Merge pull request #59 from DL4DS/minor_fix

.gitignore

@@ -169,4 +169,5 @@ code/.chainlit/translations/
 storage/logs/*
 vectorstores/*
 
-*/.files/*
+*/.files/*
+code/storage/models/

code/.chainlit/config.toml

@@ -61,11 +61,11 @@ name = "AI Tutor"
 # Large size content are by default collapsed for a cleaner ui
 default_collapse_content = true
 
-# Hide the chain of thought details from the user in the UI.
-hide_cot = true
+# Chain of Thought (CoT) display mode. Can be "hidden", "tool_call" or "full".
+cot = "hidden"
 
 # Link to your github repo. This will add a github button in the UI's header.
-# github = "https://github.com/DL4DS/dl4ds_tutor"
+github = "https://github.com/DL4DS/dl4ds_tutor"
 
 # Specify a CSS file that can be used to customize the user interface.
 # The CSS file can be served from the public directory or via an external link.
@@ -87,7 +87,7 @@ custom_meta_image_url = "https://upload.wikimedia.org/wikipedia/commons/thumb/f/
 # custom_build = "./public/build"
 
 [UI.theme]
-    default = "dark"
+    default = "light"
     #layout = "wide"
     #font_family = "Inter, sans-serif"
 # Override default MUI light theme. (Check theme.ts)
@@ -117,4 +117,4 @@ custom_meta_image_url = "https://upload.wikimedia.org/wikipedia/commons/thumb/f/
         #secondary = "#BDBDBD"
 
 [meta]
-generated_by = "1.1.306"
+generated_by = "1.1.402"

code/main.py

@@ -23,11 +23,11 @@ from chainlit.types import ThreadDict
 import time
 
 USER_TIMEOUT = 60_000
-SYSTEM = "System 🖥️"
-LLM = "LLM 🧠"
-AGENT = "Agent <>"
-YOU = "You 😃"
-ERROR = "Error 🚫"
+SYSTEM = "System"
+LLM = "AI Tutor"
+AGENT = "Agent"
+YOU = "User"
+ERROR = "Error"
 
 with open("modules/config/config.yml", "r") as f:
     config = yaml.safe_load(f)
@@ -111,11 +111,6 @@ class Chatbot:
         )  # update only llm attributes that are changed
         self.chain = self.llm_tutor.qa_bot(
             memory=conversation_list,
-            callbacks=(
-                [cl.LangchainCallbackHandler()]
-                if cl_data._data_layer and self.config["chat_logging"]["callbacks"]
-                else None
-            ),
         )
 
         cl.user_session.set("chain", self.chain)
@@ -279,7 +274,7 @@ class Chatbot:
         Returns:
             str: The renamed author.
         """
-        rename_dict = {"Chatbot": "AI Tutor"}
+        rename_dict = {"Chatbot": LLM}
         return rename_dict.get(orig_author, orig_author)
 
     async def start(self, config=None):
@@ -318,11 +313,6 @@ class Chatbot:
 
         self.chain = self.llm_tutor.qa_bot(
             memory=memory,
-            callbacks=(
-                [cl.LangchainCallbackHandler()]
-                if cl_data._data_layer and self.config["chat_logging"]["callbacks"]
-                else None
-            ),
         )
         self.question_generator = self.llm_tutor.question_generator
         cl.user_session.set("llm_tutor", self.llm_tutor)
@@ -375,7 +365,12 @@ class Chatbot:
                 "user_id": self.user["user_id"],
                 "conversation_id": self.user["session_id"],
                 "memory_window": self.config["llm_params"]["memory_window"],
-            }
+            },
+            "callbacks": (
+                [cl.LangchainCallbackHandler()]
+                if cl_data._data_layer and self.config["chat_logging"]["callbacks"]
+                else None
+            ),
         }
 
         if stream:
@@ -400,11 +395,20 @@ class Chatbot:
 
         if self.config["llm_params"]["generate_follow_up"]:
             start_time = time.time()
-            list_of_questions = self.question_generator.generate_questions(
+            config = {
+                "callbacks": (
+                    [cl.LangchainCallbackHandler()]
+                    if cl_data._data_layer and self.config["chat_logging"]["callbacks"]
+                    else None
+                )
+            }
+
+            list_of_questions = await self.question_generator.generate_questions(
                 query=user_query_dict["input"],
                 response=answer,
                 chat_history=res.get("chat_history"),
                 context=res.get("context"),
+                config=config,
             )
 
             for question in list_of_questions:
@@ -456,7 +460,11 @@ class Chatbot:
             type="user_message",
             author=self.user["user_id"],
         ).send()
-        await self.main(message)
+        async with cl.Step(
+            name="on_follow_up", type="run", parent_id=message.id
+        ) as step:
+            await self.main(message)
+            step.output = message.content
 
 
 chatbot = Chatbot(config=config)

code/modules/chat/chat_model_loader.py

@@ -28,7 +28,7 @@ class ChatModelLoader:
         elif self.config["llm_params"]["llm_loader"] == "local_llm":
             n_batch = 512  # Should be between 1 and n_ctx, consider the amount of VRAM in your GPU.
             model_path = self._verify_model_cache(
-                self.config["llm_params"]["local_llm_params"]["model"]
+                self.config["llm_params"]["local_llm_params"]["model_path"]
             )
             llm = LlamaCpp(
                 model_path=model_path,

code/modules/chat/helpers.py

@@ -110,6 +110,7 @@ def get_prompt(config, prompt_type):
         return prompts["openai"]["rephrase_prompt"]
 
 
+# TODO: Do this better
 def get_history_chat_resume(steps, k, SYSTEM, LLM):
     conversation_list = []
     count = 0
@@ -119,14 +120,17 @@ def get_history_chat_resume(steps, k, SYSTEM, LLM):
                 conversation_list.append(
                     {"type": "user_message", "content": step["output"]}
                 )
+                count += 1
             elif step["type"] == "assistant_message":
                 if step["name"] == LLM:
                     conversation_list.append(
                         {"type": "ai_message", "content": step["output"]}
                     )
+                    count += 1
             else:
-                raise ValueError("Invalid message type")
-        count += 1
+                pass
+                # raise ValueError("Invalid message type")
+        # count += 1
         if count >= 2 * k:  # 2 * k to account for both user and assistant messages
             break
     conversation_list = conversation_list[::-1]

code/modules/chat/langchain/langchain_rag.py

@@ -100,8 +100,8 @@ class QuestionGenerator:
     def __init__(self):
         pass
 
-    def generate_questions(self, query, response, chat_history, context):
-        questions = return_questions(query, response, chat_history, context)
+    def generate_questions(self, query, response, chat_history, context, config):
+        questions = return_questions(query, response, chat_history, context, config)
         return questions
 
 
@@ -204,7 +204,7 @@ class Langchain_RAG_V2(BaseRAG):
                     is_shared=True,
                 ),
             ],
-        )
+        ).with_config(run_name="Langchain_RAG_V2")
 
         if callbacks is not None:
             self.rag_chain = self.rag_chain.with_config(callbacks=callbacks)

code/modules/chat/langchain/utils.py

@@ -280,7 +280,8 @@ def create_retrieval_chain(
     return retrieval_chain
 
 
-def return_questions(query, response, chat_history_str, context):
+# TODO: Remove Hard-coded values
+async def return_questions(query, response, chat_history_str, context, config):
 
     system = (
         "You are someone that suggests a question based on the student's input and chat history. "
@@ -303,13 +304,17 @@ def return_questions(query, response, chat_history_str, context):
     )
     llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)
     question_generator = prompt | llm | StrOutputParser()
-    new_questions = question_generator.invoke(
+    question_generator = question_generator.with_config(
+        run_name="follow_up_question_generator"
+    )
+    new_questions = await question_generator.ainvoke(
         {
             "chat_history_str": chat_history_str,
             "context": context,
             "query": query,
             "response": response,
-        }
+        },
+        config=config,
     )
 
     list_of_questions = new_questions.split("...")

code/modules/config/config.yml

@@ -37,13 +37,14 @@ llm_params:
     temperature: 0.7 # float
     repo_id: 'TheBloke/TinyLlama-1.1B-Chat-v1.0-GGUF' # HuggingFace repo id
     filename: 'tinyllama-1.1b-chat-v1.0.Q5_0.gguf' # Specific name of gguf file in the repo
+    model_path: 'storage/models/tinyllama-1.1b-chat-v1.0.Q5_0.gguf' # Path to the model file
   stream: False # bool
   pdf_reader: 'gpt' # str [llama, pymupdf, gpt]
 
 chat_logging:
   log_chat: True # bool
   platform: 'literalai'
-  callbacks: False # bool
+  callbacks: True # bool
 
 splitter_options:
   use_splitter: True # bool

code/public/test.css

@@ -13,10 +13,6 @@ a[href*='https://github.com/Chainlit/chainlit'] {
     border-radius: 50%; /* Maintain circular shape */
 }
 
-/* Hide the default image */
-.MuiAvatar-root.MuiAvatar-circular.css-m2icte .MuiAvatar-img.css-1hy9t21 {
-    display: none;
-}
 
 .MuiAvatar-root.MuiAvatar-circular.css-v72an7 {
     background-image: url('/public/avatars/ai-tutor.png'); /* Replace with your custom image URL */
@@ -26,18 +22,3 @@ a[href*='https://github.com/Chainlit/chainlit'] {
     height: 40px; /* Ensure the dimensions match the original */
     border-radius: 50%; /* Maintain circular shape */
 }
-
-/* Hide the default image */
-.MuiAvatar-root.MuiAvatar-circular.css-v72an7 .MuiAvatar-img.css-1hy9t21 {
-    display: none;
-}
-
-/* Hide the new chat button
-#new-chat-button {
-    display: none;
-} */
-
-/* Hide the open sidebar button
-#open-sidebar-button {
-    display: none;
-} */

requirements.txt

@@ -23,7 +23,7 @@ llama-cpp-python
 pymupdf
 websockets
 langchain-openai
-langchain_experimental
+langchain-experimental
 html2text
 PyPDF2
 pdf2image