From af9da3c912e5bf8a2799f8b1263df3a842a07f8b Mon Sep 17 00:00:00 2001
From: kami <kamixtrash@gmail.com>
Date: Wed, 1 Jul 2026 00:17:15 +0400
Subject: [PATCH] fix(inference): raise llama HTTP request timeout 600s ->
 1800s

Align the provider's HTTP client timeout with the orchestrator per-stage
timeout (1_800_000ms). A slow local reasoning model with a large prompt and a
multi-thousand-token reasoning budget can exceed 10min on a single call; the
old 600s HTTP timeout failed the inference before the stage timeout applied,
exhausting retries on timeouts alone.
---
 .../inference/llama/cpp/LlamaCppInferenceProvider.kt         | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/infrastructure/inference/llama_cpp/src/main/kotlin/com/correx/infrastructure/inference/llama/cpp/LlamaCppInferenceProvider.kt b/infrastructure/inference/llama_cpp/src/main/kotlin/com/correx/infrastructure/inference/llama/cpp/LlamaCppInferenceProvider.kt
index 68d40cac..6ea2441a 100644
--- a/infrastructure/inference/llama_cpp/src/main/kotlin/com/correx/infrastructure/inference/llama/cpp/LlamaCppInferenceProvider.kt
+++ b/infrastructure/inference/llama_cpp/src/main/kotlin/com/correx/infrastructure/inference/llama/cpp/LlamaCppInferenceProvider.kt
@@ -31,7 +31,10 @@ import kotlinx.serialization.encodeToString
 import kotlinx.serialization.json.Json
 import org.slf4j.LoggerFactory
 
-private const val DEFAULT_REQUEST_TIMEOUT_MS = 600_000L
+// Aligned with the orchestrator's per-stage timeout (1_800_000ms). A slow local reasoning model
+// (large prompt + a multi-thousand-token reasoning budget) can take >10min for a single call;
+// a shorter HTTP timeout here fails the call before the stage timeout ever applies.
+private const val DEFAULT_REQUEST_TIMEOUT_MS = 1_800_000L
 
 private fun defaultHttpClient(): HttpClient = HttpClient(CIO) {
     install(ContentNegotiation) {