Update management of env vars

.env.example

@@ -3,9 +3,13 @@
 OPENAI_BASE_URL="https://generativelanguage.googleapis.com/v1beta/openai/"
 OPENAI_API_KEY="..."
 
+DEFAULT_PLANNER_MODEL="..."
+DEFAULT_WORKER_MODEL="..."
+
 # Embedding model
 EMBEDDING_BASE_URL="https://..."
 EMBEDDING_API_KEY="..."
+EMBEDDING_MODEL_NAME="..."
 
 # LangFuse
 LANGFUSE_SECRET_KEY="sk-lf-..."
@@ -20,9 +24,11 @@ WEAVIATE_HTTP_PORT="443" # or 8080 for localhost
 WEAVIATE_GRPC_PORT="443" # or 50051 for localhost
 WEAVIATE_HTTP_SECURE="true" # set to false for localhost
 WEAVIATE_GRPC_SECURE="true" # set to false for localhost
+WEAVIATE_COLLECTION_NAME="..."
 
 # Optionally, specify E2B.dev API key for Python Code Interpreter
 E2B_API_KEY="e2b_..."
+DEFAULT_CODE_INTERPRETER_TEMPLATE="..."
 
 WEB_SEARCH_BASE_URL="..."
 WEB_SEARCH_API_KEY="..."

src/1_basics/0_search_demo/app.py

@@ -32,7 +32,7 @@
 
 load_dotenv(verbose=True)
 
-configs = Configs.from_env_var()
+configs = Configs()
 async_weaviate_client = get_weaviate_async_client(
     http_host=configs.weaviate_http_host,
     http_port=configs.weaviate_http_port,
@@ -45,7 +45,7 @@
 async_openai_client = AsyncOpenAI()
 async_knowledgebase = AsyncWeaviateKnowledgeBase(
     async_weaviate_client,
-    collection_name="enwiki_20250520",
+    collection_name=configs.weaviate_collection_name,
 )
 
 

src/1_basics/1_react_rag/app.py

@@ -26,8 +26,6 @@
 load_dotenv(verbose=True)
 
 MAX_TURNS = 5
-AGENT_LLM_NAME = "gemini-2.5-flash"
-
 
 tools: list["ChatCompletionToolParam"] = [
     {
@@ -76,7 +74,7 @@ async def react_rag(query: str, history: list[ChatMessage]):
 
     for _ in range(MAX_TURNS):
         completion = await async_openai_client.chat.completions.create(
-            model=AGENT_LLM_NAME,
+            model=configs.default_planner_model,
             messages=oai_messages,
             tools=tools,
             reasoning_effort=None,
@@ -140,7 +138,7 @@ async def react_rag(query: str, history: list[ChatMessage]):
 )
 
 if __name__ == "__main__":
-    configs = Configs.from_env_var()
+    configs = Configs()
     async_weaviate_client = get_weaviate_async_client(
         http_host=configs.weaviate_http_host,
         http_port=configs.weaviate_http_port,
@@ -153,7 +151,7 @@ async def react_rag(query: str, history: list[ChatMessage]):
     async_openai_client = AsyncOpenAI()
     async_knowledgebase = AsyncWeaviateKnowledgeBase(
         async_weaviate_client,
-        collection_name="enwiki_20250520",
+        collection_name=configs.weaviate_collection_name,
     )
 
     signal.signal(signal.SIGINT, _handle_sigint)

src/1_basics/1_react_rag/cli.py

@@ -23,7 +23,6 @@
 load_dotenv(verbose=True)
 
 MAX_TURNS = 5
-AGENT_LLM_NAME = "gemini-2.5-flash"
 
 tools: list["ChatCompletionToolParam"] = [
     {
@@ -49,7 +48,7 @@
 
 
 async def _main():
-    configs = Configs.from_env_var()
+    configs = Configs()
     async_weaviate_client = get_weaviate_async_client(
         http_host=configs.weaviate_http_host,
         http_port=configs.weaviate_http_port,
@@ -62,7 +61,7 @@ async def _main():
     async_openai_client = AsyncOpenAI()
     async_knowledgebase = AsyncWeaviateKnowledgeBase(
         async_weaviate_client,
-        collection_name="enwiki_20250520",
+        collection_name=configs.weaviate_collection_name,
     )
 
     messages: list = [
@@ -81,7 +80,7 @@ async def _main():
         while True:
             for _ in range(MAX_TURNS):
                 completion = await async_openai_client.chat.completions.create(
-                    model=AGENT_LLM_NAME,
+                    model=configs.default_planner_model,
                     messages=messages,
                     tools=tools,
                 )

src/2_frameworks/1_react_rag/app.py

@@ -27,25 +27,6 @@
 logging.basicConfig(level=logging.INFO)
 
 
-AGENT_LLM_NAME = "gemini-2.5-flash"
-
-configs = Configs.from_env_var()
-async_weaviate_client = get_weaviate_async_client(
-    http_host=configs.weaviate_http_host,
-    http_port=configs.weaviate_http_port,
-    http_secure=configs.weaviate_http_secure,
-    grpc_host=configs.weaviate_grpc_host,
-    grpc_port=configs.weaviate_grpc_port,
-    grpc_secure=configs.weaviate_grpc_secure,
-    api_key=configs.weaviate_api_key,
-)
-async_openai_client = AsyncOpenAI()
-async_knowledgebase = AsyncWeaviateKnowledgeBase(
-    async_weaviate_client,
-    collection_name="enwiki_20250520",
-)
-
-
 async def _cleanup_clients() -> None:
     """Close async clients."""
     await async_weaviate_client.close()
@@ -65,8 +46,9 @@ async def _main(question: str, gr_messages: list[ChatMessage]):
         instructions=REACT_INSTRUCTIONS,
         tools=[agents.function_tool(async_knowledgebase.search_knowledgebase)],
         model=agents.OpenAIChatCompletionsModel(
-            model=AGENT_LLM_NAME, openai_client=async_openai_client
+            model=configs.default_planner_model, openai_client=async_openai_client
         ),
+        model_settings=agents.ModelSettings(parallel_tool_calls=True),
     )
 
     result_stream = agents.Runner.run_streamed(main_agent, input=question)
@@ -88,7 +70,7 @@ async def _main(question: str, gr_messages: list[ChatMessage]):
 
 
 if __name__ == "__main__":
-    configs = Configs.from_env_var()
+    configs = Configs()
     async_weaviate_client = get_weaviate_async_client(
         http_host=configs.weaviate_http_host,
         http_port=configs.weaviate_http_port,
@@ -100,7 +82,7 @@ async def _main(question: str, gr_messages: list[ChatMessage]):
     )
     async_knowledgebase = AsyncWeaviateKnowledgeBase(
         async_weaviate_client,
-        collection_name="enwiki_20250520",
+        collection_name=configs.weaviate_collection_name,
     )
 
     async_openai_client = AsyncOpenAI()

src/2_frameworks/1_react_rag/cli.py

@@ -24,12 +24,11 @@
 
 load_dotenv(verbose=True)
 
-AGENT_LLM_NAME = "gemini-2.5-flash"
 no_tracing_config = RunConfig(tracing_disabled=True)
 
 
 async def _main(query: str):
-    configs = Configs.from_env_var()
+    configs = Configs()
     async_weaviate_client = get_weaviate_async_client(
         http_host=configs.weaviate_http_host,
         http_port=configs.weaviate_http_port,
@@ -41,7 +40,7 @@ async def _main(query: str):
     )
     async_knowledgebase = AsyncWeaviateKnowledgeBase(
         async_weaviate_client,
-        collection_name="enwiki_20250520",
+        collection_name=configs.weaviate_collection_name,
     )
 
     async_openai_client = AsyncOpenAI()
@@ -51,7 +50,7 @@ async def _main(query: str):
         instructions=REACT_INSTRUCTIONS,
         tools=[function_tool(async_knowledgebase.search_knowledgebase)],
         model=OpenAIChatCompletionsModel(
-            model=AGENT_LLM_NAME, openai_client=async_openai_client
+            model=configs.default_planner_model, openai_client=async_openai_client
         ),
     )
 

src/2_frameworks/1_react_rag/langfuse_gradio.py

@@ -31,9 +31,8 @@
 
 set_up_logging()
 
-AGENT_LLM_NAME = "gemini-2.5-flash"
 
-configs = Configs.from_env_var()
+configs = Configs()
 async_weaviate_client = get_weaviate_async_client(
     http_host=configs.weaviate_http_host,
     http_port=configs.weaviate_http_port,
@@ -46,7 +45,7 @@
 async_openai_client = AsyncOpenAI()
 async_knowledgebase = AsyncWeaviateKnowledgeBase(
     async_weaviate_client,
-    collection_name="enwiki_20250520",
+    collection_name=configs.weaviate_collection_name,
 )
 
 
@@ -71,7 +70,7 @@ async def _main(question: str, gr_messages: list[ChatMessage]):
         instructions=REACT_INSTRUCTIONS,
         tools=[agents.function_tool(async_knowledgebase.search_knowledgebase)],
         model=agents.OpenAIChatCompletionsModel(
-            model=AGENT_LLM_NAME, openai_client=async_openai_client
+            model=configs.default_planner_model, openai_client=async_openai_client
         ),
     )
 
@@ -102,8 +101,6 @@ async def _main(question: str, gr_messages: list[ChatMessage]):
 
 
 if __name__ == "__main__":
-    configs = Configs.from_env_var()
-
     signal.signal(signal.SIGINT, _handle_sigint)
 
     try:

src/2_frameworks/2_multi_agent/efficient.py

@@ -33,12 +33,8 @@
 
 set_up_logging()
 
-AGENT_LLM_NAMES = {
-    "worker": "gemini-2.5-flash",  # less expensive,
-    "planner": "gemini-2.5-pro",  # more expensive, better at reasoning and planning
-}
 
-configs = Configs.from_env_var()
+configs = Configs()
 async_weaviate_client = get_weaviate_async_client(
     http_host=configs.weaviate_http_host,
     http_port=configs.weaviate_http_port,
@@ -51,9 +47,14 @@
 async_openai_client = AsyncOpenAI()
 async_knowledgebase = AsyncWeaviateKnowledgeBase(
     async_weaviate_client,
-    collection_name="enwiki_20250520",
+    collection_name=configs.weaviate_collection_name,
 )
 
+AGENT_LLM_NAMES = {
+    "worker": configs.default_worker_model,  # less expensive,
+    "planner": configs.default_planner_model,  # more expensive, better at reasoning and planning
+}
+
 
 async def _cleanup_clients() -> None:
     """Close async clients."""

src/2_frameworks/2_multi_agent/efficient_multiple_kbs.py

@@ -30,12 +30,8 @@
 
 set_up_logging()
 
-AGENT_LLM_NAMES = {
-    "worker": "gemini-2.5-flash",  # less expensive,
-    "planner": "gemini-2.5-pro",  # more expensive, better at reasoning and planning
-}
 
-configs = Configs.from_env_var()
+configs = Configs()
 async_weaviate_client = get_weaviate_async_client(
     http_host=configs.weaviate_http_host,
     http_port=configs.weaviate_http_port,
@@ -48,9 +44,14 @@
 async_openai_client = AsyncOpenAI()
 async_knowledgebase = AsyncWeaviateKnowledgeBase(
     async_weaviate_client,
-    collection_name="enwiki_20250520",
+    collection_name=configs.weaviate_collection_name,
 )
 
+AGENT_LLM_NAMES = {
+    "worker": configs.default_worker_model,  # less expensive,
+    "planner": configs.default_planner_model,  # more expensive, better at reasoning and planning
+}
+
 gemini_grounding_tool = GeminiGroundingWithGoogleSearch(
     model_settings=ModelSettings(model=AGENT_LLM_NAMES["worker"])
 )

src/2_frameworks/2_multi_agent/verbose.py

@@ -142,7 +142,7 @@ async def _main(question: str, gr_messages: list[ChatMessage]):
         name="Planner Agent",
         instructions=PLANNER_INSTRUCTIONS,
         model=agents.OpenAIChatCompletionsModel(
-            model="gemini-2.5-flash", openai_client=async_openai_client
+            model=configs.default_planner_model, openai_client=async_openai_client
         ),
         output_type=SearchPlan,
     )
@@ -151,7 +151,7 @@ async def _main(question: str, gr_messages: list[ChatMessage]):
         instructions=RESEARCHER_INSTRUCTIONS,
         tools=[agents.function_tool(async_knowledgebase.search_knowledgebase)],
         model=agents.OpenAIChatCompletionsModel(
-            model="gemini-2.5-flash-lite",
+            model=configs.default_worker_model,
             openai_client=async_openai_client,
         ),
         model_settings=agents.ModelSettings(tool_choice="required"),
@@ -160,7 +160,7 @@ async def _main(question: str, gr_messages: list[ChatMessage]):
         name="Writer Agent",
         instructions=WRITER_INSTRUCTIONS,
         model=agents.OpenAIChatCompletionsModel(
-            model="gemini-2.5-flash", openai_client=async_openai_client
+            model=configs.default_planner_model, openai_client=async_openai_client
         ),
         output_type=ResearchReport,
     )
@@ -213,7 +213,7 @@ async def _main(question: str, gr_messages: list[ChatMessage]):
 
 
 if __name__ == "__main__":
-    configs = Configs.from_env_var()
+    configs = Configs()
     async_weaviate_client = get_weaviate_async_client(
         http_host=configs.weaviate_http_host,
         http_port=configs.weaviate_http_port,
@@ -225,7 +225,7 @@ async def _main(question: str, gr_messages: list[ChatMessage]):
     )
     async_knowledgebase = AsyncWeaviateKnowledgeBase(
         async_weaviate_client,
-        collection_name="enwiki_20250520",
+        collection_name=configs.weaviate_collection_name,
     )
 
     async_openai_client = AsyncOpenAI()

src/2_frameworks/3_code_interpreter/app.py

@@ -5,6 +5,7 @@
 You will need your E2B API Key.
 """
 
+import os
 from pathlib import Path
 
 import agents
@@ -42,7 +43,6 @@
 but you won't be able to install packages.
 """
 
-AGENT_LLM_NAME = "gemini-2.5-flash"
 async_openai_client = AsyncOpenAI()
 code_interpreter = CodeInterpreter(
     local_files=[
@@ -65,7 +65,8 @@ async def _main(question: str, gr_messages: list[ChatMessage]):
             )
         ],
         model=agents.OpenAIChatCompletionsModel(
-            model=AGENT_LLM_NAME, openai_client=async_openai_client
+            model=os.getenv("DEFAULT_PLANNER_MODEL", "gemini-2.5-flash"),
+            openai_client=async_openai_client,
         ),
     )