Added voting logic

README.md

@@ -25,15 +25,16 @@ MQ:
   port: <MQ Port>
   server: <MQ Hostname or IP>
   users:
-    mq-chatgpt-api:
+    neon_llm_chat_gpt:
       password: <neon_chatgpt user's password>
       user: neon_chatgpt
-LLM_CHATGPT:
+LLM_CHAT_GPT:
   key: ""
   model: "gpt-3.5-turbo"
   role: "You are trying to give a short answer in less than 40 words."
   context_depth: 3
   max_tokens: 100
+  num_parallel_processes: 2
 ```
 
 For example, if your configuration resides in `~/.config`:

docker_overlay/etc/neon/diana.yaml

@@ -14,8 +14,9 @@ MQ:
     mq_handler:
       user: neon_api_utils
       password: Klatchat2021
-LLM_CHATGPT:
+LLM_CHAT_GPT:
   model: "gpt-3.5-turbo"
   role: "You are trying to give a short answer in less than 40 words."
   context_depth: 3
-  max_tokens: 100
+  max_tokens: 100
+  num_parallel_processes: 2

neon_llm_chatgpt/chatgpt.py

@@ -40,21 +40,33 @@ class ChatGPT(NeonLLM):
 
     def __init__(self, config):
         super().__init__(config)
-        self.model = config["model"]
+        self.model_name = config["model"]
         self.role = config["role"]
         self.context_depth = config["context_depth"]
         self.max_tokens = config["max_tokens"]
         self.api_key = config["key"]
         self.num_parallel_processes = config["num_parallel_processes"]
         self.warmup()
 
+    @property
+    def tokenizer(self) -> None:
+        return self._tokenizer
+
+    @property
+    def tokenizer_model_name(self) -> str:
+        return ""
+
     @property
     def model(self) -> openai:
         if self._model is None:
             openai.api_key = self.api_key
             self._model = openai
         return self._model
 
+    @property
+    def llm_model_name(self) -> str:
+        return self.model_name
+
     @property
     def _system_prompt(self) -> str:
         return self.role
@@ -72,20 +84,20 @@ def get_sorted_answer_indexes(self, question: str, answers: List[str]) -> List[i
         """
         if not answers:
             return []
-        scores = self._score(question=question, answers=answers)
+        scores = self._score(prompt=question, targets=answers)
         sorted_items = sorted(zip(range(len(answers)), scores), key=lambda x: x[1])
         sorted_items_indexes = [x[0] for x in sorted_items]
         return sorted_items_indexes
 
-    def _call_model(self, prompt: List[Dict[str]]) -> str:
+    def _call_model(self, prompt: List[Dict[str, str]]) -> str:
         """
             Wrapper for ChatGPT Model generation logic
             :param prompt: Input messages sequence
             :returns: Output text sequence generated by model
         """
 
         response = openai.ChatCompletion.create(
-            model=self.model,
+            model=self.llm_model_name,
             messages=prompt,
             temperature=0,
             max_tokens=self.max_tokens,
@@ -94,7 +106,7 @@ def _call_model(self, prompt: List[Dict[str]]) -> str:
 
         return text
 
-    def _assemble_prompt(self, message: str, chat_history: List[List[str]]) -> List[Dict[str]]:
+    def _assemble_prompt(self, message: str, chat_history: List[List[str]]) -> List[Dict[str, str]]:
         """
             Assembles prompt engineering logic
             Setup Guidance:
@@ -126,7 +138,10 @@ def _score(self, prompt: str, targets: List[str]) -> List[float]:
         scores_list = distances_from_embeddings(question_embeddings, answers_embeddings)
         return scores_list
 
-    def _embeddings(self, question: str, answers: List[str]) -> (List[str], List[List[float]]):
+    def _tokenize(self, prompt: str) -> None:
+        return None
+
+    def _embeddings(self, question: str, answers: List[str]) -> (List[float], List[List[float]]):
         """
             Computes embeddings for the list of provided answers
             :param question: Question for LLM to response to

neon_llm_chatgpt/rmq.py

@@ -39,7 +39,7 @@ def __init__(self):
 
     @property
     def name(self):
-        return "chatgpt"
+        return "chat_gpt"
 
     @property
     def model(self):