feat!: Remove WebsiteKnowledgeBase

Using a phidata knowledge base without pgvector isn't too useful. Once I learn more about databases I may re-add support with pgvector.

.env.example

@@ -18,10 +18,6 @@ REDACT_EMAIL_ADDRESSES=true
 NO_TOOLS=true
 # API key for Exa (https://exa.ai), a search engine for LLMs
 EXA_API_KEY=""
-# URL(s) that should be scraped for information
-# Multiple URLs can be passed by separating them with commas ("https://example.com,https://example.com")
-SCRAPABLE_URL=""
-
 # Show what tools are being called
 SHOW_TOOL_CALLS = false
 # Use phidata's debug mode

llmail/utils/cli_args.py

@@ -77,12 +77,12 @@ def set_argparse():
         help="Exa API key for searching with Exa (disables DuckDuckGo)",
         default=os.getenv("EXA_API_KEY") if os.getenv("EXA_API_KEY") else None,
     )
-    ai.add_argument(
-        "--scrapable-url",
-        help="URL(s) that can be scraped for information",
-        default=(os.getenv("SCRAPABLE_URL").split(",") if os.getenv("SCRAPABLE_URL") else None),
-        action="append",
-    )
+    # ai.add_argument(
+    #     "--scrapable-url",
+    #     help="URL(s) that can be scraped for information",
+    #     default=(os.getenv("SCRAPABLE_URL").split(",") if os.getenv("SCRAPABLE_URL") else None),
+    #     action="append",
+    # )
     ai.add_argument(
         "--no-tools",
         help="Do not use any tools via function calling",

llmail/utils/responding.py

@@ -12,8 +12,8 @@
 from phi.llm.ollama import Ollama
 from phi.tools.exa import ExaTools
 from phi.tools.website import WebsiteTools
-from phi.knowledge.website import WebsiteKnowledgeBase
-from phi.knowledge.combined import CombinedKnowledgeBase
+# from phi.knowledge.website import WebsiteKnowledgeBase
+# from phi.knowledge.combined import CombinedKnowledgeBase
 import ollama
 
 # Uses utils/__init__.py to import from utils/logging.py and utils/cli_args.py respectively
@@ -164,11 +164,13 @@ def fetch_and_process_emails(
             else:
                 ValueError("Invalid email thread")
             # Select tools
-            website_knowledge_base = WebsiteKnowledgeBase(
-                urls=args.scrapable_url if args.scrapable_url else []
-            )
+            # website_knowledge_base = WebsiteKnowledgeBase(
+            #     urls=args.scrapable_url if args.scrapable_url else []
+            # )
             tools = [
-                WebsiteTools(knowledge_base=website_knowledge_base),
+                # Giving WebsiteTools the knowledge base seems to allow it to add URLs to the knowledge base
+                # WebsiteTools(knowledge_base=website_knowledge_base),
+                WebsiteTools(),
                 DuckDuckGo(search=True, news=True),
             ]
             if args.exa_api_key is not None:
@@ -211,11 +213,11 @@ def fetch_and_process_emails(
                 tool_call_limit=10,
                 debug_mode=args.phidata_debug,
                 prevent_hallucinations=True,
-                knowledge_base=CombinedKnowledgeBase(
-                    sources=[
-                        website_knowledge_base,
-                    ]
-                ),
+                # knowledge_base=CombinedKnowledgeBase(
+                #     sources=[
+                #         website_knowledge_base,
+                #     ]
+                # ),
             )
             send_reply(
                 thread=tracking.get_thread_history(client, email_thread),