diff --git a/services/imajin-prompt/config.yaml b/services/imajin-prompt/config.yaml
index 795a414a..08486841 100644
--- a/services/imajin-prompt/config.yaml
+++ b/services/imajin-prompt/config.yaml
@@ -6,10 +6,10 @@
 # See service/src/llm/OLLAMA_REMOVED.md for migration details.
 
 llm:
-  # Backend selection: "http" (llama-http for mistral3 arch), "llama" (GPUBoss-coordinated GGUF)
-  # Note: Using "http" backend to connect to llama-http service
-  # Reason: Ministral-3 models use mistral3 architecture unsupported by llama-cpp-python
-  backend: http
+  # Backend selection: "http" (llama-http for mistral3 arch), "ollama" (Ollama service), "llama" (GPUBoss-coordinated GGUF)
+  # Note: Using "ollama" backend temporarily - Ollama has ministral-3:14b loaded
+  # Reason: llama-http service not running, Ollama IS running on port 11434
+  backend: ollama
 
   # Model ID from manifest.json (used by "llama" backend only)
   # Available: ministral-14b-reasoning (Ministral-3:14b GGUF)