Implement CodeRabbit rate limiting, backoff, and idempotency patterns.
Use when handling rate limit errors, implementing retry logic,
or optimizing API request throughput for CodeRabbit.
Trigger with phrases like "coderabbit rate limit", "coderabbit throttling",
"coderabbit 429", "coderabbit retry", "coderabbit backoff".
CodeRabbit rate limits apply at two levels: (1) CodeRabbit's own processing limits on how many reviews it can run concurrently, and (2) GitHub API rate limits when you build automation that queries CodeRabbit review data. This skill covers both and provides patterns for handling limits gracefully.
Prerequisites
CodeRabbit installed on repository
GitHub CLI (gh) or API access for automation
Understanding of GitHub rate limit headers
Rate Limit Tiers
CodeRabbit Review Processing
Factor
Limit
Notes
Concurrent reviews per org
Varies by plan
Free: 1, Pro: 5, Enterprise: custom
Max PR size
~3000 files
Larger PRs may timeout
Re-review cooldown
~30 seconds
Between @coderabbitai full review commands
Command rate
~10/minute/repo
PR comment commands
GitHub API (Affects Automation Scripts)
Tier
Rate Limit
Reset Window
Unauthenticated
60 req/hour
Rolling
Personal Access Token
5,000 req/hour
Rolling
GitHub App
5,000 req/hour/installation
Rolling
gh CLI
5,000 req/hour
Rolling
Instructions
Step 1: Check Current GitHub API Rate Limit
set -euo pipefail
# Check your current rate limit status
gh api rate_limit --jq '{
core: {
limit: .resources.core.limit,
remaining: .resources.core.remaining,
reset: (.resources.core.reset | todate)
},
search: {
limit: .resources.search.limit,
remaining: .resources.search.remaining,
reset: (.resources.search.reset | todate)
}
}'
Step 2: Handle Rate Limits in Automation Scripts
#!/bin/bash
# rate-safe-query.sh - GitHub API queries with rate limit awareness
set -euo pipefail
ORG="${1:?Usage: $0 <org> <repo>}"
REPO="${2:?Usage: $0 <org> <repo>}"
# Check remaining rate limit before bulk queries
REMAINING=$(gh api rate_limit --jq '.resources.core.remaining')
echo "GitHub API calls remaining: $REMAINING"
if [ "$REMAINING" -lt 100 ]; then
RESET=$(gh api rate_limit --jq '.resources.core.reset | todate')
echo "WARNING: Low rate limit. Resets at $RESET"
echo "Consider waiting or reducing query scope."
exit 1
fi
# Safe pagination: process in small batches
PAGE=1
PER_PAGE=10
while true; do
RESULT=$(gh api "repos/$ORG/$REPO/pulls?state=closed&per_page=$PER_PAGE&page=$PAGE" --jq 'length')
[ "$RESULT" -eq 0 ] && break
gh api "repos/$ORG/$REPO/pulls?state=closed&per_page=$PER_PAGE&page=$PAGE" \
--jq '.[].number' | while read -r PR_NUM; do
# Process each PR
echo "Processing PR #$PR_NUM"
# Rate-limit-safe: check remaining before each sub-query
SUB_REMAINING=$(gh api rate_limit --jq '.resources.core.remaining')
if [ "$SUB_REMAINING" -lt 50 ]; then
echo "Rate limit low ($SUB_REMAINING remaining). Pausing..."
sleep 60
fi
gh api "repos/$ORG/$REPO/pulls/$PR_NUM/reviews" \
--jq '[.[] | select(.user.login=="coderabbitai[bot]")] | length' 2>/dev/null
done
PAGE=$((PAGE + 1))
[ "$PAGE" -gt 5 ] && break # Safety limit
done
Step 3: Handle CodeRabbit Command Rate Limits
# If you send too many @coderabbitai commands in quick succession,
# CodeRabbit may not respond to all of them.
# Best practices:
1. Wait for CodeRabbit to finish one command before sending another
2. Don't spam "full review" -- one is enough, it processes the latest
3. Use "summary" instead of "full review" if you just want the walkthrough
4. Wait 2-5 minutes after PR push for the initial review before using commands
# Rate limit symptoms:
# - CodeRabbit doesn't respond to a command
# - Review appears incomplete
# - Multiple partial reviews on the same PR
# Fix: Wait 1-2 minutes and resend the command once.
Step 4: Efficient Bulk Queries with GraphQL
set -euo pipefail
ORG="${1:-your-org}"
REPO="${2:-your-repo}"
# GraphQL uses far fewer API calls than REST for bulk data
# One GraphQL call = data that would take 20+ REST calls
gh api graphql -f query='
query($owner: String!, $repo: String!) {
repository(owner: $owner, name: $repo) {
pullRequests(last: 20, states: [MERGED, CLOSED]) {
nodes {
number
title
reviews(first: 5) {
nodes {
author { login }
state
submittedAt
}
}
}
}
}
}' -f owner="$ORG" -f repo="$REPO" --jq '
.data.repository.pullRequests.nodes[] |
{
pr: .number,
title: .title,
coderabbit_reviews: [.reviews.nodes[] | select(.author.login == "coderabbitai")] | length,
coderabbit_state: ([.reviews.nodes[] | select(.author.login == "coderabbitai")] | last | .state) // "none"
}'
Step 5: Cache CodeRabbit Metrics
#!/bin/bash
# cache-coderabbit-metrics.sh - Cache review data to avoid repeated API calls
set -euo pipefail
ORG="${1:?Usage: $0 <org> <repo>}"
REPO="${2:?Usage: $0 <org> <repo>}"
CACHE_FILE="/tmp/coderabbit-metrics-$ORG-$REPO.json"
CACHE_TTL=3600 # 1 hour
# Check cache freshness
if [ -f "$CACHE_FILE" ]; then
CACHE_AGE=$(( $(date +%s) - $(stat -c %Y "$CACHE_FILE" 2>/dev/null || stat -f %m "$CACHE_FILE") ))
if [ "$CACHE_AGE" -lt "$CACHE_TTL" ]; then
echo "Using cached data (age: ${CACHE_AGE}s)"
cat "$CACHE_FILE"
exit 0
fi
fi
echo "Fetching fresh data..."
METRICS=$(gh api graphql -f query='
query($owner: String!, $repo: String!) {
repository(owner: $owner, name: $repo) {
pullRequests(last: 50, states: [MERGED, CLOSED]) {
totalCount
nodes {
number
reviews(first: 5) {
nodes {
author { login }
state
}
}
}
}
}
}' -f owner="$ORG" -f repo="$REPO" --jq '
.data.repository.pullRequests | {
total: .totalCount,
reviewed: [.nodes[] | select([.reviews.nodes[] | select(.author.login == "coderabbitai")] | length > 0)] | length,
approved: [.nodes[] | select([.reviews.nodes[] | select(.author.login == "coderabbitai" and .state == "APPROVED")] | length > 0)] | length
}')
echo "$METRICS" | tee "$CACHE_FILE"
