From 17dcb9d178dd786dba4192ea3e590d49b10a7cc5 Mon Sep 17 00:00:00 2001
From: Marcos <marcos@tracksteel.com.br>
Date: Sun, 22 Mar 2026 16:51:21 -0300
Subject: [PATCH] Increase Ollama timeout to 180s and add num_ctx

---
 ai_agent.py      | 8 +++++---
 llm_providers.py | 7 ++++---
 2 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/ai_agent.py b/ai_agent.py
index cbb7f71..eae44ae 100644
--- a/ai_agent.py
+++ b/ai_agent.py
@@ -24,12 +24,14 @@ def get_llm_response(prompt: str, provider: str, cfg: dict) -> str:
     
     elif provider == "ollama":
         ollama_host = os.getenv("OLLAMA_HOST", "http://ollama:11434")
+        model = os.getenv("OLLAMA_MODEL", "qwen2.5-coder:1.5b")
         try:
             res = requests.post(f"{ollama_host}/api/generate", json={
-                "model": os.getenv("OLLAMA_MODEL", "qwen2.5-coder:1.5b"),
+                "model": model,
                 "prompt": prompt,
-                "stream": False
-            }, timeout=30)
+                "stream": False,
+                "options": {"num_ctx": 4096}
+            }, timeout=180)
             if res.status_code == 200:
                 return res.json().get("response", "")
             return f"Erro Ollama (Status {res.status_code}): {res.text}"
diff --git a/llm_providers.py b/llm_providers.py
index a1fb928..ff64a72 100644
--- a/llm_providers.py
+++ b/llm_providers.py
@@ -337,20 +337,21 @@ def _call_ollama(model: str, prompt: str, system_prompt: str = None) -> str:
     payload = {
         "model": model,
         "prompt": prompt,
-        "stream": False
+        "stream": False,
+        "options": {"num_ctx": 4096}
     }
     
     if system_prompt:
         payload["system"] = system_prompt
     
     try:
-        res = requests.post(f"{endpoint}/api/generate", json=payload, timeout=120)
+        res = requests.post(f"{endpoint}/api/generate", json=payload, timeout=180)
         
         if res.status_code == 200:
             return res.json().get("response", "")
         return f"Erro Ollama: {res.status_code} - {res.text}"
     except requests.exceptions.Timeout:
-        return f"[TIMEOUT] Ollama não respondeu em 120s. Verifique se o serviço está rodando em {endpoint}"
+        return f"[TIMEOUT] Ollama não respondeu em 180s. Verifique se o modelo está carregado em {endpoint}"
     except requests.exceptions.ConnectionError:
         return f"[CONNECTION ERROR] Não conseguiu conectar ao Ollama em {endpoint}. Verifique se o container Ollama está na mesma rede Docker."
     except Exception as e: