/output from the call-log artifact", () => {
insertCallLog({
id: "log-1",
responseId: "resp_abc",
apiKeyId: "key-1",
detailState: "ready",
artifactRelPath: "2026-01-01/log-1.json",
});
writeArtifact("2026-01-01/log-1.json", {
clientRawRequest: { body: { input: [{ type: "message", role: "user", content: "hi" }] } },
providerRequest: { body: { input: [{ type: "message", role: "user", content: "hi" }] } },
clientResponse: {
id: "resp_abc",
output: [{ type: "message", role: "assistant", content: "hello" }],
},
});
const result = store.resolvePreviousResponseState("resp_abc", "key-1");
assert.deepEqual(result, {
input: [{ type: "message", role: "user", content: "hi" }],
output: [{ type: "message", role: "assistant", content: "hello" }],
});
});
test("resolvePreviousResponseState reads output from a wrapped (streaming) clientResponse shape", () => {
// A streaming reply's clientResponse is clientPayloadCollector.build()'s output,
// which always nests the caller-supplied summary under `.summary` (see
// createStructuredSSECollector in streamPayloadCollector.ts) rather than
// carrying `output` at the top level like a non-streaming reply does. This
// must resolve exactly like the unwrapped shape above -- it was the actual
// cause of previous_response_id continuation always failing for a streaming
// Responses-API passthrough connection (fixed alongside the clientPayload
// builder gap in open-sse/utils/stream.ts).
insertCallLog({
id: "log-1-streamed",
responseId: "resp_streamed",
apiKeyId: "key-1",
detailState: "ready",
artifactRelPath: "2026-01-01/log-1-streamed.json",
});
writeArtifact("2026-01-01/log-1-streamed.json", {
clientRawRequest: { body: { input: [{ type: "message", role: "user", content: "hi" }] } },
providerRequest: { body: { input: [{ type: "message", role: "user", content: "hi" }] } },
clientResponse: {
_streamed: true,
_format: "sse-json",
_eventCount: 1,
summary: {
id: "resp_streamed",
object: "response",
output: [{ type: "message", role: "assistant", content: "hello" }],
},
},
});
const result = store.resolvePreviousResponseState("resp_streamed", "key-1");
assert.deepEqual(result, {
input: [{ type: "message", role: "user", content: "hi" }],
output: [{ type: "message", role: "assistant", content: "hello" }],
});
});
test("resolvePreviousResponseState returns null for an unknown response id", () => {
const result = store.resolvePreviousResponseState("resp_does_not_exist", "key-1");
assert.equal(result, null);
});
test("resolvePreviousResponseState never crosses tenants (scoped by api_key_id)", () => {
insertCallLog({
id: "log-2",
responseId: "resp_tenant_a",
apiKeyId: "key-a",
detailState: "ready",
artifactRelPath: "2026-01-01/log-2.json",
});
writeArtifact("2026-01-01/log-2.json", {
clientRawRequest: { body: { input: [{ role: "user", content: "secret" }] } },
providerRequest: { body: { input: [{ role: "user", content: "secret" }] } },
clientResponse: { id: "resp_tenant_a", output: [{ role: "assistant", content: "reply" }] },
});
assert.equal(store.resolvePreviousResponseState("resp_tenant_a", "key-b"), null);
assert.equal(store.resolvePreviousResponseState("resp_tenant_a", null), null);
assert.notEqual(store.resolvePreviousResponseState("resp_tenant_a", "key-a"), null);
});
test("resolvePreviousResponseState returns null when the artifact is missing on disk", () => {
insertCallLog({
id: "log-3",
responseId: "resp_missing_file",
apiKeyId: "key-1",
detailState: "ready",
artifactRelPath: "2026-01-01/does-not-exist.json",
});
assert.equal(store.resolvePreviousResponseState("resp_missing_file", "key-1"), null);
});
test("resolvePreviousResponseState fails closed when the pipeline payload was size-limit-omitted", () => {
insertCallLog({
id: "log-4",
responseId: "resp_omitted",
apiKeyId: "key-1",
detailState: "ready",
artifactRelPath: "2026-01-01/log-4.json",
});
// A size-limit-omitted payload is replaced with a placeholder string, not
// an object -- resolvePreviousResponseState must never try to reconstruct
// from it and silently drop history.
writeArtifact("2026-01-01/log-4.json", {
clientRawRequest: { body: "[omitted: call log artifact size limit exceeded]" },
clientResponse: { id: "resp_omitted", output: [] },
});
assert.equal(store.resolvePreviousResponseState("resp_omitted", "key-1"), null);
});
test("resolvePreviousResponseState resolves input from clientRawRequest when providerRequest was translated to a different upstream wire shape", () => {
// Real shape from a live auto-routed free-tier connection: OmniRoute
// translates the client's Responses-API request into Chat Completions
// (`messages`, no `input` at all) before forwarding upstream. Reading
// `input` from providerRequest.body made this permanently unresolvable --
// previous_response_not_found on every attempt -- for any connection where
// the selected upstream isn't itself a native Responses-API passthrough.
// The client's own request is always Responses-API shaped (this store only
// fires for sourceFormat === OPENAI_RESPONSES, see chat.ts), so
// clientRawRequest is the correct source regardless of upstream shape.
insertCallLog({
id: "log-6",
responseId: "resp_gen-translate-mode",
apiKeyId: "key-1",
detailState: "ready",
artifactRelPath: "2026-01-01/log-6.json",
});
writeArtifact("2026-01-01/log-6.json", {
clientRawRequest: { body: { input: [{ type: "message", role: "user", content: "hi" }] } },
providerRequest: {
body: { model: "laguna-s-2.1-free", messages: [{ role: "user", content: "hi" }] },
},
clientResponse: {
summary: {
id: "resp_gen-translate-mode",
output: [{ type: "message", role: "assistant", content: "hello" }],
},
},
});
const result = store.resolvePreviousResponseState("resp_gen-translate-mode", "key-1");
assert.deepEqual(result, {
input: [{ type: "message", role: "user", content: "hi" }],
output: [{ type: "message", role: "assistant", content: "hello" }],
});
});
test("resolvePreviousResponseState fails closed when the stored input array was log-truncated", () => {
// Real production shape: cloneBoundedChatLogPayload (chatCore/logTruncation.ts)
// and cloneBoundedForLog (utils/requestLogger.ts) both prepend an
// `_omniroute_truncated_array` sentinel in place of the items they dropped
// once a logged array exceeds their tail-item cap (~24 items) -- routine
// for any conversation that's been going a while, not an edge case. Reading
// that sentinel back as a real Responses-API item and forwarding it upstream
// produced a live 400: "input item type 'missing' cannot be represented in
// Chat Completions" -- worse than the plain cache-miss this function is
// otherwise designed to fail into.
insertCallLog({
id: "log-7",
responseId: "resp_gen-truncated-history",
apiKeyId: "key-1",
detailState: "ready",
artifactRelPath: "2026-01-01/log-7.json",
});
writeArtifact("2026-01-01/log-7.json", {
clientRawRequest: {
body: {
input: [
{ _omniroute_truncated_array: true, originalLength: 26, retainedTailItems: 24 },
{ type: "function_call_output", call_id: "call_1", output: "ok" },
],
},
},
providerRequest: { body: { input: [] } },
clientResponse: {
summary: {
id: "resp_gen-truncated-history",
output: [{ type: "message", role: "assistant", content: "hello" }],
},
},
});
assert.equal(store.resolvePreviousResponseState("resp_gen-truncated-history", "key-1"), null);
});
test("resolvePreviousResponseState returns null when detail logging was never captured for this row", () => {
insertCallLog({
id: "log-5",
responseId: "resp_no_detail",
apiKeyId: "key-1",
detailState: "none",
artifactRelPath: null,
});
assert.equal(store.resolvePreviousResponseState("resp_no_detail", "key-1"), null);
});