remove lite_llm; distinguish if first message or whole conversation in chat querifier; Add info to README

.env

@@ -42,6 +42,5 @@ LLM_API_KEY=placeholder
 # valid Language options - "en" "de" "multi"
 LLM_LANGUAGE=de
 
-# LLM Model to be used - "mistral", "phi3", "llama3", etc.
+# LLM Model to be used - "mistral", "phi3", "llama3", etc. (Check available with <ip_of_ollama_instance>:11434/api/tags)
 LLM_MODEL_NAME=mistral
-

rag-chat-backend/src/endpoints/llm.py

@@ -38,7 +38,6 @@ SEARCH_COMPONENT = IndexSearchComponent(os_client=OS_INTERFACE)
 PROMPTS = LLM.prompt_obj
 llm_component_name = "ollama-llm"
 selected_language = "german"
-lite_llm = os.getenv("LITE_LLM")
 stream_llms = ["fu"]
 
 
@@ -97,7 +96,15 @@ def chat(stream_response, messages, tags=None, languages: str = None, start_date
             raise HTTPException(status_code=400, detail="Invalid format for 'messages'")
 
     # provides you a query based on the whole chat
-    query = LLM.llm_chat_querifier(chat=messages)
+    if len(messages) <= 2:
+        if messages[0]["role"] == "user":
+            query = messages[0]["content"]
+        elif messages[1]["role"] == "user":
+            query = messages[1]["content"]
+        else:
+            query = LLM.llm_chat_querifier(chat=messages)
+    else:
+        query = LLM.llm_chat_querifier(chat=messages)
     logger.info("Generated query: '%s'", query)
 
     # creating a new dashboard logger at the start of the project
@@ -137,7 +144,7 @@ def chat(stream_response, messages, tags=None, languages: str = None, start_date
     # check if chat is allowed for selected llm
     logger.info("Stream response: %s", stream_response)
     if stream_response == "true":
-        if lite_llm == "true" or llm_component_name in stream_llms:
+        if llm_component_name in stream_llms:
             answer = LLM.llm_chat_request(
                 chat=messages,
                 context=formatted_context,