updated core logic local

core_logic_local.py

@@ -7,20 +7,71 @@ Max Tokens: Increased for local version since there is neither the cost is incur
 . perform  thorough code review, 
 . write deeper code analysis, 
 . produce comprehensive solutions
+
+# /v1 Necessity: The /v1 is essential in the base_url for the OpenAI library to correctly route requests to Ollama's API; even though Chrome shows "Ollama is running" message at http://127.0.01:11434, i.e., without "/v1".
+
+"First Principles" breakdown of why this is necessary:
+
+1. The Browser vs. The API
+When visiting 127.0.0.1:11434 in Chrome, we hit the Base URL, Ollama sends back that simple text message just to confirm the service is alive.
+
+However, Python code doesn't just check if Ollama is alive, it tries to have a conversation; and for that, it needs to talk to a specific Endpoint (a specific door in the building).
+
+2. OpenAI Compatibility (The Industry Standard)
+Ollama was designed to be a "drop-in replacement" for OpenAI. Almost every AI library (like the openai Python library) expects a standard URL structure called the OpenAI Chat Completions API.
+
+The standard structure looks like this:
+
+Base URL: http://localhost:11434
+
+Version Prefix: /v1
+
+Action: /chat/completions
+
+When we set base_url='http://localhost:11434/v1', the OpenAI library automatically attaches /chat/completions to the end of it.
+
+3. What happens if "/v1" is removed?
+The library will try to send the data to http://localhost:11434/chat/completions, but because that URL is missing the "/v1" prefix, Ollama’s "OpenAI Compatibility" layer won't recognize the request, and either a "404 Not Found" or a "405 Method Not Allowed" may be encountered.
+
+Summary Checklist:
+In Chrome: Use 127.0.0.1:11434 - to see if it's on.
+In Python Code: Use 127.0.0.1:11434/v1  - to actually send prompts.
+
 """
 
 from openai import OpenAI
 from tools import web_search, parse_file
 import os
+import socket
+
+def get_base_url():
+    # Check if we are inside WSL
+    if os.path.exists('/proc/version'):
+        with open('/proc/version', 'r') as f:
+            if 'microsoft' in f.read().lower():
+                # if running the script from inside the Ubuntu (WSL) terminal, point to the Windows host
+                return "http://172.17.0.1:11434/v1"
+    # Otherwise, assume we are on the native Windows host, running the script from Windows Powershell/CMD, and point to localhost
+    return "http://127.0.0.1:11434/v1"
 
 # Ollama serves an OpenAI-compatible API locally at port 11434
 client = OpenAI(
-    base_url='http://localhost:11434/v1',
+    base_url=get_base_url(),
     api_key='ollama', # Required but ignored by Ollama
 )
 
+"""
+client = OpenAI(
+    base_url='http://localhost:11434/v1',
+    api_key="ollama"
+)
+"""
+
 # Use local model served by Ollama. Make sure to run: ollama serve gemma4
-model = "gemma4:latest"
+#model = "gemma4:latest"
+model = "llama3:latest" # better than llama3.2:latest and phi3:latest
+#model = "llama3.2:latest"
+#model = "phi3:latest"
 
 SYSTEM_PROMPT = """
 You are the 'Silicon Architect' — a full-stack, master-stroke creative genius in AI Engineering and Technical Architecture.