sunholo 0.66.18__tar.gz → 0.66.21__tar.gz

{sunholo-0.66.18 → sunholo-0.66.21}/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: sunholo
-Version: 0.66.18
+Version: 0.66.21
 Summary: Large Language Model DevOps - a package to help deploy LLMs to the Cloud.
 Home-page: https://github.com/sunholo-data/sunholo-py
-Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.66.18.tar.gz
+Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.66.21.tar.gz
 Author: Holosun ApS
 Author-email: multivac@sunholo.com
 License: Apache License, Version 2.0

{sunholo-0.66.18 → sunholo-0.66.21}/setup.py

@@ -1,7 +1,7 @@
 from setuptools import setup, find_packages
 
 # Define your base version
-version = '0.66.18'
+version = '0.66.21'
 
 setup(
     name='sunholo',

{sunholo-0.66.18 → sunholo-0.66.21}/sunholo/agents/flask/qna_routes.py

@@ -46,6 +46,7 @@ cache_duration = timedelta(minutes=5)  # Cache duration
 
 def make_openai_response(user_message, vector_name, answer):
     response_id = str(uuid.uuid4())
+    log.info("openai response: Q: {user_message} to VECTOR_NAME: {vector_name} - A: {answer}")
     openai_response = {
         "id": response_id,
         "object": "chat.completion",
@@ -418,9 +419,10 @@ def register_qna_routes(app, stream_interpreter, vac_interpreter):
             bot_output = parse_output(bot_output)
 
             log.info(f"Bot output: {bot_output}")
-
-            return make_openai_response(user_message, vector_name, bot_output.get('answer', ''))
-
+            if bot_output:
+                return make_openai_response(user_message, vector_name, bot_output.get('answer', ''))
+            else:
+                return make_openai_response(user_message, vector_name, 'ERROR: could not find an answer')
 
         except Exception as err:
             log.error(f"OpenAI response error: {err}")

{sunholo-0.66.18 → sunholo-0.66.21}/sunholo/agents/swagger.py

@@ -208,8 +208,8 @@ See more at https://dev.sunholo.com/
         try:
             stem = route_vac(vector_name).strip()
         except ValueError:
-            stem = f"${{{agent_type.upper()}_BACKEND_URL}}"
-            log.warning(f"Failed to find URL stem for {vector_name}/{agent_type} - using {stem} instead")
+            log.warning(f"Failed to find URL stem for {vector_name}/{agent_type} - skipping")
+            continue
         
         for method, endpoints in agent_config_paths.items():
             # controls if get requests are protected or not
@@ -273,9 +273,8 @@ See more at https://dev.sunholo.com/
         try:
             stem = route_vac(vector_name).strip()
         except ValueError:
-            stem = f"${{{agent_type.upper()}_BACKEND_URL}}"
-            log.warning(f"Failed to find URL stem for {vector_name}/{agent_type} - using {stem} instead")
-        
+            log.warning(f"Failed to find URL stem for {vector_name}/{agent_type} - skipping")
+            continue
 
         for method, endpoints in default_agent_config.items():
             # controls if get requests are protected or not

{sunholo-0.66.18 → sunholo-0.66.21}/sunholo/cli/chat_vac.py

@@ -11,6 +11,7 @@ import sys
 import subprocess
 import json
 import requests
+import os
 from pathlib import Path
 
 from rich import print
@@ -226,15 +227,30 @@ def resolve_service_url(args, no_config=False):
     if args.url_override:
 
         return args.url_override
+    
+    agent_name = load_config_key("agent", args.vac_name, kind="vacConfig")
+    agent_url = load_config_key("agent_url", args.vac_name, "vacConfig")
+    if agent_url:
+        console.print("Found agent_url within vacConfig: {agent_url}")
+    
+    # via public cloud endpoints - assumes no gcloud auth
+    if os.getenv("MULTIVAC_API_KEY"):
+        console.rule("Found MULTIVAC_API_KEY")
+        gcp_config = load_config_key("gcp_config", "global", "vacConfig")
+        endpoints_base_url = gcp_config.get("endpoints_base_url")
+        if not endpoints_base_url:
+            console.print("[bold red]MULTIVAC_API_KEY env var is set but no config.gcp_config.endpoints_base_url can be found[/bold red]")
+            sys.exit(1)
+
+        service_url = f"{endpoints_base_url}/v1/{agent_name}"
 
-    if args.no_proxy:
-        agent_url = load_config_key("agent_url", args.vac_name, "vacConfig")
-        if agent_url:
-            console.print("Found agent_url within vacConfig: {agent_url}")
+    # via direct access to agent url - requires gcloud auth access
+    elif args.no_proxy:
         
-        service_url = agent_url or get_cloud_run_service_url(args.project, args.region, args.vac_name)
+        service_url = agent_url or get_cloud_run_service_url(args.project, args.region, agent_name)
         console.print(f"No proxy, connecting directly to {service_url}")
     else:
+        # via gcloud proxy - requires gcloud auth access
         try:
             service_url = get_service_url(args.vac_name, args.project, args.region, no_config=no_config)
         except ValueError as e:

{sunholo-0.66.18 → sunholo-0.66.21}/sunholo.egg-info/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: sunholo
-Version: 0.66.18
+Version: 0.66.21
 Summary: Large Language Model DevOps - a package to help deploy LLMs to the Cloud.
 Home-page: https://github.com/sunholo-data/sunholo-py
-Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.66.18.tar.gz
+Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.66.21.tar.gz
 Author: Holosun ApS
 Author-email: multivac@sunholo.com
 License: Apache License, Version 2.0