sentinelayer-cli 0.1.2 → 0.4.4
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/README.md +998 -996
- package/bin/create-sentinelayer.js +5 -5
- package/bin/sentinelayer-cli.js +4 -4
- package/bin/sl.js +5 -5
- package/package.json +63 -54
- package/src/agents/jules/config/definition.js +209 -209
- package/src/agents/jules/config/system-prompt.js +175 -175
- package/src/agents/jules/error-intake.js +51 -51
- package/src/agents/jules/fix-cycle.js +377 -377
- package/src/agents/jules/loop.js +367 -367
- package/src/agents/jules/pulse.js +327 -319
- package/src/agents/jules/stream.js +186 -186
- package/src/agents/jules/swarm/file-scanner.js +74 -74
- package/src/agents/jules/swarm/index.js +11 -11
- package/src/agents/jules/swarm/orchestrator.js +362 -362
- package/src/agents/jules/swarm/pattern-hunter.js +123 -123
- package/src/agents/jules/swarm/sub-agent.js +308 -308
- package/src/agents/jules/tools/auth-audit.js +557 -222
- package/src/agents/jules/tools/dispatch.js +327 -327
- package/src/agents/jules/tools/file-edit.js +180 -180
- package/src/agents/jules/tools/file-read.js +100 -100
- package/src/agents/jules/tools/frontend-analyze.js +570 -570
- package/src/agents/jules/tools/glob.js +168 -168
- package/src/agents/jules/tools/grep.js +228 -228
- package/src/agents/jules/tools/index.js +29 -29
- package/src/agents/jules/tools/path-guards.js +161 -161
- package/src/agents/jules/tools/runtime-audit.js +503 -493
- package/src/agents/jules/tools/shell.js +383 -383
- package/src/agents/jules/tools/url-policy.js +100 -0
- package/src/ai/aidenid.js +972 -945
- package/src/ai/client.js +508 -508
- package/src/ai/domain-target-store.js +268 -268
- package/src/ai/identity-store.js +270 -270
- package/src/ai/site-store.js +145 -145
- package/src/audit/agents/architecture.js +180 -180
- package/src/audit/agents/compliance.js +179 -179
- package/src/audit/agents/documentation.js +165 -165
- package/src/audit/agents/performance.js +145 -145
- package/src/audit/agents/security.js +215 -215
- package/src/audit/agents/testing.js +172 -172
- package/src/audit/orchestrator.js +557 -557
- package/src/audit/package.js +204 -204
- package/src/audit/registry.js +284 -284
- package/src/audit/replay.js +103 -103
- package/src/auth/gate.js +45 -11
- package/src/auth/http.js +270 -113
- package/src/auth/service.js +891 -848
- package/src/auth/session-store.js +359 -345
- package/src/cli.js +252 -252
- package/src/commands/ai/identity-lifecycle.js +1338 -1337
- package/src/commands/ai/provision-governance.js +1272 -1246
- package/src/commands/ai/shared.js +147 -147
- package/src/commands/ai.js +11 -11
- package/src/commands/apply.js +12 -12
- package/src/commands/audit.js +1166 -1166
- package/src/commands/auth.js +375 -366
- package/src/commands/chat.js +191 -191
- package/src/commands/config.js +184 -184
- package/src/commands/cost.js +311 -311
- package/src/commands/daemon/core.js +850 -850
- package/src/commands/daemon/extended.js +1048 -1048
- package/src/commands/daemon/shared.js +213 -213
- package/src/commands/daemon.js +11 -11
- package/src/commands/guide.js +174 -174
- package/src/commands/ingest.js +58 -58
- package/src/commands/init.js +55 -55
- package/src/commands/legacy-args.js +10 -10
- package/src/commands/mcp.js +461 -404
- package/src/commands/omargate.js +15 -15
- package/src/commands/persona.js +20 -20
- package/src/commands/plugin.js +260 -260
- package/src/commands/policy.js +132 -132
- package/src/commands/prompt.js +238 -238
- package/src/commands/review.js +704 -704
- package/src/commands/scan.js +866 -788
- package/src/commands/spec.js +716 -716
- package/src/commands/swarm.js +651 -651
- package/src/commands/telemetry.js +202 -202
- package/src/commands/watch.js +510 -510
- package/src/config/agent-dictionary.js +182 -182
- package/src/config/io.js +56 -56
- package/src/config/paths.js +18 -18
- package/src/config/schema.js +55 -55
- package/src/config/service.js +184 -184
- package/src/cost/budget.js +235 -235
- package/src/cost/history.js +188 -188
- package/src/cost/tracker.js +171 -171
- package/src/daemon/artifact-lineage.js +534 -534
- package/src/daemon/assignment-ledger.js +770 -770
- package/src/daemon/ast-parser-layer.js +258 -258
- package/src/daemon/budget-governor.js +633 -633
- package/src/daemon/callgraph-overlay.js +646 -646
- package/src/daemon/error-worker.js +626 -626
- package/src/daemon/hybrid-mapper.js +929 -929
- package/src/daemon/jira-lifecycle.js +632 -632
- package/src/daemon/operator-control.js +657 -657
- package/src/daemon/reliability-lane.js +471 -471
- package/src/daemon/watchdog.js +971 -971
- package/src/guide/generator.js +316 -316
- package/src/ingest/engine.js +918 -918
- package/src/legacy-cli.js +2592 -2435
- package/src/mcp/registry.js +695 -695
- package/src/memory/blackboard.js +301 -301
- package/src/memory/retrieval.js +581 -581
- package/src/plugin/manifest.js +553 -553
- package/src/policy/packs.js +144 -144
- package/src/prompt/generator.js +118 -106
- package/src/review/ai-review.js +669 -669
- package/src/review/local-review.js +1295 -1284
- package/src/review/replay.js +235 -235
- package/src/review/report.js +664 -664
- package/src/review/spec-binding.js +487 -487
- package/src/scaffold/generator.js +67 -0
- package/src/scaffold/templates.js +150 -0
- package/src/scan/generator.js +418 -351
- package/src/scan/gh-secrets.js +107 -0
- package/src/spec/generator.js +519 -519
- package/src/spec/regenerate.js +237 -237
- package/src/spec/templates.js +91 -91
- package/src/swarm/dashboard.js +247 -247
- package/src/swarm/factory.js +363 -363
- package/src/swarm/pentest.js +934 -934
- package/src/swarm/registry.js +419 -419
- package/src/swarm/report.js +158 -158
- package/src/swarm/runtime.js +576 -576
- package/src/swarm/scenario-dsl.js +272 -272
- package/src/telemetry/ledger.js +302 -302
- package/src/telemetry/sync.js +107 -61
- package/src/ui/markdown.js +220 -220
package/src/cost/budget.js
CHANGED
|
@@ -1,235 +1,235 @@
|
|
|
1
|
-
function normalizeNumber(value, field) {
|
|
2
|
-
const normalized = Number(value || 0);
|
|
3
|
-
if (!Number.isFinite(normalized) || normalized < 0) {
|
|
4
|
-
throw new Error(`${field} must be a non-negative number.`);
|
|
5
|
-
}
|
|
6
|
-
return normalized;
|
|
7
|
-
}
|
|
8
|
-
|
|
9
|
-
function formatNumber(value, decimals = 0) {
|
|
10
|
-
return Number(value || 0).toFixed(decimals);
|
|
11
|
-
}
|
|
12
|
-
|
|
13
|
-
function normalizeWarningThresholdPercent(value) {
|
|
14
|
-
const normalized = Number(value || 0);
|
|
15
|
-
if (!Number.isFinite(normalized) || normalized < 0 || normalized > 100) {
|
|
16
|
-
throw new Error("warningThresholdPercent must be between 0 and 100.");
|
|
17
|
-
}
|
|
18
|
-
return normalized;
|
|
19
|
-
}
|
|
20
|
-
|
|
21
|
-
function collectLimitStatusWithThreshold({
|
|
22
|
-
reasons,
|
|
23
|
-
warnings,
|
|
24
|
-
usageValue,
|
|
25
|
-
limitValue,
|
|
26
|
-
warningThresholdPercent,
|
|
27
|
-
stopCode,
|
|
28
|
-
warningCode,
|
|
29
|
-
stopMessage,
|
|
30
|
-
warningMessage,
|
|
31
|
-
}) {
|
|
32
|
-
const normalizedUsage = normalizeNumber(usageValue, "usageValue");
|
|
33
|
-
const normalizedLimit = normalizeNumber(limitValue, "limitValue");
|
|
34
|
-
const normalizedWarningThresholdPercent =
|
|
35
|
-
normalizeWarningThresholdPercent(warningThresholdPercent);
|
|
36
|
-
|
|
37
|
-
if (normalizedLimit <= 0) {
|
|
38
|
-
return;
|
|
39
|
-
}
|
|
40
|
-
|
|
41
|
-
if (normalizedUsage > normalizedLimit) {
|
|
42
|
-
reasons.push({
|
|
43
|
-
code: stopCode,
|
|
44
|
-
message: stopMessage(normalizedUsage, normalizedLimit),
|
|
45
|
-
});
|
|
46
|
-
return;
|
|
47
|
-
}
|
|
48
|
-
|
|
49
|
-
if (normalizedWarningThresholdPercent <= 0) {
|
|
50
|
-
return;
|
|
51
|
-
}
|
|
52
|
-
|
|
53
|
-
const thresholdValue = (normalizedWarningThresholdPercent / 100) * normalizedLimit;
|
|
54
|
-
if (normalizedUsage >= thresholdValue) {
|
|
55
|
-
warnings.push({
|
|
56
|
-
code: warningCode,
|
|
57
|
-
message: warningMessage(normalizedUsage, normalizedLimit, normalizedWarningThresholdPercent),
|
|
58
|
-
});
|
|
59
|
-
}
|
|
60
|
-
}
|
|
61
|
-
|
|
62
|
-
/**
|
|
63
|
-
* Evaluate runtime/cost/token/tool/no-progress budgets and return deterministic stop/warning signals.
|
|
64
|
-
*
|
|
65
|
-
* @param {{
|
|
66
|
-
* sessionSummary?: {
|
|
67
|
-
* costUsd?: number,
|
|
68
|
-
* outputTokens?: number,
|
|
69
|
-
* noProgressStreak?: number,
|
|
70
|
-
* durationMs?: number,
|
|
71
|
-
* toolCalls?: number
|
|
72
|
-
* },
|
|
73
|
-
* maxCostUsd?: number,
|
|
74
|
-
* maxOutputTokens?: number,
|
|
75
|
-
* maxNoProgress?: number,
|
|
76
|
-
* maxRuntimeMs?: number,
|
|
77
|
-
* maxToolCalls?: number,
|
|
78
|
-
* warningThresholdPercent?: number
|
|
79
|
-
* }} [options]
|
|
80
|
-
* @returns {{
|
|
81
|
-
* blocking: boolean,
|
|
82
|
-
* warnings: Array<{ code: string, message: string }>,
|
|
83
|
-
* reasons: Array<{ code: string, message: string }>,
|
|
84
|
-
* limits: {
|
|
85
|
-
* maxCostUsd: number,
|
|
86
|
-
* maxOutputTokens: number,
|
|
87
|
-
* maxNoProgress: number,
|
|
88
|
-
* maxRuntimeMs: number,
|
|
89
|
-
* maxToolCalls: number,
|
|
90
|
-
* warningThresholdPercent: number
|
|
91
|
-
* },
|
|
92
|
-
* usage: {
|
|
93
|
-
* costUsd: number,
|
|
94
|
-
* outputTokens: number,
|
|
95
|
-
* noProgressStreak: number,
|
|
96
|
-
* runtimeMs: number,
|
|
97
|
-
* toolCalls: number
|
|
98
|
-
* }
|
|
99
|
-
* }}
|
|
100
|
-
*/
|
|
101
|
-
export function evaluateBudget({
|
|
102
|
-
sessionSummary = {},
|
|
103
|
-
maxCostUsd = 1.0,
|
|
104
|
-
maxOutputTokens = 0,
|
|
105
|
-
maxNoProgress = 3,
|
|
106
|
-
maxRuntimeMs = 0,
|
|
107
|
-
maxToolCalls = 0,
|
|
108
|
-
warningThresholdPercent = 80,
|
|
109
|
-
} = {}) {
|
|
110
|
-
const normalizedMaxCost = normalizeNumber(maxCostUsd, "maxCostUsd");
|
|
111
|
-
const normalizedMaxOutputTokens = normalizeNumber(maxOutputTokens, "maxOutputTokens");
|
|
112
|
-
const normalizedMaxNoProgress = Math.max(1, normalizeNumber(maxNoProgress, "maxNoProgress"));
|
|
113
|
-
const normalizedMaxRuntimeMs = normalizeNumber(maxRuntimeMs, "maxRuntimeMs");
|
|
114
|
-
const normalizedMaxToolCalls = normalizeNumber(maxToolCalls, "maxToolCalls");
|
|
115
|
-
const normalizedWarningThresholdPercent = normalizeWarningThresholdPercent(
|
|
116
|
-
warningThresholdPercent
|
|
117
|
-
);
|
|
118
|
-
|
|
119
|
-
const totalCostUsd = normalizeNumber(sessionSummary.costUsd || 0, "sessionSummary.costUsd");
|
|
120
|
-
const totalOutputTokens = normalizeNumber(
|
|
121
|
-
sessionSummary.outputTokens || 0,
|
|
122
|
-
"sessionSummary.outputTokens"
|
|
123
|
-
);
|
|
124
|
-
const noProgressStreak = normalizeNumber(
|
|
125
|
-
sessionSummary.noProgressStreak || 0,
|
|
126
|
-
"sessionSummary.noProgressStreak"
|
|
127
|
-
);
|
|
128
|
-
const totalRuntimeMs = normalizeNumber(sessionSummary.durationMs || 0, "sessionSummary.durationMs");
|
|
129
|
-
const totalToolCalls = normalizeNumber(
|
|
130
|
-
sessionSummary.toolCalls || 0,
|
|
131
|
-
"sessionSummary.toolCalls"
|
|
132
|
-
);
|
|
133
|
-
|
|
134
|
-
const reasons = [];
|
|
135
|
-
const warnings = [];
|
|
136
|
-
|
|
137
|
-
collectLimitStatusWithThreshold({
|
|
138
|
-
reasons,
|
|
139
|
-
warnings,
|
|
140
|
-
usageValue: totalCostUsd,
|
|
141
|
-
limitValue: normalizedMaxCost,
|
|
142
|
-
warningThresholdPercent: normalizedWarningThresholdPercent,
|
|
143
|
-
stopCode: "MAX_COST_EXCEEDED",
|
|
144
|
-
warningCode: "COST_BUDGET_NEAR_LIMIT",
|
|
145
|
-
stopMessage: (usage, limit) =>
|
|
146
|
-
`Cost budget exceeded (${formatNumber(usage, 6)} > ${formatNumber(limit, 6)}).`,
|
|
147
|
-
warningMessage: (usage, limit, thresholdPercent) =>
|
|
148
|
-
`Cost budget near limit (${formatNumber(usage, 6)} / ${formatNumber(limit, 6)} at ${formatNumber(
|
|
149
|
-
thresholdPercent,
|
|
150
|
-
0
|
|
151
|
-
)}%).`,
|
|
152
|
-
});
|
|
153
|
-
|
|
154
|
-
collectLimitStatusWithThreshold({
|
|
155
|
-
reasons,
|
|
156
|
-
warnings,
|
|
157
|
-
usageValue: totalOutputTokens,
|
|
158
|
-
limitValue: normalizedMaxOutputTokens,
|
|
159
|
-
warningThresholdPercent: normalizedWarningThresholdPercent,
|
|
160
|
-
stopCode: "MAX_OUTPUT_TOKENS_EXCEEDED",
|
|
161
|
-
warningCode: "OUTPUT_TOKENS_NEAR_LIMIT",
|
|
162
|
-
stopMessage: (usage, limit) =>
|
|
163
|
-
`Output token budget exceeded (${formatNumber(usage, 0)} > ${formatNumber(limit, 0)}).`,
|
|
164
|
-
warningMessage: (usage, limit, thresholdPercent) =>
|
|
165
|
-
`Output token budget near limit (${formatNumber(usage, 0)} / ${formatNumber(
|
|
166
|
-
limit,
|
|
167
|
-
0
|
|
168
|
-
)} at ${formatNumber(thresholdPercent, 0)}%).`,
|
|
169
|
-
});
|
|
170
|
-
|
|
171
|
-
collectLimitStatusWithThreshold({
|
|
172
|
-
reasons,
|
|
173
|
-
warnings,
|
|
174
|
-
usageValue: totalRuntimeMs,
|
|
175
|
-
limitValue: normalizedMaxRuntimeMs,
|
|
176
|
-
warningThresholdPercent: normalizedWarningThresholdPercent,
|
|
177
|
-
stopCode: "MAX_RUNTIME_MS_EXCEEDED",
|
|
178
|
-
warningCode: "RUNTIME_MS_NEAR_LIMIT",
|
|
179
|
-
stopMessage: (usage, limit) =>
|
|
180
|
-
`Runtime budget exceeded (${formatNumber(usage, 0)}ms > ${formatNumber(limit, 0)}ms).`,
|
|
181
|
-
warningMessage: (usage, limit, thresholdPercent) =>
|
|
182
|
-
`Runtime budget near limit (${formatNumber(usage, 0)}ms / ${formatNumber(
|
|
183
|
-
limit,
|
|
184
|
-
0
|
|
185
|
-
)}ms at ${formatNumber(thresholdPercent, 0)}%).`,
|
|
186
|
-
});
|
|
187
|
-
|
|
188
|
-
collectLimitStatusWithThreshold({
|
|
189
|
-
reasons,
|
|
190
|
-
warnings,
|
|
191
|
-
usageValue: totalToolCalls,
|
|
192
|
-
limitValue: normalizedMaxToolCalls,
|
|
193
|
-
warningThresholdPercent: normalizedWarningThresholdPercent,
|
|
194
|
-
stopCode: "MAX_TOOL_CALLS_EXCEEDED",
|
|
195
|
-
warningCode: "TOOL_CALLS_NEAR_LIMIT",
|
|
196
|
-
stopMessage: (usage, limit) =>
|
|
197
|
-
`Tool-call budget exceeded (${formatNumber(usage, 0)} > ${formatNumber(limit, 0)}).`,
|
|
198
|
-
warningMessage: (usage, limit, thresholdPercent) =>
|
|
199
|
-
`Tool-call budget near limit (${formatNumber(usage, 0)} / ${formatNumber(
|
|
200
|
-
limit,
|
|
201
|
-
0
|
|
202
|
-
)} at ${formatNumber(thresholdPercent, 0)}%).`,
|
|
203
|
-
});
|
|
204
|
-
|
|
205
|
-
if (noProgressStreak >= normalizedMaxNoProgress) {
|
|
206
|
-
reasons.push({
|
|
207
|
-
code: "DIMINISHING_RETURNS",
|
|
208
|
-
message: `No-progress streak reached ${formatNumber(noProgressStreak, 0)} (threshold ${formatNumber(
|
|
209
|
-
normalizedMaxNoProgress,
|
|
210
|
-
0
|
|
211
|
-
)}).`,
|
|
212
|
-
});
|
|
213
|
-
}
|
|
214
|
-
|
|
215
|
-
return {
|
|
216
|
-
blocking: reasons.length > 0,
|
|
217
|
-
warnings,
|
|
218
|
-
reasons,
|
|
219
|
-
limits: {
|
|
220
|
-
maxCostUsd: normalizedMaxCost,
|
|
221
|
-
maxOutputTokens: normalizedMaxOutputTokens,
|
|
222
|
-
maxNoProgress: normalizedMaxNoProgress,
|
|
223
|
-
maxRuntimeMs: normalizedMaxRuntimeMs,
|
|
224
|
-
maxToolCalls: normalizedMaxToolCalls,
|
|
225
|
-
warningThresholdPercent: normalizedWarningThresholdPercent,
|
|
226
|
-
},
|
|
227
|
-
usage: {
|
|
228
|
-
costUsd: totalCostUsd,
|
|
229
|
-
outputTokens: totalOutputTokens,
|
|
230
|
-
noProgressStreak,
|
|
231
|
-
runtimeMs: totalRuntimeMs,
|
|
232
|
-
toolCalls: totalToolCalls,
|
|
233
|
-
},
|
|
234
|
-
};
|
|
235
|
-
}
|
|
1
|
+
function normalizeNumber(value, field) {
|
|
2
|
+
const normalized = Number(value || 0);
|
|
3
|
+
if (!Number.isFinite(normalized) || normalized < 0) {
|
|
4
|
+
throw new Error(`${field} must be a non-negative number.`);
|
|
5
|
+
}
|
|
6
|
+
return normalized;
|
|
7
|
+
}
|
|
8
|
+
|
|
9
|
+
function formatNumber(value, decimals = 0) {
|
|
10
|
+
return Number(value || 0).toFixed(decimals);
|
|
11
|
+
}
|
|
12
|
+
|
|
13
|
+
function normalizeWarningThresholdPercent(value) {
|
|
14
|
+
const normalized = Number(value || 0);
|
|
15
|
+
if (!Number.isFinite(normalized) || normalized < 0 || normalized > 100) {
|
|
16
|
+
throw new Error("warningThresholdPercent must be between 0 and 100.");
|
|
17
|
+
}
|
|
18
|
+
return normalized;
|
|
19
|
+
}
|
|
20
|
+
|
|
21
|
+
function collectLimitStatusWithThreshold({
|
|
22
|
+
reasons,
|
|
23
|
+
warnings,
|
|
24
|
+
usageValue,
|
|
25
|
+
limitValue,
|
|
26
|
+
warningThresholdPercent,
|
|
27
|
+
stopCode,
|
|
28
|
+
warningCode,
|
|
29
|
+
stopMessage,
|
|
30
|
+
warningMessage,
|
|
31
|
+
}) {
|
|
32
|
+
const normalizedUsage = normalizeNumber(usageValue, "usageValue");
|
|
33
|
+
const normalizedLimit = normalizeNumber(limitValue, "limitValue");
|
|
34
|
+
const normalizedWarningThresholdPercent =
|
|
35
|
+
normalizeWarningThresholdPercent(warningThresholdPercent);
|
|
36
|
+
|
|
37
|
+
if (normalizedLimit <= 0) {
|
|
38
|
+
return;
|
|
39
|
+
}
|
|
40
|
+
|
|
41
|
+
if (normalizedUsage > normalizedLimit) {
|
|
42
|
+
reasons.push({
|
|
43
|
+
code: stopCode,
|
|
44
|
+
message: stopMessage(normalizedUsage, normalizedLimit),
|
|
45
|
+
});
|
|
46
|
+
return;
|
|
47
|
+
}
|
|
48
|
+
|
|
49
|
+
if (normalizedWarningThresholdPercent <= 0) {
|
|
50
|
+
return;
|
|
51
|
+
}
|
|
52
|
+
|
|
53
|
+
const thresholdValue = (normalizedWarningThresholdPercent / 100) * normalizedLimit;
|
|
54
|
+
if (normalizedUsage >= thresholdValue) {
|
|
55
|
+
warnings.push({
|
|
56
|
+
code: warningCode,
|
|
57
|
+
message: warningMessage(normalizedUsage, normalizedLimit, normalizedWarningThresholdPercent),
|
|
58
|
+
});
|
|
59
|
+
}
|
|
60
|
+
}
|
|
61
|
+
|
|
62
|
+
/**
|
|
63
|
+
* Evaluate runtime/cost/token/tool/no-progress budgets and return deterministic stop/warning signals.
|
|
64
|
+
*
|
|
65
|
+
* @param {{
|
|
66
|
+
* sessionSummary?: {
|
|
67
|
+
* costUsd?: number,
|
|
68
|
+
* outputTokens?: number,
|
|
69
|
+
* noProgressStreak?: number,
|
|
70
|
+
* durationMs?: number,
|
|
71
|
+
* toolCalls?: number
|
|
72
|
+
* },
|
|
73
|
+
* maxCostUsd?: number,
|
|
74
|
+
* maxOutputTokens?: number,
|
|
75
|
+
* maxNoProgress?: number,
|
|
76
|
+
* maxRuntimeMs?: number,
|
|
77
|
+
* maxToolCalls?: number,
|
|
78
|
+
* warningThresholdPercent?: number
|
|
79
|
+
* }} [options]
|
|
80
|
+
* @returns {{
|
|
81
|
+
* blocking: boolean,
|
|
82
|
+
* warnings: Array<{ code: string, message: string }>,
|
|
83
|
+
* reasons: Array<{ code: string, message: string }>,
|
|
84
|
+
* limits: {
|
|
85
|
+
* maxCostUsd: number,
|
|
86
|
+
* maxOutputTokens: number,
|
|
87
|
+
* maxNoProgress: number,
|
|
88
|
+
* maxRuntimeMs: number,
|
|
89
|
+
* maxToolCalls: number,
|
|
90
|
+
* warningThresholdPercent: number
|
|
91
|
+
* },
|
|
92
|
+
* usage: {
|
|
93
|
+
* costUsd: number,
|
|
94
|
+
* outputTokens: number,
|
|
95
|
+
* noProgressStreak: number,
|
|
96
|
+
* runtimeMs: number,
|
|
97
|
+
* toolCalls: number
|
|
98
|
+
* }
|
|
99
|
+
* }}
|
|
100
|
+
*/
|
|
101
|
+
export function evaluateBudget({
|
|
102
|
+
sessionSummary = {},
|
|
103
|
+
maxCostUsd = 1.0,
|
|
104
|
+
maxOutputTokens = 0,
|
|
105
|
+
maxNoProgress = 3,
|
|
106
|
+
maxRuntimeMs = 0,
|
|
107
|
+
maxToolCalls = 0,
|
|
108
|
+
warningThresholdPercent = 80,
|
|
109
|
+
} = {}) {
|
|
110
|
+
const normalizedMaxCost = normalizeNumber(maxCostUsd, "maxCostUsd");
|
|
111
|
+
const normalizedMaxOutputTokens = normalizeNumber(maxOutputTokens, "maxOutputTokens");
|
|
112
|
+
const normalizedMaxNoProgress = Math.max(1, normalizeNumber(maxNoProgress, "maxNoProgress"));
|
|
113
|
+
const normalizedMaxRuntimeMs = normalizeNumber(maxRuntimeMs, "maxRuntimeMs");
|
|
114
|
+
const normalizedMaxToolCalls = normalizeNumber(maxToolCalls, "maxToolCalls");
|
|
115
|
+
const normalizedWarningThresholdPercent = normalizeWarningThresholdPercent(
|
|
116
|
+
warningThresholdPercent
|
|
117
|
+
);
|
|
118
|
+
|
|
119
|
+
const totalCostUsd = normalizeNumber(sessionSummary.costUsd || 0, "sessionSummary.costUsd");
|
|
120
|
+
const totalOutputTokens = normalizeNumber(
|
|
121
|
+
sessionSummary.outputTokens || 0,
|
|
122
|
+
"sessionSummary.outputTokens"
|
|
123
|
+
);
|
|
124
|
+
const noProgressStreak = normalizeNumber(
|
|
125
|
+
sessionSummary.noProgressStreak || 0,
|
|
126
|
+
"sessionSummary.noProgressStreak"
|
|
127
|
+
);
|
|
128
|
+
const totalRuntimeMs = normalizeNumber(sessionSummary.durationMs || 0, "sessionSummary.durationMs");
|
|
129
|
+
const totalToolCalls = normalizeNumber(
|
|
130
|
+
sessionSummary.toolCalls || 0,
|
|
131
|
+
"sessionSummary.toolCalls"
|
|
132
|
+
);
|
|
133
|
+
|
|
134
|
+
const reasons = [];
|
|
135
|
+
const warnings = [];
|
|
136
|
+
|
|
137
|
+
collectLimitStatusWithThreshold({
|
|
138
|
+
reasons,
|
|
139
|
+
warnings,
|
|
140
|
+
usageValue: totalCostUsd,
|
|
141
|
+
limitValue: normalizedMaxCost,
|
|
142
|
+
warningThresholdPercent: normalizedWarningThresholdPercent,
|
|
143
|
+
stopCode: "MAX_COST_EXCEEDED",
|
|
144
|
+
warningCode: "COST_BUDGET_NEAR_LIMIT",
|
|
145
|
+
stopMessage: (usage, limit) =>
|
|
146
|
+
`Cost budget exceeded (${formatNumber(usage, 6)} > ${formatNumber(limit, 6)}).`,
|
|
147
|
+
warningMessage: (usage, limit, thresholdPercent) =>
|
|
148
|
+
`Cost budget near limit (${formatNumber(usage, 6)} / ${formatNumber(limit, 6)} at ${formatNumber(
|
|
149
|
+
thresholdPercent,
|
|
150
|
+
0
|
|
151
|
+
)}%).`,
|
|
152
|
+
});
|
|
153
|
+
|
|
154
|
+
collectLimitStatusWithThreshold({
|
|
155
|
+
reasons,
|
|
156
|
+
warnings,
|
|
157
|
+
usageValue: totalOutputTokens,
|
|
158
|
+
limitValue: normalizedMaxOutputTokens,
|
|
159
|
+
warningThresholdPercent: normalizedWarningThresholdPercent,
|
|
160
|
+
stopCode: "MAX_OUTPUT_TOKENS_EXCEEDED",
|
|
161
|
+
warningCode: "OUTPUT_TOKENS_NEAR_LIMIT",
|
|
162
|
+
stopMessage: (usage, limit) =>
|
|
163
|
+
`Output token budget exceeded (${formatNumber(usage, 0)} > ${formatNumber(limit, 0)}).`,
|
|
164
|
+
warningMessage: (usage, limit, thresholdPercent) =>
|
|
165
|
+
`Output token budget near limit (${formatNumber(usage, 0)} / ${formatNumber(
|
|
166
|
+
limit,
|
|
167
|
+
0
|
|
168
|
+
)} at ${formatNumber(thresholdPercent, 0)}%).`,
|
|
169
|
+
});
|
|
170
|
+
|
|
171
|
+
collectLimitStatusWithThreshold({
|
|
172
|
+
reasons,
|
|
173
|
+
warnings,
|
|
174
|
+
usageValue: totalRuntimeMs,
|
|
175
|
+
limitValue: normalizedMaxRuntimeMs,
|
|
176
|
+
warningThresholdPercent: normalizedWarningThresholdPercent,
|
|
177
|
+
stopCode: "MAX_RUNTIME_MS_EXCEEDED",
|
|
178
|
+
warningCode: "RUNTIME_MS_NEAR_LIMIT",
|
|
179
|
+
stopMessage: (usage, limit) =>
|
|
180
|
+
`Runtime budget exceeded (${formatNumber(usage, 0)}ms > ${formatNumber(limit, 0)}ms).`,
|
|
181
|
+
warningMessage: (usage, limit, thresholdPercent) =>
|
|
182
|
+
`Runtime budget near limit (${formatNumber(usage, 0)}ms / ${formatNumber(
|
|
183
|
+
limit,
|
|
184
|
+
0
|
|
185
|
+
)}ms at ${formatNumber(thresholdPercent, 0)}%).`,
|
|
186
|
+
});
|
|
187
|
+
|
|
188
|
+
collectLimitStatusWithThreshold({
|
|
189
|
+
reasons,
|
|
190
|
+
warnings,
|
|
191
|
+
usageValue: totalToolCalls,
|
|
192
|
+
limitValue: normalizedMaxToolCalls,
|
|
193
|
+
warningThresholdPercent: normalizedWarningThresholdPercent,
|
|
194
|
+
stopCode: "MAX_TOOL_CALLS_EXCEEDED",
|
|
195
|
+
warningCode: "TOOL_CALLS_NEAR_LIMIT",
|
|
196
|
+
stopMessage: (usage, limit) =>
|
|
197
|
+
`Tool-call budget exceeded (${formatNumber(usage, 0)} > ${formatNumber(limit, 0)}).`,
|
|
198
|
+
warningMessage: (usage, limit, thresholdPercent) =>
|
|
199
|
+
`Tool-call budget near limit (${formatNumber(usage, 0)} / ${formatNumber(
|
|
200
|
+
limit,
|
|
201
|
+
0
|
|
202
|
+
)} at ${formatNumber(thresholdPercent, 0)}%).`,
|
|
203
|
+
});
|
|
204
|
+
|
|
205
|
+
if (noProgressStreak >= normalizedMaxNoProgress) {
|
|
206
|
+
reasons.push({
|
|
207
|
+
code: "DIMINISHING_RETURNS",
|
|
208
|
+
message: `No-progress streak reached ${formatNumber(noProgressStreak, 0)} (threshold ${formatNumber(
|
|
209
|
+
normalizedMaxNoProgress,
|
|
210
|
+
0
|
|
211
|
+
)}).`,
|
|
212
|
+
});
|
|
213
|
+
}
|
|
214
|
+
|
|
215
|
+
return {
|
|
216
|
+
blocking: reasons.length > 0,
|
|
217
|
+
warnings,
|
|
218
|
+
reasons,
|
|
219
|
+
limits: {
|
|
220
|
+
maxCostUsd: normalizedMaxCost,
|
|
221
|
+
maxOutputTokens: normalizedMaxOutputTokens,
|
|
222
|
+
maxNoProgress: normalizedMaxNoProgress,
|
|
223
|
+
maxRuntimeMs: normalizedMaxRuntimeMs,
|
|
224
|
+
maxToolCalls: normalizedMaxToolCalls,
|
|
225
|
+
warningThresholdPercent: normalizedWarningThresholdPercent,
|
|
226
|
+
},
|
|
227
|
+
usage: {
|
|
228
|
+
costUsd: totalCostUsd,
|
|
229
|
+
outputTokens: totalOutputTokens,
|
|
230
|
+
noProgressStreak,
|
|
231
|
+
runtimeMs: totalRuntimeMs,
|
|
232
|
+
toolCalls: totalToolCalls,
|
|
233
|
+
},
|
|
234
|
+
};
|
|
235
|
+
}
|