Merge pull request #13 from filip-michalsky/model_name_config

Add the ability to configure model name when streaming. Right now supports openAI but need to add abstraction to swap in any LLM.

examples/streaming_generator_example.py

@@ -1,5 +1,6 @@
 import os
-from sales_gpt import SalesGPT
+from salesgpt.agents import SalesGPT
+
 from langchain.chat_models import ChatOpenAI
 
 with open('.env','r') as f:
@@ -25,7 +26,7 @@
 sales_agent.seed_agent()
 
 # get generator of the LLM output
-generator = sales_agent.step(return_streaming_generator=True)
+generator = sales_agent.step(return_streaming_generator=True, model_name="gpt-3.5-turbo-0613")
 
 # operate on streaming LLM output in near-real time
 # for instance, do something after each full sentence is generated

salesgpt/agents.py

@@ -71,7 +71,7 @@ def human_step(self, human_input):
         self.conversation_history.append(human_input)
 
     @time_logger
-    def step(self, return_streaming_generator: bool = False):
+    def step(self, return_streaming_generator: bool = False, model_name="gpt-3.5-turbo-0613"):
         """
         Args:
             return_streaming_generator (bool): whether or not return
@@ -80,11 +80,11 @@ def step(self, return_streaming_generator: bool = False):
         if not return_streaming_generator:
             self._call(inputs={})
         else:
-            return self._streaming_generator()
+            return self._streaming_generator(model_name=model_name)
 
     # TO-DO change this override "run" override the "run method" in the SalesConversation chain!
     @time_logger
-    def _streaming_generator(self):
+    def _streaming_generator(self, model_name="gpt-3.5-turbo-0613"):
         """
         Sometimes, the sales agent wants to take an action before the full LLM output is available.
         For instance, if we want to do text to speech on the partial LLM output.
@@ -128,7 +128,7 @@ def _streaming_generator(self):
             messages=messages,
             stop="<END_OF_TURN>",
             stream=True,
-            model="gpt-3.5-turbo-0613",
+            model=model_name,
         )
 
     def _call(self, inputs: Dict[str, Any]) -> None:

salesgpt/version.py

@@ -1,3 +1,3 @@
 """Version information."""
 
-__version__ = "0.0.2"
+__version__ = "0.0.3"

tests/test_salesgpt.py

@@ -36,3 +36,40 @@ def test_valid_inference(self, load_env):
         assert agent_output is not None, "Agent output cannot be None."
         assert isinstance(agent_output, str), "Agent output needs to be of type str"
         assert len(agent_output) > 0, "Length of output needs to be greater than 0."
+
+
+    def test_valid_inference_stream(self, load_env):
+        """Test that the agent will start and generate the first utterance when streaming."""
+
+        llm = ChatOpenAI(temperature=0.9)
+        model_name = 'gpt-3.5-turbo'
+
+        sales_agent = SalesGPT.from_llm(
+            llm,
+            verbose=False,
+            salesperson_name="Ted Lasso",
+            salesperson_role="Sales Representative",
+            company_name="Sleep Haven",
+            company_business="""Sleep Haven 
+                                    is a premium mattress company that provides
+                                    customers with the most comfortable and
+                                    supportive sleeping experience possible. 
+                                    We offer a range of high-quality mattresses,
+                                    pillows, and bedding accessories 
+                                    that are designed to meet the unique 
+                                    needs of our customers.""",
+        )
+
+        sales_agent.seed_agent()
+        sales_agent.determine_conversation_stage()  # optional for demonstration, built into the prompt
+
+        # agent output sample
+        stream_generator = sales_agent.step(return_streaming_generator=True, model_name=model_name)
+        agent_output=''
+        for chunk in stream_generator:
+            token = chunk["choices"][0]["delta"].get("content", "")
+            agent_output += token
+
+        assert agent_output is not None, "Agent output cannot be None."
+        assert isinstance(agent_output, str), "Agent output needs to be of type str"
+        assert len(agent_output) > 0, "Length of output needs to be greater than 0."