From 0ef2b715c46c7ee0e2a49d66d0ef1ec2089e5a34 Mon Sep 17 00:00:00 2001
From: asepharyana <superaseph@gmail.com>
Date: Sat, 1 Aug 2026 15:00:59 +0700
Subject: [PATCH] =?UTF-8?q?fix(gateway):=20enable=20stream=20for=20all=20L?=
 =?UTF-8?q?LM=20calls=20=E2=80=94=20router=20always=20streams=20SSE?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Audit lanjutan: 6x 'LLM API request failed: Request was aborted' per jam.
Root cause: 9router/omniroute SELALU balas SSE (data: chunks) walau request
tanpa stream:true — SDK OpenAI non-stream menunggu FULL body sebelum parse,
jadi batch moderasi besar yang upstream-nya lambat kena timeout 30-60s dan
di-abort. llmClient sudah punya agregasi streaming (chunks → ChatCompletion).

Fix: stream:true di llmCaller (moderasi batch/individual), llmVision,
cultureLearner, userProfileLearner. Verified: SDK stream test 806ms vs
sebelumnya abort. Caller lain (recovery worker dll) lewat llmCaller sama.
---
 .../src/modules/ai-moderation/cultureLearner.ts            | 1 +
 .../discord-gateway/src/modules/ai-moderation/llmCaller.ts | 7 +++++++
 .../discord-gateway/src/modules/ai-moderation/llmClient.ts | 1 +
 .../src/modules/ai-moderation/userProfileLearner.ts        | 1 +
 4 files changed, 10 insertions(+)

diff --git a/services/discord-gateway/src/modules/ai-moderation/cultureLearner.ts b/services/discord-gateway/src/modules/ai-moderation/cultureLearner.ts
index e71d537..49b2c2a 100644
--- a/services/discord-gateway/src/modules/ai-moderation/cultureLearner.ts
+++ b/services/discord-gateway/src/modules/ai-moderation/cultureLearner.ts
@@ -87,6 +87,7 @@ hal dasar ke newbie tanpa judge."`;
       max_tokens: 500,
       temperature: 0.7, // Higher temp for summarization
       retries: 2,
+      stream: true, // router always streams SSE; non-stream waits for full body and times out
     });
 
     if (!completion) throw new Error("Empty response from LLM");
diff --git a/services/discord-gateway/src/modules/ai-moderation/llmCaller.ts b/services/discord-gateway/src/modules/ai-moderation/llmCaller.ts
index 4ed3eb2..f2f471a 100644
--- a/services/discord-gateway/src/modules/ai-moderation/llmCaller.ts
+++ b/services/discord-gateway/src/modules/ai-moderation/llmCaller.ts
@@ -79,6 +79,13 @@ export async function callModerationLLM(
             jsonResponse: { type: "json_object" },
             retries: 0,
             signal,
+            // Router (9router/omniroute) always streams SSE even when the
+            // request omits `stream`. In non-stream mode the OpenAI SDK waits
+            // for the FULL body before parsing, so slow/long upstream streams
+            // hit the 30s/60s timeout and abort mid-generation. Streaming mode
+            // consumes chunks incrementally — timeout only fires on a real
+            // stall. llmClient aggregates the stream into a ChatCompletion.
+            stream: true,
           });
 
           if (!completion)
diff --git a/services/discord-gateway/src/modules/ai-moderation/llmClient.ts b/services/discord-gateway/src/modules/ai-moderation/llmClient.ts
index 3969423..230b4a5 100644
--- a/services/discord-gateway/src/modules/ai-moderation/llmClient.ts
+++ b/services/discord-gateway/src/modules/ai-moderation/llmClient.ts
@@ -272,6 +272,7 @@ export async function llmVision(
     temperature: 0.1,
     top_p: 0.9,
     retries: 0,
+    stream: true, // router always streams SSE; non-stream waits for full body and times out
   });
 
   if (!completion) return null;
diff --git a/services/discord-gateway/src/modules/ai-moderation/userProfileLearner.ts b/services/discord-gateway/src/modules/ai-moderation/userProfileLearner.ts
index 18e4400..ac8d8aa 100644
--- a/services/discord-gateway/src/modules/ai-moderation/userProfileLearner.ts
+++ b/services/discord-gateway/src/modules/ai-moderation/userProfileLearner.ts
@@ -116,6 +116,7 @@ atau konten SARA, itu akan SANGAT tidak sesuai dengan karakternya dan patut dicu
       max_tokens: 2000,
       temperature: 0.7, // Higher temp for summarization
       retries: 2,
+      stream: true, // router always streams SSE; non-stream waits for full body and times out
     });
 
     if (!completion) throw new Error("Empty response from LLM");