1 files changed, 77 insertions, 0 deletions
diff --git a/internal/retry/backoff.go b/internal/retry/backoff.go
new file mode 100644
index 0000000..b91abc4
--- /dev/null
+++ b/internal/retry/backoff.go
@@ -0,0 +1,77 @@
+// Package retry provides exponential-backoff retry helpers used across the
+// codebase for rate-limit-aware HTTP/subprocess calls.
+package retry
+
+import (
+	"context"
+	"fmt"
+	"regexp"
+	"strconv"
+	"strings"
+	"time"
+)
+
+var retryAfterRe = regexp.MustCompile(`(?i)retry[-_ ]after[:\s]+(\d+)`)
+
+const maxBackoffDelay = 5 * time.Minute
+
+// IsRateLimitError returns true if err looks like a transient rate-limit
+// (e.g. HTTP 429, "too many requests", "overloaded") that is worth retrying.
+func IsRateLimitError(err error) bool {
+	if err == nil {
+		return false
+	}
+	msg := strings.ToLower(err.Error())
+	return strings.Contains(msg, "rate limit") ||
+		strings.Contains(msg, "too many requests") ||
+		strings.Contains(msg, "429") ||
+		strings.Contains(msg, "overloaded")
+}
+
+// ParseRetryAfter extracts a Retry-After duration from an error message.
+// Returns 0 if no retry-after value is found.
+func ParseRetryAfter(msg string) time.Duration {
+	m := retryAfterRe.FindStringSubmatch(msg)
+	if m == nil {
+		return 0
+	}
+	secs, err := strconv.Atoi(m[1])
+	if err != nil || secs <= 0 {
+		return 0
+	}
+	return time.Duration(secs) * time.Second
+}
+
+// RunWithBackoff calls fn repeatedly on rate-limit errors, using exponential backoff.
+// maxRetries is the max number of retry attempts (not counting the initial call).
+// baseDelay is the initial backoff duration (doubled each retry).
+func RunWithBackoff(ctx context.Context, maxRetries int, baseDelay time.Duration, fn func() error) error {
+	var lastErr error
+	for attempt := 0; attempt <= maxRetries; attempt++ {
+		lastErr = fn()
+		if lastErr == nil {
+			return nil
+		}
+		if !IsRateLimitError(lastErr) {
+			return lastErr
+		}
+		if attempt == maxRetries {
+			break
+		}
+
+		delay := baseDelay * (1 << attempt)
+		if delay > maxBackoffDelay {
+			delay = maxBackoffDelay
+		}
+		if ra := ParseRetryAfter(lastErr.Error()); ra > 0 {
+			delay = ra
+		}
+
+		select {
+		case <-ctx.Done():
+			return fmt.Errorf("context cancelled during rate-limit backoff: %w", ctx.Err())
+		case <-time.After(delay):
+		}
+	}
+	return lastErr
+}