ollama service exposed to host and tunnelled via tailnet to securely serve ollama inference to dev machine in tailnet (tailscale internal network)

Author: himmat12

.env.example

@@ -1,4 +1,4 @@
-# Core config (both modes)
+ENV=dev # Set to 'prod' for production environment
 OLLAMA_HOST=http://ollama:11434
 AGENT_WORKSPACE=/workspace
 OLLAMA_MODEL=llama3.2:1b

agent/app/agentic_loop.py

@@ -12,7 +12,7 @@
 
 load_dotenv()
 
-OLLAMA_HOST = "http://ollama:11434"
+OLLAMA_HOST = lambda: "http://100.88.9.73:11434" if os.getenv("ENV") == "dev" else "http://ollama:11434"
 OLLAMA_MODEL = os.getenv("OLLAMA_MODEL", "llama3.2:3b")
 
 
@@ -21,7 +21,7 @@ def ollama_generate(prompt: str, is_json_response: bool = True):
     payload = {
         "model": OLLAMA_MODEL,
         "prompt": prompt,
-        "stream": True,
+        "stream": False,
     }
 
     if is_json_response:

docker-compose.prod.yaml

@@ -1,6 +1,7 @@
 services:
   ollama:
-    ports: []
+    ports: 
+      - "127.0.0.1:11434:11434"
 
   agent:
     image: ghcr.io/himmat12/local-code-agent:latest