Skip to content

Improve Mistral rate-limit handling and add cooldown probe scripts #69

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Android-PowerUser merged 2 commits into from
Mar 31, 2026
Merged

Improve Mistral rate-limit handling and add cooldown probe scripts #69

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
8993e22
Honor Mistral server retry headers for cooldown scheduling
Android-PowerUser Mar 31, 2026
fe8b023
Merge branch 'codex/analyze-technical-debt-and-complexity-o6bm9p' int…
Android-PowerUser Mar 31, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
32 changes: 28 additions & 4 deletions app/src/main/kotlin/com/google/ai/sample/feature/multimodal/PhotoReasoningViewModel.kt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -71,6 +71,7 @@ import com.google.ai.sample.webrtc.WebRTCSender
import com.google.ai.sample.webrtc.SignalingClient
import org.webrtc.IceCandidate
import kotlin.math.max
import kotlin.math.roundToLong

class PhotoReasoningViewModel(
application: Application,
Expand Down Expand Up @@ -183,11 +184,11 @@ class PhotoReasoningViewModel(
// to avoid re-executing already-executed commands
private var incrementalCommandCount = 0

// Mistral rate limiting per API key (4 seconds between requests with same key)
// Mistral rate limiting per API key (1.5 seconds between requests with same key)
private val mistralNextAllowedRequestAtMsByKey = mutableMapOf<String, Long>()
private var lastMistralTokenTimeMs = 0L
private var lastMistralTokenKey: String? = null
private val MISTRAL_MIN_INTERVAL_MS = 4000L
private val MISTRAL_MIN_INTERVAL_MS = 1500L

// Accumulated full text during streaming for incremental command parsing
private var streamingAccumulatedText = StringBuilder()
Expand Down Expand Up @@ -1140,11 +1141,31 @@ private fun reasonWithMistral(
mistralNextAllowedRequestAtMsByKey[key] = max(existing, nextAllowedAt)
}

fun markKeyCooldown(key: String, referenceTimeMs: Long, extraDelayMs: Long) {
val normalizedExtraDelay = extraDelayMs.coerceAtLeast(0L)
val nextAllowedAt = referenceTimeMs + max(MISTRAL_MIN_INTERVAL_MS, normalizedExtraDelay)
val existing = mistralNextAllowedRequestAtMsByKey[key] ?: 0L
mistralNextAllowedRequestAtMsByKey[key] = max(existing, nextAllowedAt)
}

fun remainingWaitForKeyMs(key: String, nowMs: Long): Long {
val nextAllowedAt = mistralNextAllowedRequestAtMsByKey[key] ?: 0L
return (nextAllowedAt - nowMs).coerceAtLeast(0L)
}

fun parseRetryAfterMs(headerValue: String?): Long? {
if (headerValue.isNullOrBlank()) return null
val seconds = headerValue.trim().toDoubleOrNull() ?: return null
return (seconds * 1000.0).roundToLong().coerceAtLeast(0L)
}

fun parseRateLimitResetDelayMs(response: okhttp3.Response, nowMs: Long): Long? {
val resetHeader = response.header("x-ratelimit-reset") ?: return null
val resetEpochSeconds = resetHeader.trim().toLongOrNull() ?: return null
val resetMs = resetEpochSeconds * 1000L
return (resetMs - nowMs).coerceAtLeast(0L)
}

fun isRetryableMistralFailure(code: Int): Boolean {
return code == 429 || code >= 500
}
Expand Down Expand Up @@ -1176,7 +1197,10 @@ private fun reasonWithMistral(
try {
val attemptResponse = client.newCall(buildRequest(selectedKey)).execute()
val requestEndMs = System.currentTimeMillis()
markKeyCooldown(selectedKey, requestEndMs)
val retryAfterMs = parseRetryAfterMs(attemptResponse.header("Retry-After"))
val resetDelayMs = parseRateLimitResetDelayMs(attemptResponse, requestEndMs)
val serverRequestedDelayMs = max(retryAfterMs ?: 0L, resetDelayMs ?: 0L)
markKeyCooldown(selectedKey, requestEndMs, serverRequestedDelayMs)

if (attemptResponse.isSuccessful) {
response = attemptResponse
Expand All @@ -1195,7 +1219,7 @@ private fun reasonWithMistral(
consecutiveFailures++
withContext(Dispatchers.Main) {
replaceAiMessageText(
"Mistral temporär nicht verfügbar (Versuch $consecutiveFailures/$maxAttempts). Wiederhole...",
"Mistral temporär nicht verfügbar (Versuch $consecutiveFailures/$maxAttempts). Warte auf Server-Rate-Limit und wiederhole...",
isPending = true
)
}
Expand Down