pr-shepherd 0.53.0 → 0.53.1

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.
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "name": "pr-shepherd",
3
3
  "description": "Autonomous PR CI monitor and review-comment resolver for agentic coding tools",
4
- "version": "0.53.0",
4
+ "version": "0.53.1",
5
5
  "author": {
6
6
  "name": "Jonathan Ong",
7
7
  "email": "jonathanrichardong@gmail.com"
@@ -9,7 +9,7 @@ export function formatQuotaWarning(warning) {
9
9
  if (warning === undefined)
10
10
  return null;
11
11
  const used = warning.used === undefined ? "" : ` · used ${warning.used}`;
12
- return [
12
+ const lines = [
13
13
  "## GitHub API quota warning",
14
14
  "",
15
15
  `- Resource: \`${warning.resource}\``,
@@ -18,8 +18,22 @@ export function formatQuotaWarning(warning) {
18
18
  `- Reset: ${resetTime(warning.resetAt)}`,
19
19
  `- Recommended poll interval: ${warning.pollIntervalMinutes} minutes`,
20
20
  `- Recommended bounded CLI timeout: ${warning.pollTimeoutMinutes} minutes`,
21
- "- Recommendation: keep polling pr-shepherd at the cadence above; for incidental PR operations prefer REST `gh` (`gh pr view`, `gh pr review`, `gh api`); do not substitute `gh pr checks`/`gh pr watch`",
22
- ].join("\n");
21
+ recommendation(warning),
22
+ ];
23
+ for (const budget of warning.budgets ?? []) {
24
+ const budgetUsed = budget.used === undefined ? "" : ` · used ${budget.used}`;
25
+ lines.push(`- Budget \`${budget.resource}\`: ${budget.remaining}/${budget.limit} remaining${budgetUsed} · crossed ${budget.thresholdPercent}% · resets ${resetTime(budget.resetAt)}`);
26
+ }
27
+ return lines.join("\n");
28
+ }
29
+ function recommendation(warning) {
30
+ if (warning.resource === "combined") {
31
+ return "- Recommendation: keep polling pr-shepherd at the cadence above; both GraphQL and REST core are below their warning thresholds, so do not shift work between them. Do not substitute `gh pr checks` or `gh pr watch`";
32
+ }
33
+ if (warning.resource === "core") {
34
+ return "- Recommendation: keep polling pr-shepherd at the cadence above; do not add incidental REST `gh` calls (`gh pr view`, `gh pr review`, `gh api`) while REST core is low. Do not substitute `gh pr checks` or `gh pr watch`";
35
+ }
36
+ return "- Recommendation: keep polling pr-shepherd at the cadence above; for incidental PR operations prefer REST `gh` (`gh pr view`, `gh pr review`, `gh api`); do not substitute `gh pr checks` or `gh pr watch`";
23
37
  }
24
38
  export function formatApiUsage(usage) {
25
39
  if (usage === undefined)
@@ -1,6 +1,6 @@
1
1
  import { loadConfig } from "../../config/load.mjs";
2
- import { summarizeApiTelemetry, withGraphqlCredentialFingerprint, } from "../../github/api-telemetry.mjs";
3
- import { evaluateWorktreeGraphqlQuotaWarning } from "../../state/graphql-quota-warnings.mjs";
2
+ import { summarizeApiTelemetry } from "../../github/api-telemetry.mjs";
3
+ import { selectQuotaWarning } from "../quota-selection.mjs";
4
4
  import { buildQuotaAwareContinuation } from "../../quota-warning.mjs";
5
5
  function shouldWarn(result) {
6
6
  return ["wait", "mark_ready", "merge", "fix_code"].includes(result.action);
@@ -10,14 +10,15 @@ export async function attachApiUsage(result, persistWarning, preservePersistedWa
10
10
  if (apiUsage === undefined)
11
11
  return result;
12
12
  let quotaWarning = preservePersistedWarning ? result.quotaWarning : undefined;
13
- if (quotaWarning === undefined && apiUsage.graphql !== undefined && shouldWarn(result)) {
13
+ const core = apiUsage.rest?.find((item) => item.resource === "core");
14
+ if (quotaWarning === undefined && shouldWarn(result) && (apiUsage.graphql || core)) {
14
15
  const [owner, repo] = result.repo.split("/");
15
16
  if (owner && repo) {
16
17
  const bands = loadConfig().watch.graphqlQuotaWarnings.map((band) => ({
17
18
  ...band,
18
19
  pollIntervalMinutes: Math.max(band.pollIntervalMinutes, minimumPollIntervalMinutes),
19
20
  }));
20
- quotaWarning = await evaluateWorktreeGraphqlQuotaWarning({ owner, repo }, bands, withGraphqlCredentialFingerprint(apiUsage.graphql), persistWarning);
21
+ quotaWarning = await selectQuotaWarning({ owner, repo }, bands, apiUsage, persistWarning);
21
22
  }
22
23
  }
23
24
  const { quotaWarning: _deferredWarning, ...baseResult } = result;
@@ -1,10 +1,22 @@
1
1
  import type { GraphqlQuotaWarningBand } from "../config/load.mts";
2
- import type { GraphqlApiUsage, PollSummaryResult } from "../types.mts";
3
- /** Sleep at least `--interval`, and at least the active crossed quota band. */
4
- export declare function graphqlQuotaPollIntervalMs(bands: GraphqlQuotaWarningBand[], usage: Pick<GraphqlApiUsage, "remaining" | "limit"> | undefined, fallbackMs: number, maxMs: number): number;
2
+ import type { ApiResourceUsage, GraphqlApiUsage, PollSummaryResult } from "../types.mts";
3
+ /** Slow the poll for whichever of GraphQL or REST core is in a tighter band. */
4
+ export declare function quotaPollIntervalMs(bands: GraphqlQuotaWarningBand[], usage: {
5
+ graphql?: Pick<GraphqlApiUsage, "remaining" | "limit">;
6
+ rest?: Pick<ApiResourceUsage, "resource" | "remaining" | "limit">[];
7
+ } | undefined, fallbackMs: number, maxMs: number): number;
8
+ export interface RateLimitRetry {
9
+ ms: number;
10
+ resource: string;
11
+ remaining?: number;
12
+ limit?: number;
13
+ resetAt?: number;
14
+ }
5
15
  /**
6
- * Retry delay for `--until-terminal` when GitHub returns a GraphQL 429 / secondary
7
- * limit. `null` means the error is not a retryable rate limit.
16
+ * Retry delay for `--until-terminal` when GitHub exhausts a primary quota or
17
+ * returns a secondary limit. `null` means the error is not a retryable rate limit.
8
18
  */
9
- export declare function pollGraphQlRetryAfterMs(err: unknown): number | null;
19
+ export declare function pollRateLimitRetryAfterMs(err: unknown): RateLimitRetry | null;
20
+ /** Stderr line naming the exhausted budget and when the sleep ends. */
21
+ export declare function formatRateLimitRetryLine(tickLabel: string, elapsedSeconds: number, retry: RateLimitRetry): string;
10
22
  export declare function aggregateQuotaWarning(result: PollSummaryResult, bands: GraphqlQuotaWarningBand[], intervalSeconds: number): Promise<PollSummaryResult["quotaWarning"]>;
@@ -1,10 +1,10 @@
1
- import { evaluateWorktreeGraphqlQuotaWarning } from "../state/graphql-quota-warnings.mjs";
2
- import { summarizeApiTelemetry, withGraphqlCredentialFingerprint, } from "../github/api-telemetry.mjs";
1
+ import { summarizeApiTelemetry } from "../github/api-telemetry.mjs";
3
2
  import { GitHubRequestError } from "../github/errors.mjs";
4
3
  import { isRateLimitMessage } from "../comments/rate-limit.mjs";
4
+ import { selectQuotaWarning } from "./quota-selection.mjs";
5
5
  const GRAPHQL_RETRY_AFTER_DEFAULT_MS = 60_000;
6
6
  /** Sleep at least `--interval`, and at least the active crossed quota band. */
7
- export function graphqlQuotaPollIntervalMs(bands, usage, fallbackMs, maxMs) {
7
+ function graphqlQuotaPollIntervalMs(bands, usage, fallbackMs, maxMs) {
8
8
  if (usage === undefined || usage.limit <= 0 || bands.length === 0) {
9
9
  return Math.min(fallbackMs, maxMs);
10
10
  }
@@ -15,34 +15,72 @@ export function graphqlQuotaPollIntervalMs(bands, usage, fallbackMs, maxMs) {
15
15
  const bandMs = active.pollIntervalMinutes * 60_000;
16
16
  return Math.min(Math.max(fallbackMs, bandMs), maxMs);
17
17
  }
18
+ /** Slow the poll for whichever of GraphQL or REST core is in a tighter band. */
19
+ export function quotaPollIntervalMs(bands, usage, fallbackMs, maxMs) {
20
+ const core = usage?.rest?.find((item) => item.resource === "core");
21
+ return Math.max(graphqlQuotaPollIntervalMs(bands, usage?.graphql, fallbackMs, maxMs), graphqlQuotaPollIntervalMs(bands, core, fallbackMs, maxMs));
22
+ }
18
23
  /**
19
- * Retry delay for `--until-terminal` when GitHub returns a GraphQL 429 / secondary
20
- * limit. `null` means the error is not a retryable rate limit.
24
+ * Retry delay for `--until-terminal` when GitHub exhausts a primary quota or
25
+ * returns a secondary limit. `null` means the error is not a retryable rate limit.
21
26
  */
22
- export function pollGraphQlRetryAfterMs(err) {
27
+ export function pollRateLimitRetryAfterMs(err) {
23
28
  if (!(err instanceof GitHubRequestError))
24
29
  return null;
25
- const retryable = err.status === 429 ||
26
- err.retryAfterSeconds !== undefined ||
27
- isRateLimitMessage(err.message) ||
28
- (err.graphqlErrors?.some((error) => isRateLimitMessage(error.message)) ?? false) ||
29
- (err.rateLimit !== undefined && err.rateLimit.remaining <= 0);
30
+ const rateLimitMessage = isRateLimitMessage(err.message) ||
31
+ (err.graphqlErrors?.some((error) => isRateLimitMessage(error.message)) ?? false);
32
+ const exhausted = err.rateLimit !== undefined && err.rateLimit.remaining <= 0;
33
+ const retryable = err.status === 429 || err.retryAfterSeconds !== undefined || rateLimitMessage || exhausted;
30
34
  if (!retryable)
31
35
  return null;
32
- if (err.retryAfterSeconds !== undefined)
33
- return Math.max(err.retryAfterSeconds, 0) * 1000;
34
- if (err.rateLimit !== undefined && err.rateLimit.remaining <= 0) {
35
- return Math.max(err.rateLimit.resetAt * 1000 - Date.now(), 0);
36
+ const resource = retryResource(err);
37
+ const rateLimit = err.rateLimit;
38
+ const details = {
39
+ resource,
40
+ ...(rateLimit?.remaining !== undefined && { remaining: rateLimit.remaining }),
41
+ ...(rateLimit?.limit !== undefined && { limit: rateLimit.limit }),
42
+ ...(rateLimit?.resetAt !== undefined && { resetAt: rateLimit.resetAt }),
43
+ };
44
+ if (err.retryAfterSeconds !== undefined) {
45
+ return { ...details, ms: Math.max(err.retryAfterSeconds, 0) * 1000 };
36
46
  }
37
- return GRAPHQL_RETRY_AFTER_DEFAULT_MS;
47
+ if (exhausted && rateLimit !== undefined) {
48
+ return { ...details, ms: Math.max(rateLimit.resetAt * 1000 - Date.now(), 0) };
49
+ }
50
+ return { ...details, ms: GRAPHQL_RETRY_AFTER_DEFAULT_MS };
51
+ }
52
+ function retryResource(err) {
53
+ if (err.rateLimit?.resource)
54
+ return err.rateLimit.resource;
55
+ const secondary = err.status === 429 ||
56
+ err.retryAfterSeconds !== undefined ||
57
+ /secondary/i.test(err.message) ||
58
+ (err.graphqlErrors?.some((error) => /secondary/i.test(error.message)) ?? false);
59
+ return secondary ? "secondary" : "graphql";
60
+ }
61
+ /** Stderr line naming the exhausted budget and when the sleep ends. */
62
+ export function formatRateLimitRetryLine(tickLabel, elapsedSeconds, retry) {
63
+ const resetAt = retry.resetAt ?? Math.ceil((Date.now() + retry.ms) / 1000);
64
+ const clock = `${new Date(resetAt * 1000).toISOString().slice(11, 19)}Z`;
65
+ const label = retry.resource === "graphql"
66
+ ? "GitHub GraphQL rate limit"
67
+ : retry.resource === "secondary"
68
+ ? "GitHub secondary rate limit"
69
+ : `GitHub REST ${retry.resource} rate limit`;
70
+ const counts = retry.remaining !== undefined && retry.limit !== undefined
71
+ ? ` (${retry.remaining}/${retry.limit})`
72
+ : "";
73
+ return `[${tickLabel} / +${elapsedSeconds}s] ${label}${counts} — retrying at ${clock} (in ${Math.round(retry.ms / 1000)}s)\n`;
38
74
  }
39
75
  export async function aggregateQuotaWarning(result, bands, intervalSeconds) {
40
- const usage = summarizeApiTelemetry()?.graphql;
76
+ const usage = summarizeApiTelemetry();
41
77
  const [owner, repo] = result.repo.split("/");
42
78
  if (!usage || !owner || !repo)
43
79
  return undefined;
44
- return evaluateWorktreeGraphqlQuotaWarning({ owner, repo }, bands.map((band) => ({
80
+ if (!usage.graphql && !usage.rest?.some((item) => item.resource === "core"))
81
+ return undefined;
82
+ return selectQuotaWarning({ owner, repo }, bands.map((band) => ({
45
83
  ...band,
46
84
  pollIntervalMinutes: Math.max(band.pollIntervalMinutes, intervalSeconds / 60),
47
- })), withGraphqlCredentialFingerprint(usage), true);
85
+ })), usage, true);
48
86
  }
@@ -5,7 +5,7 @@ import { getRepoInfo } from "../github/client.mjs";
5
5
  import { withApiTelemetryScope, summarizeApiTelemetry } from "../github/api-telemetry.mjs";
6
6
  import { fetchPollSummary } from "../github/poll-summary.mjs";
7
7
  import { sleep } from "../util/sleep.mjs";
8
- import { aggregateQuotaWarning, graphqlQuotaPollIntervalMs, pollGraphQlRetryAfterMs, } from "./poll-quota.mjs";
8
+ import { aggregateQuotaWarning, formatRateLimitRetryLine, pollRateLimitRetryAfterMs, quotaPollIntervalMs, } from "./poll-quota.mjs";
9
9
  import { planPollSummary, withPollSummaryInstructions } from "./poll-summary-instructions.mjs";
10
10
  import { summaryStatusSignature } from "./poll-summary-signature.mjs";
11
11
  import { applyStackStallGuard } from "./stack-stall.mjs";
@@ -83,12 +83,12 @@ async function runAggregatePollCore(opts) {
83
83
  ...(pendingQuotaWarning && { quotaWarning: pendingQuotaWarning }),
84
84
  }, opts.merge);
85
85
  }
86
- const retryMs = opts.untilTerminal ? pollGraphQlRetryAfterMs(error) : null;
87
- if (retryMs === null || rateLimitRetries >= 1)
86
+ const retry = opts.untilTerminal ? pollRateLimitRetryAfterMs(error) : null;
87
+ if (retry === null || rateLimitRetries >= 1)
88
88
  throw error;
89
89
  rateLimitRetries += 1;
90
- process.stderr.write(`[aggregate poll tick ${tick} / +${Math.round((Date.now() - start) / 1000)}s] GraphQL rate limit — retrying in ${Math.round(retryMs / 1000)}s\n`);
91
- await sleep(retryMs);
90
+ process.stderr.write(formatRateLimitRetryLine(`aggregate poll tick ${tick}`, Math.round((Date.now() - start) / 1000), retry));
91
+ await sleep(retry.ms);
92
92
  continue;
93
93
  }
94
94
  const allTerminal = last.selection.kind === "stack"
@@ -142,7 +142,7 @@ async function runAggregatePollCore(opts) {
142
142
  const elapsedMs = Date.now() - start;
143
143
  const sleepMs = debounceUntil
144
144
  ? Math.min(intervalMs, Math.max(debounceUntil - Date.now(), 0))
145
- : graphqlQuotaPollIntervalMs(quotaBands, summarizeApiTelemetry()?.graphql, intervalMs, MAX_TIMER_MS);
145
+ : quotaPollIntervalMs(quotaBands, summarizeApiTelemetry(), intervalMs, MAX_TIMER_MS);
146
146
  if (!opts.untilTerminal && debounceUntil === null) {
147
147
  const remainingMs = timeoutMs - elapsedMs;
148
148
  if (remainingMs <= 0 || remainingMs + TIMER_DRIFT_TOLERANCE_MS < sleepMs) {
@@ -2,7 +2,7 @@ import { runIterate } from "./iterate/index.mjs";
2
2
  import { sleep } from "../util/sleep.mjs";
3
3
  import { withPollApiUsage } from "./poll-run.mjs";
4
4
  import { loadConfig } from "../config/load.mjs";
5
- import { graphqlQuotaPollIntervalMs, pollGraphQlRetryAfterMs } from "./poll-quota.mjs";
5
+ import { formatRateLimitRetryLine, pollRateLimitRetryAfterMs, quotaPollIntervalMs, } from "./poll-quota.mjs";
6
6
  import { writeDebounceProgress, writeWaitProgress } from "./poll-progress.mjs";
7
7
  const DEFAULT_POLL_DEBOUNCE_SECONDS = 60;
8
8
  const MAX_TIMER_MS = 2 ** 31 - 1;
@@ -54,12 +54,12 @@ async function runPollCore(opts) {
54
54
  return result;
55
55
  }
56
56
  catch (err) {
57
- const retryMs = untilTerminal ? pollGraphQlRetryAfterMs(err) : null;
58
- if (retryMs === null || rateLimitRetries >= 1)
57
+ const retry = untilTerminal ? pollRateLimitRetryAfterMs(err) : null;
58
+ if (retry === null || rateLimitRetries >= 1)
59
59
  throw err;
60
60
  rateLimitRetries += 1;
61
- process.stderr.write(`[poll tick ${tick} / +${Math.round((Date.now() - start) / 1000)}s] GraphQL rate limit — retrying in ${Math.round(retryMs / 1000)}s\n`);
62
- await sleep(retryMs);
61
+ process.stderr.write(formatRateLimitRetryLine(`poll tick ${tick}`, Math.round((Date.now() - start) / 1000), retry));
62
+ await sleep(retry.ms);
63
63
  const result = await iterateTick(fingerprintCache);
64
64
  rateLimitRetries = 0;
65
65
  return result;
@@ -101,7 +101,7 @@ async function runPollCore(opts) {
101
101
  if (pendingQuotaWarning === undefined)
102
102
  debounceUntil = null;
103
103
  const elapsedMs = Date.now() - start;
104
- const sleepMs = graphqlQuotaPollIntervalMs(quotaBands, lastResult.apiUsage?.graphql, intervalMs, MAX_TIMER_MS);
104
+ const sleepMs = quotaPollIntervalMs(quotaBands, lastResult.apiUsage, intervalMs, MAX_TIMER_MS);
105
105
  if (!untilTerminal) {
106
106
  const remainingMs = timeoutMs - elapsedMs;
107
107
  if (remainingMs <= 0 || remainingMs + TIMER_DRIFT_TOLERANCE_MS < sleepMs) {
@@ -126,7 +126,7 @@ async function runPollCore(opts) {
126
126
  !pastDebounce) {
127
127
  debounceUntil = null;
128
128
  const elapsedMs = Date.now() - start;
129
- const sleepMs = graphqlQuotaPollIntervalMs(quotaBands, lastResult.apiUsage?.graphql, intervalMs, MAX_TIMER_MS);
129
+ const sleepMs = quotaPollIntervalMs(quotaBands, lastResult.apiUsage, intervalMs, MAX_TIMER_MS);
130
130
  if (!untilTerminal) {
131
131
  const remainingMs = timeoutMs - elapsedMs;
132
132
  if (remainingMs <= 0 || remainingMs + TIMER_DRIFT_TOLERANCE_MS < sleepMs) {
@@ -0,0 +1,7 @@
1
+ import type { GraphqlQuotaWarningBand } from "../config/load.mts";
2
+ import type { ApiUsage, GraphqlQuotaWarning } from "../types.mts";
3
+ /** Claim GraphQL and REST core warnings separately, then present one result. */
4
+ export declare function selectQuotaWarning(key: {
5
+ owner: string;
6
+ repo: string;
7
+ }, bands: GraphqlQuotaWarningBand[], usage: ApiUsage, persist: boolean): Promise<GraphqlQuotaWarning | undefined>;
@@ -0,0 +1,20 @@
1
+ import { withGraphqlCredentialFingerprint } from "../github/api-telemetry.mjs";
2
+ import { composeQuotaWarning } from "../quota-budgets.mjs";
3
+ import { evaluateWorktreeGraphqlQuotaWarning } from "../state/graphql-quota-warnings.mjs";
4
+ /** Claim GraphQL and REST core warnings separately, then present one result. */
5
+ export async function selectQuotaWarning(key, bands, usage, persist) {
6
+ const core = usage.rest?.find((item) => item.resource === "core");
7
+ const graphqlWarning = usage.graphql
8
+ ? await evaluateWorktreeGraphqlQuotaWarning(key, bands, withGraphqlCredentialFingerprint(usage.graphql), persist)
9
+ : undefined;
10
+ const coreWarning = core
11
+ ? await evaluateWorktreeGraphqlQuotaWarning(key, bands, withGraphqlCredentialFingerprint(core), persist)
12
+ : undefined;
13
+ return composeQuotaWarning({
14
+ graphqlWarning,
15
+ coreWarning,
16
+ graphql: usage.graphql,
17
+ core,
18
+ bands,
19
+ });
20
+ }
@@ -0,0 +1,16 @@
1
+ import type { GraphqlQuotaWarningBand } from "./config/load.mts";
2
+ import type { ApiResourceUsage, GraphqlQuotaWarning } from "./types.mts";
3
+ type Usage = Pick<ApiResourceUsage, "resource" | "remaining" | "limit" | "used" | "resetAt">;
4
+ /**
5
+ * One warning for the budgets that are low on this tick. A GraphQL warning
6
+ * recommends REST only when REST core is still above its bands. When both are
7
+ * low, the result uses the later reset and does not recommend a switch.
8
+ */
9
+ export declare function composeQuotaWarning(input: {
10
+ graphqlWarning?: GraphqlQuotaWarning;
11
+ coreWarning?: GraphqlQuotaWarning;
12
+ graphql?: Usage;
13
+ core?: Usage;
14
+ bands: GraphqlQuotaWarningBand[];
15
+ }): GraphqlQuotaWarning | undefined;
16
+ export {};
@@ -0,0 +1,65 @@
1
+ /** True when this sample is at or below a configured remaining-percent band. */
2
+ function budgetBelowBand(usage, bands) {
3
+ if (usage === undefined || usage.limit <= 0 || bands.length === 0)
4
+ return false;
5
+ return bands.some((band) => usage.remaining * 100 <= usage.limit * band.remainingPercent);
6
+ }
7
+ /**
8
+ * One warning for the budgets that are low on this tick. A GraphQL warning
9
+ * recommends REST only when REST core is still above its bands. When both are
10
+ * low, the result uses the later reset and does not recommend a switch.
11
+ */
12
+ export function composeQuotaWarning(input) {
13
+ const graphqlBelow = budgetBelowBand(input.graphql, input.bands);
14
+ const coreBelow = budgetBelowBand(input.core, input.bands);
15
+ const graphql = input.graphqlWarning;
16
+ const core = input.coreWarning;
17
+ if (graphql && core)
18
+ return combineWarnings(graphql, core);
19
+ if (graphql && coreBelow && input.core)
20
+ return combineWarnings(graphql, describeBudget(input.core, input.bands));
21
+ if (core && graphqlBelow && input.graphql) {
22
+ return combineWarnings(core, describeBudget(input.graphql, input.bands));
23
+ }
24
+ return graphql ?? core;
25
+ }
26
+ function describeBudget(usage, bands) {
27
+ const active = bands
28
+ .filter((band) => usage.remaining * 100 <= usage.limit * band.remainingPercent)
29
+ .reduce((lowest, band) => (band.remainingPercent < lowest.remainingPercent ? band : lowest));
30
+ return {
31
+ resource: usage.resource === "core" ? "core" : "graphql",
32
+ thresholdPercent: active.remainingPercent,
33
+ remaining: usage.remaining,
34
+ limit: usage.limit,
35
+ ...(usage.used !== undefined && { used: usage.used }),
36
+ resetAt: usage.resetAt,
37
+ pollIntervalMinutes: active.pollIntervalMinutes,
38
+ pollTimeoutMinutes: active.pollIntervalMinutes * 2,
39
+ };
40
+ }
41
+ function combineWarnings(left, right) {
42
+ const later = left.resetAt >= right.resetAt ? left : right;
43
+ const interval = Math.max(left.pollIntervalMinutes, right.pollIntervalMinutes);
44
+ return {
45
+ resource: "combined",
46
+ thresholdPercent: Math.min(left.thresholdPercent, right.thresholdPercent),
47
+ remaining: later.remaining,
48
+ limit: later.limit,
49
+ ...(later.used !== undefined && { used: later.used }),
50
+ resetAt: later.resetAt,
51
+ pollIntervalMinutes: interval,
52
+ pollTimeoutMinutes: interval * 2,
53
+ budgets: [toBudget(left), toBudget(right)].sort((a, b) => a.resource.localeCompare(b.resource)),
54
+ };
55
+ }
56
+ function toBudget(warning) {
57
+ return {
58
+ resource: warning.resource === "core" ? "core" : "graphql",
59
+ thresholdPercent: warning.thresholdPercent,
60
+ remaining: warning.remaining,
61
+ limit: warning.limit,
62
+ ...(warning.used !== undefined && { used: warning.used }),
63
+ resetAt: warning.resetAt,
64
+ };
65
+ }
@@ -2,5 +2,15 @@ export function buildQuotaAwareContinuation(warning, prefix) {
2
2
  const interval = `${warning.pollIntervalMinutes}m`;
3
3
  const timeout = `${warning.pollTimeoutMinutes}m`;
4
4
  const resetTime = new Date(warning.resetAt * 1000).toISOString();
5
- return `${prefix} GitHub's GraphQL API quota is low (crossed the ${warning.thresholdPercent}% remaining threshold). Keep using pr-shepherd at the cadence below; for incidental PR operations that do not need Shepherd's full snapshot, prefer non-GraphQL \`gh\` CLI commands (e.g. \`gh pr view\`, \`gh pr review\`, \`gh api\` REST endpoints) — they draw on the separate REST budget, not the depleted GraphQL pool. Do not substitute \`gh pr checks\` or \`gh pr watch\` for the Shepherd loop. Resume full-cadence pr-shepherd after the GraphQL quota resets at ${resetTime}. If you must keep polling before then, poll no more often than every ${warning.pollIntervalMinutes} minutes. With a polling CLI command, preserve the other options, raise any shorter interval and timeout flags to at least \`--interval ${interval} --timeout ${timeout}\`, keep any longer cadence, and omit \`--timeout\` when using \`--until-terminal\`. With a single-tick CLI, API, or MCP call, wait at least ${warning.pollIntervalMinutes} minutes before the next tick.`;
5
+ const opening = warning.resource === "combined"
6
+ ? "GitHub's GraphQL and REST core quotas are both low. Keep using pr-shepherd at the cadence below. Do not shift incidental calls between GraphQL and REST."
7
+ : warning.resource === "core"
8
+ ? `GitHub's REST core quota is low (crossed the ${warning.thresholdPercent}% remaining threshold). Keep using pr-shepherd at the cadence below. Do not add incidental REST \`gh\` calls (\`gh pr view\`, \`gh pr review\`, \`gh api\`) while REST core is below its warning threshold.`
9
+ : `GitHub's GraphQL API quota is low (crossed the ${warning.thresholdPercent}% remaining threshold). Keep using pr-shepherd at the cadence below; for incidental PR operations that do not need Shepherd's full snapshot, prefer non-GraphQL \`gh\` CLI commands (e.g. \`gh pr view\`, \`gh pr review\`, \`gh api\` REST endpoints) — they draw on the separate REST budget, not the depleted GraphQL pool.`;
10
+ const resetLabel = warning.resource === "combined"
11
+ ? "both quotas have reset"
12
+ : warning.resource === "core"
13
+ ? "the REST core quota resets"
14
+ : "the GraphQL quota resets";
15
+ return `${prefix} ${opening} Do not substitute \`gh pr checks\` or \`gh pr watch\` for the Shepherd loop. Resume full-cadence pr-shepherd after ${resetLabel} at ${resetTime}. If you must keep polling before then, poll no more often than every ${warning.pollIntervalMinutes} minutes. With a polling CLI command, preserve the other options, raise any shorter interval and timeout flags to at least \`--interval ${interval} --timeout ${timeout}\`, keep any longer cadence, and omit \`--timeout\` when using \`--until-terminal\`. With a single-tick CLI, API, or MCP call, wait at least ${warning.pollIntervalMinutes} minutes before the next tick.`;
6
16
  }
@@ -32,7 +32,7 @@ export function evaluateGraphqlQuotaWarning(bands, sample, previous, observedAt
32
32
  return { state };
33
33
  return {
34
34
  warning: {
35
- resource: "graphql",
35
+ resource: sample.resource === "core" ? "core" : "graphql",
36
36
  thresholdPercent: active.remainingPercent,
37
37
  remaining: effective.remaining,
38
38
  limit: effective.limit,
@@ -1,6 +1,6 @@
1
1
  import type { GraphqlQuotaWarningBand } from "../config/load.mts";
2
- import type { GraphqlQuotaWarning, GraphqlApiUsage } from "../types.mts";
2
+ import type { ApiResourceUsage, GraphqlQuotaWarning } from "../types.mts";
3
3
  export declare function evaluateWorktreeGraphqlQuotaWarning(key: {
4
4
  owner: string;
5
5
  repo: string;
6
- }, bands: GraphqlQuotaWarningBand[], sample: GraphqlApiUsage, persist: boolean, now?: number): Promise<GraphqlQuotaWarning | undefined>;
6
+ }, bands: GraphqlQuotaWarningBand[], sample: ApiResourceUsage, persist: boolean, now?: number): Promise<GraphqlQuotaWarning | undefined>;
@@ -10,9 +10,10 @@ const sessionStates = new Map();
10
10
  export async function evaluateWorktreeGraphqlQuotaWarning(key, bands, sample, persist, now = Date.now() / 1000) {
11
11
  if (bands.length === 0)
12
12
  return undefined;
13
- const path = await warningStatePath(key);
13
+ const quotaResource = sample.resource === "core" ? "core" : "graphql";
14
+ const path = await warningStatePath(key, quotaResource);
14
15
  if (path === undefined) {
15
- const sessionKey = `${key.owner}/${key.repo}`;
16
+ const sessionKey = `${key.owner}/${key.repo}/${quotaResource}`;
16
17
  if (!persist) {
17
18
  return evaluateGraphqlQuotaWarning(bands, sample, sessionStates.get(sessionKey) ?? null, now)
18
19
  .warning;
@@ -50,7 +51,7 @@ async function serializeStateUpdate(key, update) {
50
51
  });
51
52
  return result;
52
53
  }
53
- async function warningStatePath(key) {
54
+ async function warningStatePath(key, resource) {
54
55
  let worktreeKey;
55
56
  try {
56
57
  worktreeKey = await getWorktreeKey();
@@ -58,7 +59,7 @@ async function warningStatePath(key) {
58
59
  catch {
59
60
  return undefined;
60
61
  }
61
- return join(resolveRepoStateDir(key), "worktrees", `${worktreeKey}-graphql-quota-warnings.json`);
62
+ return join(resolveRepoStateDir(key), "worktrees", `${worktreeKey}-${resource}-quota-warnings.json`);
62
63
  }
63
64
  async function readState(path) {
64
65
  try {
@@ -5,6 +5,8 @@ export interface ApiResourceUsage {
5
5
  used?: number;
6
6
  remaining: number;
7
7
  resetAt: number;
8
+ /** Truncated SHA-256 of the credential. Quota state only; summarized usage omits it. */
9
+ credentialFingerprint?: string;
8
10
  }
9
11
  export interface GraphqlApiUsage extends ApiResourceUsage {
10
12
  /** Exact sum reported by rateLimit.cost for GraphQL queries in this command. */
@@ -13,8 +15,15 @@ export interface GraphqlApiUsage extends ApiResourceUsage {
13
15
  unmeasuredRequestCount: number;
14
16
  /** Exact sum reported by rateLimit.nodeCount for measured GraphQL queries. */
15
17
  nodeCount: number;
16
- /** Truncated SHA-256 of the credential. Quota state only; summarized usage omits it. */
17
- credentialFingerprint?: string;
18
+ }
19
+ /** One budget inside a combined GraphQL and REST core warning. */
20
+ export interface QuotaWarningBudget {
21
+ resource: "graphql" | "core";
22
+ thresholdPercent: number;
23
+ remaining: number;
24
+ limit: number;
25
+ used?: number;
26
+ resetAt: number;
18
27
  }
19
28
  export interface ApiUsage {
20
29
  credentialSources: string[];
@@ -22,7 +31,7 @@ export interface ApiUsage {
22
31
  rest?: ApiResourceUsage[];
23
32
  }
24
33
  export interface GraphqlQuotaWarning {
25
- resource: "graphql";
34
+ resource: "graphql" | "core" | "combined";
26
35
  thresholdPercent: number;
27
36
  remaining: number;
28
37
  limit: number;
@@ -30,4 +39,6 @@ export interface GraphqlQuotaWarning {
30
39
  resetAt: number;
31
40
  pollIntervalMinutes: number;
32
41
  pollTimeoutMinutes: number;
42
+ /** Both budgets when GraphQL and REST core are low together. */
43
+ budgets?: QuotaWarningBudget[];
33
44
  }
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "pr-shepherd",
3
- "version": "0.53.0",
3
+ "version": "0.53.1",
4
4
  "description": "Autonomous PR CI monitor and review-comment resolver for agentic coding tools",
5
5
  "keywords": [
6
6
  "automation",
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "pr-shepherd",
3
- "version": "0.53.0",
3
+ "version": "0.53.1",
4
4
  "description": "Autonomous PR CI monitor and review-comment resolver for Codex.",
5
5
  "author": {
6
6
  "name": "Jonathan Ong",
@@ -2,7 +2,7 @@
2
2
  "mcpServers": {
3
3
  "pr-shepherd": {
4
4
  "command": "npx",
5
- "args": ["--yes", "--package", "pr-shepherd@0.53.0", "pr-shepherd-mcp"]
5
+ "args": ["--yes", "--package", "pr-shepherd@0.53.1", "pr-shepherd-mcp"]
6
6
  }
7
7
  }
8
8
  }
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "pr-shepherd": {
3
3
  "command": "npx",
4
- "args": ["--yes", "--package", "pr-shepherd@0.53.0", "pr-shepherd-mcp"]
4
+ "args": ["--yes", "--package", "pr-shepherd@0.53.1", "pr-shepherd-mcp"]
5
5
  }
6
6
  }