haansi 0.1.136 → 0.1.137
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/haansi.js +77 -34
- package/package.json +1 -1
package/dist/haansi.js
CHANGED
|
@@ -95,7 +95,7 @@ var require_package = __commonJS({
|
|
|
95
95
|
"package.json"(exports2, module2) {
|
|
96
96
|
module2.exports = {
|
|
97
97
|
name: "haansi",
|
|
98
|
-
version: "0.1.
|
|
98
|
+
version: "0.1.137",
|
|
99
99
|
description: "Haansi CLI - Session collector and MCP server for Claude Code & OpenAI Codex",
|
|
100
100
|
bin: {
|
|
101
101
|
haansi: "./dist/haansi.js"
|
|
@@ -2417,6 +2417,14 @@ var init_scaling_metrics = __esm({
|
|
|
2417
2417
|
treatMissingData: "breaching",
|
|
2418
2418
|
missingDataRationale: "The reconciler is the only publisher and runs on a schedule independent of every worker. Its silence means the reconciler itself is broken, and the backlog it was watching is unobserved rather than empty. Scaling out against an unknown backlog wastes money; scaling in loses the queue. Breaching keeps capacity up until a human looks."
|
|
2419
2419
|
}),
|
|
2420
|
+
MINING_PENDING_DECISIONS: spec({
|
|
2421
|
+
name: "haansi/mining/pending_decisions",
|
|
2422
|
+
unit: "Count",
|
|
2423
|
+
description: "Context decisions queued for the LLM (mining_pending_decisions, state 'pending'), platform-resident orgs.",
|
|
2424
|
+
dimensions: ["environment"],
|
|
2425
|
+
treatMissingData: "breaching",
|
|
2426
|
+
missingDataRationale: "Read only by the adaptive controller as a guard on scaling canonical ingestion: faster ingestion just grows this queue. Absent means the queue is unobserved, and the controller holds ingestion capacity rather than guessing in either direction."
|
|
2427
|
+
}),
|
|
2420
2428
|
MINING_REPLAYS_IN_PROGRESS: spec({
|
|
2421
2429
|
name: "haansi/mining/replays_in_progress",
|
|
2422
2430
|
unit: "Count",
|
|
@@ -88810,27 +88818,14 @@ function baseNameOf(path2) {
|
|
|
88810
88818
|
const name = /\.py$/.test(file2) ? file2.replace(/\.py$/, "").replace(/^test_(?=.)|(?<=.)_test$/, "") : file2.replace(/\.[cm]?[jt]sx?$/, "").replace(/\.(?:test|spec)$/, "");
|
|
88811
88819
|
return (name === "index" || name === "__init__") && parts.length ? parts[parts.length - 1] : name;
|
|
88812
88820
|
}
|
|
88821
|
+
function isUnitTestFile(path2) {
|
|
88822
|
+
const lang = adapterFor(path2);
|
|
88823
|
+
return !!lang && lang.isTestFile(path2) && !NOT_UNIT.test(path2);
|
|
88824
|
+
}
|
|
88813
88825
|
async function candidateTests(show, task) {
|
|
88814
|
-
const code = new Set(task.codeFiles);
|
|
88815
|
-
const tests = new Set(task.testFiles);
|
|
88816
|
-
const related = [];
|
|
88817
|
-
const placed = [];
|
|
88818
|
-
for (const file2 of task.testFiles) {
|
|
88819
|
-
const direct = await hopOf(show, task.targetSha, [file2], tests);
|
|
88820
|
-
const second = await hopOf(
|
|
88821
|
-
show,
|
|
88822
|
-
task.targetSha,
|
|
88823
|
-
direct,
|
|
88824
|
-
/* @__PURE__ */ new Set([...tests, ...direct])
|
|
88825
|
-
);
|
|
88826
|
-
if (direct.length) placed.push(file2);
|
|
88827
|
-
if ([...direct, ...second].some((m) => code.has(m))) related.push(file2);
|
|
88828
|
-
}
|
|
88829
|
-
const files = related.length ? related : placed.length ? [] : [...task.testFiles];
|
|
88830
88826
|
const out = [];
|
|
88831
|
-
for (const file2 of
|
|
88827
|
+
for (const file2 of task.testFiles.filter(isUnitTestFile)) {
|
|
88832
88828
|
const lang = adapterFor(file2);
|
|
88833
|
-
if (!lang) continue;
|
|
88834
88829
|
const source = await show(task.targetSha, file2);
|
|
88835
88830
|
if (source === null) continue;
|
|
88836
88831
|
const before = await show(task.baseSha, file2);
|
|
@@ -88840,7 +88835,7 @@ async function candidateTests(show, task) {
|
|
|
88840
88835
|
}
|
|
88841
88836
|
return out;
|
|
88842
88837
|
}
|
|
88843
|
-
var MAX_TITLES, MAX_TITLE_CHARS, MAX_FAILING_PATHS, MAX_TYPE_DECLARATIONS, MAX_HOP_MODULES, norm;
|
|
88838
|
+
var MAX_TITLES, MAX_TITLE_CHARS, MAX_FAILING_PATHS, MAX_TYPE_DECLARATIONS, MAX_HOP_MODULES, norm, NOT_UNIT;
|
|
88844
88839
|
var init_replay_spec = __esm({
|
|
88845
88840
|
"src/commands/replay-spec.ts"() {
|
|
88846
88841
|
"use strict";
|
|
@@ -88853,6 +88848,7 @@ var init_replay_spec = __esm({
|
|
|
88853
88848
|
MAX_TYPE_DECLARATIONS = 40;
|
|
88854
88849
|
MAX_HOP_MODULES = 24;
|
|
88855
88850
|
norm = (s) => s.replace(/\s+/g, " ").trim();
|
|
88851
|
+
NOT_UNIT = /(^|\/)(tests\/integration|integration|e2e|playwright|cypress)\/|\.(integration|e2e)\.(test|spec)\.[a-z]+$|\.snap$|\.snapshot\.txt$/i;
|
|
88856
88852
|
}
|
|
88857
88853
|
});
|
|
88858
88854
|
|
|
@@ -89017,6 +89013,7 @@ function answerOf(text) {
|
|
|
89017
89013
|
function parseSelection(text, candidates) {
|
|
89018
89014
|
const v = answerOf(text);
|
|
89019
89015
|
if (!v || !Array.isArray(v.tests)) return null;
|
|
89016
|
+
if (v.tests.length === 0) return [];
|
|
89020
89017
|
const picked = new Set(
|
|
89021
89018
|
v.tests.filter((t) => typeof t === "string").map(key)
|
|
89022
89019
|
);
|
|
@@ -89074,8 +89071,9 @@ async function selectExpectedTests(deps, input) {
|
|
|
89074
89071
|
);
|
|
89075
89072
|
return { tests: all, picked: false };
|
|
89076
89073
|
}
|
|
89074
|
+
const added = input.candidates.filter((c) => c.isNew).length;
|
|
89077
89075
|
deps.log(
|
|
89078
|
-
`\u2026 ${input.label}: graded on ${tests.length} of ${offered.length} candidate test(s)
|
|
89076
|
+
`\u2026 ${input.label}: graded on the ${added} test(s) the change added and ${tests.filter((t) => !input.candidates.find((c) => c.name === t)?.isNew).length} more picked, of ${offered.length} candidate test(s)`
|
|
89079
89077
|
);
|
|
89080
89078
|
return { tests, picked: true };
|
|
89081
89079
|
} finally {
|
|
@@ -89192,8 +89190,7 @@ async function expectedTestsFor(deps, task, pick2) {
|
|
|
89192
89190
|
const candidates = await candidateTests(deps.showFile, {
|
|
89193
89191
|
baseSha: task.baseSha,
|
|
89194
89192
|
targetSha: task.targetSha,
|
|
89195
|
-
testFiles: task.testFiles
|
|
89196
|
-
codeFiles
|
|
89193
|
+
testFiles: task.testFiles
|
|
89197
89194
|
});
|
|
89198
89195
|
if (candidates.length === 0) return [];
|
|
89199
89196
|
const stat = codeFiles.length ? await (deps.gitStdout ?? deps.git)(
|
|
@@ -89212,7 +89209,10 @@ async function expectedTestsFor(deps, task, pick2) {
|
|
|
89212
89209
|
const change = stat.code === 0 ? stat.output.trim().split("\n").map((l) => l.split(" ")).filter((c) => c.length >= 3).map(
|
|
89213
89210
|
([a, r, path2]) => a === "-" ? `- ${path2} (binary)` : `- ${path2} (+${a} \u2212${r})`
|
|
89214
89211
|
).join("\n") : codeFiles.map((f) => `- ${f}`).join("\n");
|
|
89215
|
-
const
|
|
89212
|
+
const picked = new Set(
|
|
89213
|
+
pick2 ? await pick2(task, candidates, change) : candidates.map((c) => c.name)
|
|
89214
|
+
);
|
|
89215
|
+
const tests = candidates.filter((c) => c.isNew || picked.has(c.name)).map((c) => c.name);
|
|
89216
89216
|
return tests.length <= MAX_EXPECTED_TESTS ? tests : [];
|
|
89217
89217
|
}
|
|
89218
89218
|
function parseCalibrateArgs(argv) {
|
|
@@ -90493,6 +90493,9 @@ var init_replay_recurring = __esm({
|
|
|
90493
90493
|
});
|
|
90494
90494
|
|
|
90495
90495
|
// src/commands/replay-context.ts
|
|
90496
|
+
function meetsMinimum(score, min) {
|
|
90497
|
+
return score.onPoint >= (min?.onPoint ?? 0) && score.nearby >= (min?.nearby ?? 0);
|
|
90498
|
+
}
|
|
90496
90499
|
function storedScoreOf(task) {
|
|
90497
90500
|
return {
|
|
90498
90501
|
onPoint: task.context?.onPoint ?? 0,
|
|
@@ -90741,11 +90744,17 @@ async function gateOnContext(deps, tasks, options) {
|
|
|
90741
90744
|
let noContext = 0;
|
|
90742
90745
|
let nearby = 0;
|
|
90743
90746
|
let unknown2 = 0;
|
|
90747
|
+
let below = 0;
|
|
90744
90748
|
for (const t of tasks) {
|
|
90745
90749
|
const label = t.name ?? t.taskKey;
|
|
90746
90750
|
const s = states.get(t.taskId);
|
|
90747
90751
|
const tier = gateTierOf(s, recurringOf(t).length) ?? (s === "none" && options.history?.(t) ? "history" : null);
|
|
90748
|
-
if (tier) {
|
|
90752
|
+
if (tier && !meetsMinimum(scoreOf(t), options.minimum)) {
|
|
90753
|
+
below += 1;
|
|
90754
|
+
deps.log(
|
|
90755
|
+
`skip ${label}: ${scoreOf(t).onPoint} on point and ${scoreOf(t).nearby} nearby in Haansi's search, under the run's minimum (${options.minimum?.onPoint ?? 0} on point, ${options.minimum?.nearby ?? 0} nearby)`
|
|
90756
|
+
);
|
|
90757
|
+
} else if (tier) {
|
|
90749
90758
|
tiers.set(t.taskId, tier);
|
|
90750
90759
|
if (tier === "missed")
|
|
90751
90760
|
deps.log(
|
|
@@ -90780,6 +90789,7 @@ async function gateOnContext(deps, tasks, options) {
|
|
|
90780
90789
|
return {
|
|
90781
90790
|
replay: replay2,
|
|
90782
90791
|
nearby,
|
|
90792
|
+
belowMinimum: below,
|
|
90783
90793
|
missed: count4("missed"),
|
|
90784
90794
|
recurring: count4("recurring"),
|
|
90785
90795
|
history: count4("history"),
|
|
@@ -90788,10 +90798,10 @@ async function gateOnContext(deps, tasks, options) {
|
|
|
90788
90798
|
tiers
|
|
90789
90799
|
};
|
|
90790
90800
|
}
|
|
90791
|
-
function fromChecks(tasks, limit, shard, history) {
|
|
90801
|
+
function fromChecks(tasks, limit, shard, history, minimum) {
|
|
90792
90802
|
const tierOf = (t) => gateTierOf(t.context?.state, t.context?.recurringIds?.length ?? 0) ?? (t.context?.state === "none" && history?.(t) ? "history" : null);
|
|
90793
90803
|
const ready = tasks.filter(
|
|
90794
|
-
(t) => !needsCheck(t) && tierOf(t) !== null && t.calibration?.state === "valid" && t.calibration.inputsFingerprint === t.inputsFingerprint
|
|
90804
|
+
(t) => !needsCheck(t) && tierOf(t) !== null && meetsMinimum(storedScoreOf(t), minimum) && t.calibration?.state === "valid" && t.calibration.inputsFingerprint === t.inputsFingerprint
|
|
90795
90805
|
);
|
|
90796
90806
|
const ranked = rankFound(ready, storedScoreOf, (t) => tierOf(t));
|
|
90797
90807
|
const top = limit ? ranked.slice(0, limit) : ranked;
|
|
@@ -91261,6 +91271,10 @@ async function runStoredReplays(options, deps) {
|
|
|
91261
91271
|
set2 = { ...set2, tasks: fresh };
|
|
91262
91272
|
if (fresh.length === 0) return { results: [], notValid: 0, recorded: null };
|
|
91263
91273
|
}
|
|
91274
|
+
const minimum = {
|
|
91275
|
+
onPoint: options.minOnPoint ?? 0,
|
|
91276
|
+
nearby: options.minNearby ?? 0
|
|
91277
|
+
};
|
|
91264
91278
|
const historyTier = options.arms.includes("history") ? (t) => hasHistory(
|
|
91265
91279
|
t,
|
|
91266
91280
|
historyScope === "all" && set2.historyActivity === true ? "all" : "git"
|
|
@@ -91272,7 +91286,8 @@ async function runStoredReplays(options, deps) {
|
|
|
91272
91286
|
set2.tasks,
|
|
91273
91287
|
options.limit,
|
|
91274
91288
|
options.shard ?? null,
|
|
91275
|
-
historyTier
|
|
91289
|
+
historyTier,
|
|
91290
|
+
minimum
|
|
91276
91291
|
);
|
|
91277
91292
|
mine = picked.mine;
|
|
91278
91293
|
tiers = picked.tiers;
|
|
@@ -91393,7 +91408,8 @@ async function runStoredReplays(options, deps) {
|
|
|
91393
91408
|
targetOf,
|
|
91394
91409
|
prepare: interfaceOf,
|
|
91395
91410
|
prompt: promptForRelevance,
|
|
91396
|
-
history: historyTier
|
|
91411
|
+
history: historyTier,
|
|
91412
|
+
minimum
|
|
91397
91413
|
});
|
|
91398
91414
|
gated = gate.replay;
|
|
91399
91415
|
nearby = gate.nearby;
|
|
@@ -91523,7 +91539,14 @@ async function runStoredReplays(options, deps) {
|
|
|
91523
91539
|
);
|
|
91524
91540
|
return;
|
|
91525
91541
|
}
|
|
91526
|
-
|
|
91542
|
+
const repeats = options.repeats ?? DEFAULT_REPEATS;
|
|
91543
|
+
const runs = Array.from(
|
|
91544
|
+
{ length: repeats },
|
|
91545
|
+
(_, i) => arms.map((arm) => ({ arm, run: i + 1 }))
|
|
91546
|
+
).flat();
|
|
91547
|
+
for (const { arm, run: run3 } of runs) {
|
|
91548
|
+
if (repeats > 1 && arm === arms[0])
|
|
91549
|
+
deps.log(`\u2026 ${label}: run ${run3}/${repeats}`);
|
|
91527
91550
|
if (arm === "history" && history.length === 0) {
|
|
91528
91551
|
deps.log(
|
|
91529
91552
|
`skip history ${label}: nothing of its workstream before it began`
|
|
@@ -91847,6 +91870,8 @@ var init_replay_tasks = __esm({
|
|
|
91847
91870
|
// src/commands/replay.ts
|
|
91848
91871
|
var replay_exports = {};
|
|
91849
91872
|
__export(replay_exports, {
|
|
91873
|
+
DEFAULT_REPEATS: () => DEFAULT_REPEATS,
|
|
91874
|
+
MAX_REPEATS: () => MAX_REPEATS,
|
|
91850
91875
|
OUTPUT_HEAD_CHARS: () => OUTPUT_HEAD_CHARS,
|
|
91851
91876
|
OUTPUT_TAIL_CHARS: () => OUTPUT_TAIL_CHARS,
|
|
91852
91877
|
SHA: () => SHA2,
|
|
@@ -91893,7 +91918,8 @@ function parseReplayArgs(argv) {
|
|
|
91893
91918
|
concurrency: 1,
|
|
91894
91919
|
setupCache: null,
|
|
91895
91920
|
force: false,
|
|
91896
|
-
maxHours: null
|
|
91921
|
+
maxHours: null,
|
|
91922
|
+
repeats: DEFAULT_REPEATS
|
|
91897
91923
|
};
|
|
91898
91924
|
let sub;
|
|
91899
91925
|
const errors = [];
|
|
@@ -91911,6 +91937,9 @@ function parseReplayArgs(argv) {
|
|
|
91911
91937
|
else if (a === "--test-cmd") options.testCmd = next();
|
|
91912
91938
|
else if (a === "--setup-cmd") options.setupCmd = next() || null;
|
|
91913
91939
|
else if (a === "--attempts") options.attempts = Number(next());
|
|
91940
|
+
else if (a === "--repeats") options.repeats = Number(next());
|
|
91941
|
+
else if (a === "--min-on-point") options.minOnPoint = Number(next());
|
|
91942
|
+
else if (a === "--min-nearby") options.minNearby = Number(next());
|
|
91914
91943
|
else if (a === "--limit") options.limit = Number(next());
|
|
91915
91944
|
else if (a === "--timeout-min") options.timeoutMs = Number(next()) * 6e4;
|
|
91916
91945
|
else if (a === "--agent-name") options.agentName = next();
|
|
@@ -91945,6 +91974,14 @@ function parseReplayArgs(argv) {
|
|
|
91945
91974
|
);
|
|
91946
91975
|
if (!Number.isInteger(options.attempts) || options.attempts < 1 || options.attempts > 20)
|
|
91947
91976
|
errors.push("--attempts must be 1 to 20.");
|
|
91977
|
+
if (!Number.isInteger(options.repeats) || options.repeats < 1 || options.repeats > MAX_REPEATS)
|
|
91978
|
+
errors.push(`--repeats must be 1 to ${MAX_REPEATS}.`);
|
|
91979
|
+
for (const [flag, v] of [
|
|
91980
|
+
["--min-on-point", options.minOnPoint],
|
|
91981
|
+
["--min-nearby", options.minNearby]
|
|
91982
|
+
])
|
|
91983
|
+
if (v !== void 0 && (!Number.isInteger(v) || v < 0))
|
|
91984
|
+
errors.push(`${flag} must be a whole number, 0 or more.`);
|
|
91948
91985
|
if (options.limit !== null && (!Number.isInteger(options.limit) || options.limit < 1))
|
|
91949
91986
|
errors.push("--limit must be a positive whole number.");
|
|
91950
91987
|
if (!Number.isFinite(options.timeoutMs) || options.timeoutMs <= 0)
|
|
@@ -92537,7 +92574,7 @@ Baseline: ${arm("baseline")} passed \xB7 With Haansi: ${arm("haansi")} passed${h
|
|
|
92537
92574
|
removeWorkDir(work);
|
|
92538
92575
|
}
|
|
92539
92576
|
}
|
|
92540
|
-
var import_node_child_process5, import_node_fs14, import_node_os11, import_node_path19, usesHaansi, SHA2, ARMS, DEFAULT_ARMS, FEEDBACK_CHARS, FAILURE_LINES2, REPLAY_BRANCH, OUTPUT_HEAD_CHARS, OUTPUT_TAIL_CHARS, USAGE2;
|
|
92577
|
+
var import_node_child_process5, import_node_fs14, import_node_os11, import_node_path19, DEFAULT_REPEATS, MAX_REPEATS, usesHaansi, SHA2, ARMS, DEFAULT_ARMS, FEEDBACK_CHARS, FAILURE_LINES2, REPLAY_BRANCH, OUTPUT_HEAD_CHARS, OUTPUT_TAIL_CHARS, USAGE2;
|
|
92541
92578
|
var init_replay = __esm({
|
|
92542
92579
|
"src/commands/replay.ts"() {
|
|
92543
92580
|
"use strict";
|
|
@@ -92556,6 +92593,8 @@ var init_replay = __esm({
|
|
|
92556
92593
|
init_replay_usage();
|
|
92557
92594
|
init_replay_preflight();
|
|
92558
92595
|
init_replay_scale();
|
|
92596
|
+
DEFAULT_REPEATS = 2;
|
|
92597
|
+
MAX_REPEATS = 5;
|
|
92559
92598
|
usesHaansi = (arm) => arm !== "baseline";
|
|
92560
92599
|
SHA2 = /^[0-9a-f]{7,40}$/i;
|
|
92561
92600
|
ARMS = ["baseline", "haansi", "history"];
|
|
@@ -92566,7 +92605,8 @@ var init_replay = __esm({
|
|
|
92566
92605
|
OUTPUT_HEAD_CHARS = 4e4;
|
|
92567
92606
|
OUTPUT_TAIL_CHARS = 2e5;
|
|
92568
92607
|
USAGE2 = `Usage: haansi replay run --agent-cmd "<cmd>" --test-cmd "<cmd>"
|
|
92569
|
-
[--setup-cmd "<cmd>"] [--attempts 2] [--arms baseline,haansi,history] [--history git|all] [--limit n]
|
|
92608
|
+
[--setup-cmd "<cmd>"] [--attempts 2] [--repeats 2] [--arms baseline,haansi,history] [--history git|all] [--limit n]
|
|
92609
|
+
[--min-on-point n] [--min-nearby n]
|
|
92570
92610
|
[--timeout-min 20] [--agent-name <name>] [--model <name>]
|
|
92571
92611
|
[--judge-cmd "<cmd>"]
|
|
92572
92612
|
[--run-ref <url>] [--dry-run] [--repo-commands] [--check-commands] [--no-diffs]
|
|
@@ -92590,7 +92630,10 @@ test output on this machine and reports only how far they match.
|
|
|
92590
92630
|
At scale: --shard i/n takes this job's share of the tasks (a CI matrix of n
|
|
92591
92631
|
jobs); --concurrency k replays k tasks at once here; an arm already measured
|
|
92592
92632
|
on the same card with the same agent and model is skipped (--force runs it
|
|
92593
|
-
again); --
|
|
92633
|
+
again); --repeats n runs each arm n times per task (default 2), interleaved
|
|
92634
|
+
(each arm once, then each again), each run its own result; --min-on-point
|
|
92635
|
+
and --min-nearby replay only tasks whose context check returned at least
|
|
92636
|
+
that many on-point and nearby items; --new-only leaves out every task replayed before on all its arms with
|
|
92594
92637
|
the same agent and model, in any version, before --limit picks; --max-hours stops starting arms after h hours. --setup-cache
|
|
92595
92638
|
package-lock.json --setup-cache-path '**/node_modules' keeps the setup's output
|
|
92596
92639
|
(every workspace's node_modules) per lockfile (in ~/.cache/haansi-replay, or --setup-cache-dir) and restores
|