Vito
0095db8cbc
* feat(inspect): bounded browser PNG reader matching the server's
The Inspect page will read a chosen file entirely in the browser, and the
record it shows must equal what Images shows for the same bytes. This adds
to web/js/png-metadata.js (loaded by the legacy pages as before, with
parsePngTextChunks unchanged):
- readPngMetadataFile(file, {signal, prefix, yieldControl, decompress}):
a port of utils/prompt_inspector.py's _png_metadata, _PngRecords,
_png_text, _inflate and _batch_normalized. It slices only chunk headers,
80-byte keyword prefixes and the retained prompt, workflow, parameters
and comfyassets_prompt_pairs records, CRC-checks what it keeps, applies
the same budgets (2 MiB shared role budget encoded and decoded, 4 MiB
workflow that becomes limited, 8 MiB and 4 MiB duplicate budgets, 4 MiB
duplicate decoded parameters, 10,000 chunks, the batch-index rule for
parameters copies) and compares duplicate chunks byte for byte. Every
range is checked against file.size before slicing and charged against a
19,794,389-byte ceiling, so a multi-gigabyte PNG costs only its metadata.
Compressed text inflates through DecompressionStream with one 256-byte
write in flight and the output capped at the record's allowance; an
AbortSignal rejects the read with an AbortError and leaves no unhandled
rejection; yieldControl runs about every 8 ms of work.
- classifyPrefix(bytes): png, jpeg, webp, iso-bmff, matroska, avi, asf or
null from the first 32 bytes.
- readImageHeader(file, kind): JPEG (SOFn walk as Pillow) and WebP (VP8,
VP8L, VP8X) dimensions from at most 64 KiB.
tests/test_local_inspect_parity.py drives tests/js/support/local-inspect-
oracle.mjs over the committed and test-time inspection parity corpus plus
compressed, oversized, Latin-1, batch-duplicate and all-four-record shapes,
and asserts records, record formats, dimensions and readability equal
read_inspection_file on each file; the two JPEG/WebP fixtures are checked
against Pillow. tests/js/png-local-reader.test.js pins each server rule,
the sparse 3 GiB read, the prefix reuse, inflation back-pressure and abort.
tests/fixtures/local-inspect/tiny.{jpg,webp} are CC0 FFmpeg test patterns.
* feat(inspect): bounded browser video tag readers
Add web/js/workspace/local-video.mjs: ISO-BMFF (MP4/MOV/M4V) and Matroska
(MKV/WebM) walkers that read only the container tags the server asks ffprobe
for, slicing box and element headers, carrier names and retained values within
the server's budgets and never a byte inside mdat or a Cluster; a port of
utils/video_metadata.select_video_carrier with the record budgets of
read_inspection_file; and a byte-exact json.loads/json.dumps round trip so
envelope objects reserialize as the server spells them.
Every naming rule mirrors ffprobe 9.0.2 on hand-built files: legacy ilst atoms,
mdta keys, joined repeats, data types, Matroska language and default handling,
and Tags behind a Cluster found only through a SeekHead.
Tests: node units over byte-exact container builders and the committed
fixtures; a Python parity test comparing the local carrier with the server's
selection from the same tag maps (mocked probe) and from the same files (real
ffprobe, including 1 MiB + 2.5 MiB per-key records); an ffprobe
characterization of the recorded fixture maps.
* feat(inspect): server-identical local projection and payload
Port utils/inspection_response.py to the browser: a strict parse with Python's
Decimal spelling, duplicate-key rejection, NaN/Infinity as null and integers
past 2^53-1 as text; the prompt, workflow and carrier projection with the
depth-64 and 20,000-node ceilings; _metadata; and the 8 MiB omission steps
(workflow graph, then prompt graph, then all raw) measured as UTF-8 bytes.
local-inspection.mjs builds the inspection_version 1 wire for one chosen File
from the Task 1 and Task 2 readers (observation selected-local-file, interim
unresolved roles, unavailable library), fits it to the response budget, and
exposes localFileAdapter for the shared controller, inspector and viewer:
handles are session-owned by a private token, so a tampered or hand-made
descriptor is rejected and blob: URLs are never resolved through the API
client. projectInspectionPayload admits selected-local-file only for a
local-file identity.
A Python oracle test compares projectMetadata with _metadata byte for byte over
the parity PNG corpus, the committed video fixtures and tag maps, 75 committed
adversarial record sets (projection-cases.json) and node-ceiling sets built at
test time.
* test(inspect): import oracle modules by file URL for Windows
The local-inspect oracle imported product modules with path.join(...), which
Node parses as a URL scheme on Windows (D:\... -> ERR_UNSUPPORTED_ESM_URL_SCHEME),
failing every tests/test_local_inspect_parity.py oracle test on the Windows CI
job. Every dynamic import in tests/js/support now goes through moduleUrl, a
pathToFileURL wrapper, and a node unit proves the Windows form on any host and
scans the oracles so a path import cannot return.
* fix(inspect): read video tags as ffprobe reports them
Same bytes now give the server's record in five shapes where the local
reader was stricter or kept different tags (PR #277 final review I1, I2,
M1, M2, M4):
- Invalid UTF-8 in a tag value is replaced as ffprobe's default
string_validation=replace does (av_utf8_decode, one U+FFFD per failing
decode step), instead of failing the read. Key names and TagNames that
do not decode are no carrier and are skipped.
- An ilst item whose keys index is 0 or out of range is skipped, as
ffmpeg does, instead of failing the read.
- Only the first keys box is read, as ffmpeg does.
- A first per-key workflow over 4 MiB keeps its place in the carrier
choice as an unread marker, so the server's carrier (per-key, workflow
limited) is chosen; when the server's first probe would overflow, the
re-probe leaves it out and its own bound applies.
- An envelope integer of more than 4300 digits fails the Python json
port, so that envelope is skipped as CPython's json.loads skips it.
Each shape was checked against ffprobe 9.0.2 and read_inspection_file;
the real-ffprobe parity test now builds them, and a 600-case random-byte
comparison of the replacement against ffprobe matched.
* perf(inspect): measure the local response size natively and yield between steps
fitInspectionPayload counted the UTF-8 size of an 8-12 MB wire with a
per-character loop and ran every omission step on the main thread in one
turn (PR #277 final review M7). The size is now counted by TextEncoder
encodeInto through one bounded scratch buffer, and each re-measure
follows a yieldControl turn, with an abort signal checked between steps.
The fit is async; the fitted payload is the same as before.
* fix(inspect): hold video tags as ffmpeg's dictionary does, decode Mac Roman text
Three more same-bytes divergences, found while verifying the review
fixes against ffprobe 9.0.2 and read_inspection_file:
- ffmpeg keeps one tag per ASCII-folded name under the spelling set
last (ISO-BMFF joins the values with ";", Matroska keeps the last), so
PROMPT and Prompt are one record on the server; the local reader kept
both and failed the read as a duplicate. A legacy comment atom and a
later COMMENT key likewise leave no exact-case envelope.
- ffprobe's format_tags filter folds ASCII case only, so a name that is
"workflow" only under Unicode case rules is never reported; the local
reader took it.
- ISO-BMFF data types 0 and 3 are Mac Roman text to ffmpeg, written into
twice the input size (a long value is cut); the local reader decoded
type 0 as UTF-8 and skipped type 3.
The real-ffprobe parity test builds each shape.
* docs(testing): list the local Inspect divergences; raise touched coverage floors
TESTING.md now tracks the local Inspect (#249 PR L1) divergences that
remain after the review fixes, each checked on a host with ffprobe 9.0.2:
media kind by bytes versus extension (parity is for the same bytes with a
matching extension), JPEG dimensions past the 64 KiB header read, classic
QuickTime udta text atoms, raw versus escaped tag bounds, the 10,000
element ceiling, unknown EBML DocTypes, the JSON depth approximation and
the C _decimal assumption.
coverage:update raised the branch floors of the three local Inspect
modules this PR touches; the unrelated activity-scan raise it proposed is
left at main's values.
250 lines
20 KiB
JavaScript
250 lines
20 KiB
JavaScript
// web/js/workspace/local-projection.mjs ports utils/inspection_response.py: the
|
||
// strict Decimal/duplicate-key parse, the prompt/workflow/carrier projection, the
|
||
// depth and node ceilings, _metadata and the 8 MiB omission steps. These units pin
|
||
// the rules the plan names; tests/test_local_inspect_parity.py proves byte parity
|
||
// with the server over the committed corpus.
|
||
import test from "node:test";
|
||
import assert from "node:assert/strict";
|
||
import fs from "node:fs";
|
||
import path from "node:path";
|
||
import { fileURLToPath } from "node:url";
|
||
|
||
const here = path.dirname(fileURLToPath(import.meta.url));
|
||
const m = await import("../../web/js/workspace/local-projection.mjs").catch((error) => {
|
||
if (error.code !== "ERR_MODULE_NOT_FOUND") throw error;
|
||
return {};
|
||
});
|
||
const CASES = JSON.parse(fs.readFileSync(path.join(here, "..", "fixtures", "local-inspect", "projection-cases.json"), "utf8")).cases;
|
||
|
||
const plain = (value) => JSON.parse(JSON.stringify(value));
|
||
const bytes = (value) => Buffer.byteLength(JSON.stringify(value), "utf8");
|
||
const graphOf = (text, kind = "workflow") => plain(m.projectGraph(text, kind));
|
||
const nested = (depth, close = "]", open = "[") => open.repeat(depth) + close.repeat(depth);
|
||
const nodes = (count) => `{${Array.from({ length: count }, (_, i) => `"${i}":{"class_type":"A","inputs":{}}`).join(",")}}`;
|
||
|
||
test("contract", () => {
|
||
for (const name of ["projectGraph", "projectMetadata", "fitInspectionPayload", "inspectionBytes", "parseStrictJson"]) {
|
||
assert.equal(typeof m[name], "function", `${name} is exported`);
|
||
}
|
||
assert.equal(m.MAX_RESPONSE_BYTES, 8388608);
|
||
assert.equal(m.MAX_PROJECTION_DEPTH, 64);
|
||
assert.equal(m.GRAPH_NODE_CEILING, 20000);
|
||
});
|
||
|
||
test("projection spells numbers like the server", () => {
|
||
// literal -> Python str(Decimal(literal)) (or the int when it is a safe integer)
|
||
const table = [
|
||
["1e3", "1E+3"], ["1e0", "1"], ["-0.0", "-0.0"], ["6.50", "6.50"], ["0.0100", "0.0100"], ["1.0e2", "1.0E+2"],
|
||
["0e5", "0E+5"], ["0.0000001", "1E-7"], ["0.000001", "0.000001"], ["1E+3", "1E+3"], ["1e-0", "1"], ["0.5e1", "5"],
|
||
["0.05e1", "0.5"], ["0.0e0", "0.0"], ["1.5e+300", "1.5E+300"], ["123e-8", "0.00000123"], ["2.5e-3", "0.0025"],
|
||
["1.0", "1.0"], ["-1.50", "-1.50"], ["0.0", "0.0"], ["100.0", "100.0"], ["1234567e-6", "1.234567"], ["0.5E-6", "5E-7"],
|
||
["0.5E-5", "0.000005"], ["1e22", "1E+22"], ["0.000000", "0.000000"], ["-0.000000e-3", "-0E-9"], ["7.0e-6", "0.0000070"],
|
||
["7.0e-7", "7.0E-7"], ["12345678901234567890.123", "12345678901234567890.123"], ["9007199254740992.0", "9007199254740992.0"],
|
||
["1e999999999999999999", "1E+999999999999999999"], ["1e-1999999999999999997", "1E-1999999999999999997"],
|
||
["0.0e1000000000000000000", "0E+999999999999999999"], ["1.5e999999999999999999", "1.5E+999999999999999999"],
|
||
["0.1e-1999999999999999996", "1E-1999999999999999997"],
|
||
["9007199254740991", 9007199254740991], ["-9007199254740991", -9007199254740991], ["9007199254740992", "9007199254740992"],
|
||
["-9007199254740992", "-9007199254740992"], ["10000000000000000", "10000000000000000"], ["123456789012345678901234567890", "123456789012345678901234567890"],
|
||
["0", 0], ["-0", 0], ["42", 42], ["-7", -7],
|
||
];
|
||
for (const [literal, expected] of table) {
|
||
const graph = graphOf(`{"n":${literal}}`);
|
||
assert.equal(graph.status, "parsed", literal);
|
||
assert.deepEqual(graph.value, { n: expected }, literal);
|
||
}
|
||
assert.deepEqual(graphOf('{"n":[NaN,Infinity,-Infinity]}').value, { n: [null, null, null] });
|
||
// A 4300-digit integer is Python's int() limit; one more digit fails the parse.
|
||
assert.deepEqual(graphOf(`{"n":${"1".repeat(4300)}}`).value, { n: "1".repeat(4300) });
|
||
assert.equal(graphOf(`{"n":${"1".repeat(4301)}}`).status, "malformed");
|
||
assert.equal(graphOf(`{"n":-${"1".repeat(4301)}}`).status, "malformed");
|
||
// Exact Decimal construction fails past the C library's exponent range.
|
||
for (const literal of ["1e1000000000000000000", "15e999999999999999999", "1e-1999999999999999998", "1.5e-1999999999999999997",
|
||
"0e1000000000000000000", "1e9223372036854775808", "0.0e1000000000000000001", "1e99999999999999999999999999"]) {
|
||
assert.equal(graphOf(`{"n":${literal}}`).status, "malformed", literal);
|
||
}
|
||
// Escapes: surrogate pairs join, lone surrogates stay, control characters decode.
|
||
const text = graphOf('{"s":"\\u00E9\\uD83D\\uDE00\\uD800\\u0000\\u001f\\/\\\\\\"\\b\\f\\n\\r\\t\\uD800\\u0041"}').value.s;
|
||
assert.equal(text, "é\u{1f600}\ud800\u0000\u001f/\\\"\b\f\n\r\t\ud800A");
|
||
assert.equal(JSON.stringify(m.projectGraph('{"s":"\\ud800"}', "workflow").value), '{"s":"\\ud800"}');
|
||
});
|
||
|
||
test("projection rejects what Python's json rejects and nothing else", () => {
|
||
const rejected = ["{}", "\f1", " 1", '{"a":1,}', "[1,]", "{'a':1}", '{"a"}', "[1 2]", "tru", "nul", "-NaN", "1.", ".5", "01", "+1",
|
||
"1e", "-", '"\\x"', '"abc', "", " ", '"\u0001"', '{"a":1}x', '"\\u12"', '{"a":1,"a":2}', '{"a":{"b":1,"b":1}}', "[1,[2,{\"k\":1,\"k\":2}]]",
|
||
'{"a":1 "b":2}', "[,1]", "{,}", '{"a":}', "[1}", "{1:2}", "
[]", "[]\u000b", "0x10", "1_000", "--1", "1e+", "1E-", '"\\uD800\\uDC"'];
|
||
for (const text of rejected) assert.equal(graphOf(text).status, "malformed", JSON.stringify(text));
|
||
for (const text of [" \t\r\n{}\n\t ", '{"a":[1,{"b":null}],"c":"d"}', '{"\\u0000":1}', '{"a":1,"A":1}', '{"":[]}']) {
|
||
assert.equal(graphOf(text).status, "parsed", JSON.stringify(text));
|
||
}
|
||
// Top-level values that are not objects parse but are not graphs.
|
||
for (const text of ["[]", '"x"', "1", "null", "true", "NaN", "1.5", "-Infinity"]) assert.equal(graphOf(text).status, "malformed", text);
|
||
assert.equal(graphOf(undefined).status, "malformed");
|
||
assert.equal(graphOf(null).status, "malformed");
|
||
});
|
||
|
||
test("projection limits", () => {
|
||
// A workflow's root is depth 1; the 64th nested container is kept, the 65th is limited.
|
||
assert.equal(graphOf(`{"a":${nested(63)}}`).status, "parsed");
|
||
assert.equal(graphOf(`{"a":${nested(64)}}`).status, "limited");
|
||
// Prompt: root 1, node 2, inputs 3; a value chain reaching depth 64 is kept, 65 is limited.
|
||
assert.equal(graphOf(`{"1":{"class_type":"A","inputs":{"v":${nested(61)}}}}`, "prompt").status, "parsed");
|
||
assert.equal(graphOf(`{"1":{"class_type":"A","inputs":{"v":${nested(62)}}}}`, "prompt").status, "limited");
|
||
// Dropped node keys are still validated for depth.
|
||
assert.equal(graphOf(`{"1":{"class_type":"A","inputs":{},"_meta":{"v":${nested(61)}}}}`, "prompt").status, "parsed");
|
||
assert.equal(graphOf(`{"1":{"class_type":"A","inputs":{},"_meta":{"v":${nested(62)}}}}`, "prompt").status, "limited");
|
||
// A workflow is normalized before its shape is checked, so a deep non-object is limited; a
|
||
// prompt checks its root first. Depth never overflows the parser (100,000 nested arrays).
|
||
assert.equal(graphOf(nested(10)).status, "malformed");
|
||
assert.equal(graphOf(nested(5000)).status, "limited");
|
||
assert.equal(graphOf(nested(5000), "prompt").status, "malformed");
|
||
assert.equal(graphOf(`{"a":${nested(100000)}}`).status, "limited");
|
||
// Node ceilings: 20,000 API nodes parse, 20,001 are limited.
|
||
assert.equal(graphOf(nodes(20000), "prompt").status, "parsed");
|
||
assert.equal(graphOf(nodes(20001), "prompt").status, "limited");
|
||
const list = (count) => `[${Array.from({ length: count }, () => "{}").join(",")}]`;
|
||
assert.equal(graphOf(`{"nodes":${list(20000)}}`).status, "parsed");
|
||
assert.equal(graphOf(`{"nodes":${list(20001)}}`).status, "limited");
|
||
assert.equal(graphOf(`{"nodes":${list(10001)},"definitions":{"subgraphs":[{"nodes":${list(10000)}},5,{"nodes":3}]}}`).status, "limited");
|
||
assert.equal(graphOf(`{"nodes":${list(10000)},"definitions":{"subgraphs":[{"nodes":${list(10000)}}]}}`).status, "parsed");
|
||
assert.equal(graphOf(`{"nodes":{"${"x".repeat(5)}":1},"definitions":[]}`).status, "parsed");
|
||
// Duplicate keys anywhere are malformed, including inside dropped node keys.
|
||
assert.equal(graphOf('{"1":{"class_type":"A","inputs":{},"_meta":{"t":1,"t":2}}}', "prompt").status, "malformed");
|
||
assert.equal(graphOf('{"1":{"class_type":"A","inputs":{"a":1,"a":1}}}', "prompt").status, "malformed");
|
||
});
|
||
|
||
test("prompt nodes keep only class_type and inputs on null-prototype objects", () => {
|
||
const text = '{"3":{"class_type":"KSampler","inputs":{"seed":9007199254740993,"cfg":8.0,"n":NaN},"_meta":{"title":"x"},"is_changed":[1]},"4":5,"5":[1,2.50],"__proto__":{"class_type":"P","inputs":{}},"constructor":"c"}';
|
||
const graph = m.projectGraph(text, "prompt");
|
||
assert.equal(graph.status, "parsed");
|
||
assert.equal(Object.getPrototypeOf(graph.value), null);
|
||
assert.equal(Object.getPrototypeOf(graph.value["3"]), null);
|
||
assert.equal(Object.getPrototypeOf(graph.value["3"].inputs), null);
|
||
// JSON.parse, like the projection, makes "__proto__" an own key; an object literal would not.
|
||
assert.deepEqual(plain(graph.value), JSON.parse('{"3":{"class_type":"KSampler","inputs":{"seed":"9007199254740993","cfg":"8.0","n":null}},"4":5,"5":[1,"2.50"],"__proto__":{"class_type":"P","inputs":{}},"constructor":"c"}'));
|
||
assert.ok(Object.hasOwn(graph.value, "__proto__"));
|
||
assert.deepEqual(plain(graph.value["__proto__"]), { class_type: "P", inputs: {} });
|
||
assert.equal(graph.value.constructor, "c");
|
||
for (const text of ["[]", '"3"', "5", "null", '{"nodes":[]}x']) assert.equal(m.projectGraph(text, "prompt").status, text === '{"nodes":[]}x' ? "malformed" : "malformed", text);
|
||
// A workflow keeps everything, normalized.
|
||
assert.deepEqual(graphOf('{"nodes":[{"id":1,"widgets_values":[7,"fixed",8.0,1e3]}],"links":[],"version":0.4}').value,
|
||
{ nodes: [{ id: 1, widgets_values: [7, "fixed", "8.0", "1E+3"] }], links: [], version: "0.4" });
|
||
});
|
||
|
||
test("projectMetadata mirrors _metadata for records, formats and carriers", () => {
|
||
const prompt = '{"1":{"class_type":"A","inputs":{"x":1.0}}}', workflow = '{"nodes":[],"version":0.4}';
|
||
const read = (records, record_formats, carrierText = null, limited = []) => ({ records, record_formats, carrierText, limited });
|
||
assert.deepEqual(plain(m.projectMetadata(null)), { status: "unreadable", omitted_reason: null, raw: {}, record_formats: {}, parameters: null,
|
||
prompt_graph: { status: "absent", value: null }, workflow_graph: { status: "absent", value: null } });
|
||
assert.deepEqual(plain(m.projectMetadata(read({}, {}))), { status: "absent", omitted_reason: null, raw: {}, record_formats: {}, parameters: null,
|
||
prompt_graph: { status: "absent", value: null }, workflow_graph: { status: "absent", value: null } });
|
||
const full = plain(m.projectMetadata(read({ prompt, workflow, parameters: "cat\nSteps: 1", comfyassets_prompt_pairs: "{}", other: "x" },
|
||
{ prompt: "original-text", workflow: "original-text", parameters: "original-text", comfyassets_prompt_pairs: "original-text", other: "original-text" })));
|
||
assert.deepEqual(full, { status: "available", omitted_reason: null,
|
||
raw: { prompt, workflow, parameters: "cat\nSteps: 1", comfyassets_prompt_pairs: "{}" },
|
||
record_formats: { prompt: "original-text", workflow: "original-text", parameters: "original-text", comfyassets_prompt_pairs: "original-text" },
|
||
parameters: "cat\nSteps: 1", prompt_graph: { status: "parsed", value: { 1: { class_type: "A", inputs: { x: "1.0" } } } },
|
||
workflow_graph: { status: "parsed", value: { nodes: [], version: "0.4" } } });
|
||
// limited records lose their text; invalid is malformed; reserialized without a carrier is limited.
|
||
const limited = plain(m.projectMetadata(read({ prompt, workflow }, { prompt: "original-text" }, null, ["workflow"])));
|
||
assert.deepEqual(limited.raw, { prompt });
|
||
assert.deepEqual(limited.record_formats, { prompt: "original-text", workflow: "limited" });
|
||
assert.equal(limited.workflow_graph.status, "limited");
|
||
assert.equal(plain(m.projectMetadata(read({ prompt }, { prompt: "original-text", workflow: "invalid" }))).workflow_graph.status, "malformed");
|
||
assert.equal(plain(m.projectMetadata(read({ workflow }, { workflow: "reserialized-json" }))).workflow_graph.status, "limited");
|
||
assert.equal(plain(m.projectMetadata(read({ prompt: "{" }, { prompt: "original-text" }))).prompt_graph.status, "malformed");
|
||
// Unknown formats are dropped from record_formats but the raw text still projects.
|
||
const odd = plain(m.projectMetadata(read({ prompt }, { prompt: "weird" })));
|
||
assert.deepEqual(odd.record_formats, {});
|
||
assert.equal(odd.prompt_graph.status, "parsed");
|
||
// A non-text record vanishes from raw and its named graph is malformed.
|
||
assert.equal(plain(m.projectMetadata(read({ prompt: 5 }, { prompt: "original-text" }))).prompt_graph.status, "malformed");
|
||
// Carriers: strings re-parse by kind, objects were projected at depth 2; a bad carrier fails every present graph.
|
||
const envelope = JSON.stringify({ prompt, workflow: { nodes: [], v: 1 } });
|
||
const carried = plain(m.projectMetadata(read({ prompt, workflow: '{"nodes": [], "v": 1}' }, { prompt: "original-text", workflow: "reserialized-json" }, envelope)));
|
||
assert.deepEqual(carried.prompt_graph, { status: "parsed", value: { 1: { class_type: "A", inputs: { x: "1.0" } } } });
|
||
assert.deepEqual(carried.workflow_graph, { status: "parsed", value: { nodes: [], v: 1 } });
|
||
const broken = plain(m.projectMetadata(read({ prompt, workflow }, { prompt: "original-text", workflow: "original-text" }, "[1]")));
|
||
assert.deepEqual([broken.prompt_graph.status, broken.workflow_graph.status], ["malformed", "malformed"]);
|
||
const deep = plain(m.projectMetadata(read({ prompt }, { prompt: "original-text" }, `{"title":${nested(70)},"prompt":${JSON.stringify(prompt)}}`)));
|
||
assert.equal(deep.prompt_graph.status, "limited");
|
||
const partial = plain(m.projectMetadata(read({ prompt, workflow }, { prompt: "original-text", workflow: "original-text" }, '{"prompt":"{","workflow":5}')));
|
||
assert.deepEqual([partial.prompt_graph.status, partial.workflow_graph.status], ["malformed", "malformed"]);
|
||
const missing = plain(m.projectMetadata(read({ prompt }, { prompt: "original-text" }, '{"workflow":{}}')));
|
||
assert.deepEqual([missing.prompt_graph.status, missing.workflow_graph.status], ["malformed", "absent"]);
|
||
// The carrier's own prompt object is projected: depth 2, node keys filtered.
|
||
const objectPrompt = plain(m.projectMetadata(read({ prompt: "x" }, { prompt: "reserialized-json" }, '{"prompt":{"1":{"class_type":"A","inputs":{},"_meta":{}}}}')));
|
||
assert.deepEqual(objectPrompt.prompt_graph, { status: "parsed", value: { 1: { class_type: "A", inputs: {} } } });
|
||
});
|
||
|
||
test("fitInspectionPayload omits the workflow graph, then the prompt graph, then all raw atomically", async () => {
|
||
const payload = {
|
||
success: true, inspection_version: 1, metadata: {
|
||
status: "available", omitted_reason: null, raw: { prompt: "é".repeat(100), workflow: "{}" }, record_formats: { prompt: "original-text", workflow: "original-text" },
|
||
parameters: null, prompt_graph: { status: "parsed", value: { 1: { class_type: "A", inputs: { t: "\ud800".repeat(50) } } } },
|
||
workflow_graph: { status: "parsed", value: { nodes: Array.from({ length: 50 }, (_, i) => ({ id: i })) } },
|
||
},
|
||
};
|
||
const text = JSON.stringify(payload);
|
||
// UTF-8 bytes, not UTF-16 code units: 'é' is 2 bytes, an escaped lone surrogate 6.
|
||
assert.equal(m.inspectionBytes(payload), Buffer.byteLength(text, "utf8"));
|
||
assert.ok(m.inspectionBytes(payload) > text.length);
|
||
// Measured in bounded chunks: pairs and 2-byte characters across every chunk edge count once.
|
||
for (const shift of [0, 1, 2, 3]) {
|
||
const wide = { s: "x".repeat(shift) + "é😀".repeat(200000) + "\ud800", t: "a😀".repeat(70000) };
|
||
assert.equal(m.inspectionBytes(wide), Buffer.byteLength(JSON.stringify(wide), "utf8"), `shift ${shift}`);
|
||
}
|
||
const whole = await m.fitInspectionPayload(payload, bytes(payload));
|
||
assert.equal(whole, payload, "a payload within the limit is returned as is");
|
||
const first = await m.fitInspectionPayload(payload, bytes(payload) - 1);
|
||
assert.equal(first.metadata.workflow_graph.status, "limited");
|
||
assert.equal(first.metadata.prompt_graph.status, "parsed");
|
||
assert.deepEqual(first.metadata.raw, payload.metadata.raw);
|
||
assert.ok(bytes(first) <= bytes(payload) - 1);
|
||
const second = await m.fitInspectionPayload(payload, bytes(first) - 1);
|
||
assert.deepEqual([second.metadata.workflow_graph.status, second.metadata.prompt_graph.status], ["limited", "limited"]);
|
||
assert.equal(second.metadata.status, "available");
|
||
assert.deepEqual(second.metadata.raw, payload.metadata.raw);
|
||
const third = await m.fitInspectionPayload(payload, bytes(second) - 1);
|
||
assert.deepEqual(third.metadata, { ...second.metadata, raw: {}, parameters: null, status: "limited", omitted_reason: "response-size-limit" });
|
||
assert.ok(bytes(third) <= bytes(second) - 1);
|
||
await assert.rejects(m.fitInspectionPayload(payload, bytes(third) - 1), /exceeds/);
|
||
// The caller's payload is never mutated and absent graphs are left alone.
|
||
assert.equal(JSON.stringify(payload), text);
|
||
const absent = { ...payload, metadata: { ...payload.metadata, workflow_graph: { status: "absent", value: null } } };
|
||
const fitted = await m.fitInspectionPayload(absent, bytes(absent) - 1);
|
||
assert.equal(fitted.metadata.workflow_graph.status, "absent");
|
||
assert.equal(fitted.metadata.prompt_graph.status, "limited");
|
||
assert.equal((await m.fitInspectionPayload(payload)).metadata.status, "available", "the default limit is 8 MiB");
|
||
// Every omission step after the first measure hands the event loop a turn, and an
|
||
// abort between steps rejects with AbortError; the answer itself is unchanged.
|
||
let turns = 0;
|
||
const counted = await m.fitInspectionPayload(payload, bytes(second) - 1, { yieldControl: async () => { turns += 1; } });
|
||
assert.equal(JSON.stringify(counted), JSON.stringify(third));
|
||
assert.equal(turns, 3);
|
||
assert.equal(await m.fitInspectionPayload(payload, bytes(payload), { yieldControl: async () => { turns += 1; } }), payload);
|
||
assert.equal(turns, 3, "a payload within the limit is measured once, without a turn");
|
||
const controller = new AbortController();
|
||
await assert.rejects(m.fitInspectionPayload(payload, bytes(second) - 1, { signal: controller.signal, yieldControl: async () => controller.abort() }), { name: "AbortError" });
|
||
});
|
||
|
||
test("the committed projection corpus has every outcome and parses with the oracle entry shape", () => {
|
||
assert.ok(CASES.length >= 40);
|
||
const names = new Set(), statuses = new Set(), graphs = new Set();
|
||
for (const entry of CASES) {
|
||
assert.equal(typeof entry.name, "string");
|
||
assert.ok(!names.has(entry.name), `duplicate case ${entry.name}`);
|
||
names.add(entry.name);
|
||
const read = entry.records === null ? null : { records: entry.records, record_formats: entry.record_formats ?? Object.fromEntries(Object.keys(entry.records).map((key) => [key, "original-text"])),
|
||
carrierText: entry.carrier_text ?? null, limited: entry.limited ?? [] };
|
||
const metadata = m.projectMetadata(read);
|
||
statuses.add(metadata.status);
|
||
graphs.add(metadata.prompt_graph.status);
|
||
graphs.add(metadata.workflow_graph.status);
|
||
if (entry.expect) {
|
||
for (const [field, status] of Object.entries(entry.expect)) assert.equal(metadata[field].status, status, `${entry.name} ${field}`);
|
||
}
|
||
}
|
||
assert.deepEqual([...statuses].sort(), ["absent", "available", "unreadable"]);
|
||
assert.deepEqual([...graphs].sort(), ["absent", "limited", "malformed", "parsed"]);
|
||
});
|