add links and openai model

app.py

@@ -37,18 +37,31 @@ else:
     os.system('pip install llama-cpp-python')
 
 css="""
-#col-container {max-width: 700px; margin-left: auto; margin-right: auto;}
+#col-container {max-width: 1500px; margin-left: auto; margin-right: auto;}
 """
 
 title = """
-<div style="text-align: center;max-width: 700px;">
-    <h1>Chat with Documents 📚 - Falcon, Llama-2</h1>
-    <p style="text-align: center;">Upload txt, pdf, doc, docx, enex, epub, html, md, odt, ptt, pttx; click the "Click to Upload Files" button, <br />
-    Wait for the Status to show Loaded documents, start typing your questions. <br />
-    The app is set to store chat-history</p>
+<div style="text-align: center;max-width: 1500px;">
+    <h3>Chat with Documents 📚 - Falcon, Llama-2</h3>
+    <p style="text-align: center;">Upload txt, pdf, doc, docx, enex, epub, html, md, odt, ptt and pttx. 
+    Wait for the Status to show Loaded documents, start typing your questions. This is a demo of <a href="https://github.com/R3gm/ConversaDocs">ConversaDocs</a>.<br /></p>
 </div>
 """
 
+description = """
+# Application Information
+
+- Notebook for run ConversaDocs in Colab [![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/R3gm/ConversaDocs/blob/main/ConversaDocs_Colab.ipynb)
+
+- Oficial Repository [![a](https://img.shields.io/badge/GitHub-Repository-black?style=flat-square&logo=github)](https://github.com/R3gm/ConversaDocs/)
+
+- This application works on both CPU and GPU. For fast inference with GGML models, use the GPU.
+
+- You can clone the 'space' but to make it work, you need to set My_hf_token in secrets with a valid huggingface [token](https://huggingface.co/settings/tokens) 
+
+- For more information about what GGML models are, you can visit this notebook [![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/R3gm/InsightSolver-Colab/blob/main/LLM_Inference_with_llama_cpp_python__Llama_2_13b_chat.ipynb)
+"""
+
 theme='aliabid94/new-theme'
 
 def flag():
@@ -58,9 +71,10 @@ def upload_file(files, max_docs):
     file_paths = [file.name for file in files]
     return dc.call_load_db(file_paths, max_docs)
 
-def predict(message, chat_history, max_k):
+def predict(message, chat_history, max_k, check_memory):
         print(message)
-        bot_message = dc.convchain(message, max_k)
+        print(check_memory)
+        bot_message = dc.convchain(message, max_k, check_memory)
         print(bot_message)
         return "", dc.get_chats()
 
@@ -77,6 +91,10 @@ def convert():
     data_docs += f"<hr><h3 style='color:red;'>{pg}</h2><p>{txt}</p><p>{sc}</p>"
   return data_docs
 
+def clear_api_key(api_key):
+  return 'api_key...', dc.openai_model(api_key)
+
+
 # Max values in generation
 DOC_DB_LIMIT = 10
 MAX_NEW_TOKENS = 2048
@@ -89,25 +107,25 @@ if "SET_LIMIT" == os.getenv("DEMO"):
 with gr.Blocks(theme=theme, css=css) as demo:
   with gr.Tab("Chat"):
 
-    with gr.Column(elem_id="col-container"):
+    with gr.Column():
         gr.HTML(title)
-        upload_button = gr.UploadButton("Click to Upload Files", file_types=["pdf"], file_count="multiple")
+        upload_button = gr.UploadButton("Click to Upload Files", file_count="multiple")
         file_output = gr.HTML()
-        chatbot = gr.Chatbot([], elem_id="chatbot").style(height=300)
+
+        chatbot = gr.Chatbot([], elem_id="chatbot") #.style(height=300)
         msg = gr.Textbox(label="Question", placeholder="Type your question and hit Enter ")
+        with gr.Row():
+            check_memory = gr.inputs.Checkbox(label="Remember previous messages")
+            clear_button = gr.Button("CLEAR CHAT HISTORY", )
+            max_docs = gr.inputs.Slider(1, DOC_DB_LIMIT, default=3, label="Maximum querys to the DB.", step=1)
 
     with gr.Column():
+        link_output = gr.HTML("")
         sou = gr.HTML("")
 
-  with gr.Tab("Chat Options"):
-    max_docs = gr.inputs.Slider(1, DOC_DB_LIMIT, default=3, label="Maximum querys to the DB.", step=1)
-    row_table = gr.HTML("<hr><h4> </h2>")
-    clear_button = gr.Button("CLEAR CHAT HISTORY", )
-    link_output = gr.HTML("")
     clear_button.click(flag,[],[link_output]).then(dc.clr_history,[], [link_output]).then(lambda: None, None, chatbot, queue=False)
-
-    upload_button.upload(flag,[],[file_output]).then(upload_file, [upload_button, max_docs], file_output)
-
+    upload_button.upload(flag,[],[file_output]).then(upload_file, [upload_button, max_docs], file_output).then(dc.clr_history,[], [link_output]).then(lambda: None, None, chatbot, queue=False)
+    
   with gr.Tab("Change model"):
     gr.HTML("<h3>Only models from the GGML library are accepted.</h3>")
     repo_ = gr.Textbox(label="Repository" ,value="TheBloke/Llama-2-7B-Chat-GGML")
@@ -118,14 +136,25 @@ with gr.Blocks(theme=theme, css=css) as demo:
     top_p = gr.inputs.Slider(0, 100, default=50, label="Top P", step=1)
     repeat_penalty = gr.inputs.Slider(0.1, 100., default=1.2, label="Repeat penalty", step=0.1)
     change_model_button = gr.Button("Load GGML Model")
-    model_verify = gr.HTML("Loaded model Falcon 7B-instruct")
-    default_model = gr.HTML("<hr><h4>Default Model</h2>")
+    
+    default_model = gr.HTML("<hr>Default Model</h2>")
     falcon_button = gr.Button("Load FALCON 7B-Instruct")
 
-  msg.submit(predict,[msg, chatbot, max_docs],[msg, chatbot]).then(convert,[],[sou])
+    openai_gpt_model = gr.HTML("<hr>OpenAI Model gpt-3.5-turbo</h2>")
+    api_key = gr.Textbox(label="API KEY", value="api_key...")
+    openai_button = gr.Button("Load gpt-3.5-turbo")
+
+    line_ = gr.HTML("<hr> </h2>")
+    model_verify = gr.HTML("Loaded model Falcon 7B-instruct")
+
+  with gr.Tab("About"):
+    description_md = gr.Markdown(description)
+
+  msg.submit(predict,[msg, chatbot, max_docs, check_memory],[msg, chatbot]).then(convert,[],[sou])
 
   change_model_button.click(dc.change_llm,[repo_, file_, max_tokens, temperature, top_p, top_k, repeat_penalty, max_docs],[model_verify])
 
   falcon_button.click(dc.default_falcon_model, [], [model_verify])
+  openai_button.click(clear_api_key, [api_key], [api_key, model_verify])
     
-demo.launch(enable_queue=True)
+demo.launch(debug=True,share=True,  enable_queue=True)

conversadocs/bones.py

@@ -38,7 +38,7 @@ llm_api=HuggingFaceHub(
 
 
 #alter
-def load_db(files, chain_type, k, llm):
+def load_db(files):
     EXTENSIONS = {
         ".txt": (TextLoader, {"encoding": "utf8"}),
         ".pdf": (PyPDFLoader, {}),
@@ -81,7 +81,9 @@ def load_db(files, chain_type, k, llm):
 
     # create vector database from data
     db = DocArrayInMemorySearch.from_documents(docs, embeddings)
-    # define retriever
+    return db
+
+def q_a(db, chain_type="stuff", k=3, llm=None):
     retriever = db.as_retriever(search_type="similarity", search_kwargs={"k": k})
     # create a chatbot chain. Memory is managed externally.
     qa = ConversationalRetrievalChain.from_llm(
@@ -94,6 +96,7 @@ def load_db(files, chain_type, k, llm):
     return qa
 
 
+
 class DocChat(param.Parameterized):
     chat_history = param.List([])
     answer = param.String("")
@@ -106,28 +109,42 @@ class DocChat(param.Parameterized):
     def __init__(self,  **params):
         super(DocChat, self).__init__( **params)
         self.loaded_file = "demo_docs/demo.txt"
-        self.qa = load_db(self.loaded_file,"stuff", self.k_value, self.llm)
+        self.db = load_db(self.loaded_file)
+        self.qa = q_a(self.db, "stuff", self.k_value, self.llm)
+        
 
     def call_load_db(self, path_file, k):
         if not os.path.exists(path_file[0]):  # init or no file specified
             return "No file loaded"
         else:
           try:
-            self.qa = load_db(path_file, "stuff", k, self.llm)
+            self.db = load_db(path_file)
             self.loaded_file = path_file
+            self.qa = q_a(self.db, "stuff", k, self.llm)
+            self.k_value = k
+            #self.clr_history()
+            return f"New DB created and history cleared | Loaded File: {self.loaded_file}"
           except:
             return f'No valid file'
-        self.clr_history()
-        return f"New DB created | Loaded File: {self.loaded_file}"
+
 
     # chat
-    def convchain(self, query, k_max):
+    def convchain(self, query, k_max, recall_previous_messages):
         if k_max != self.k_value:
-          print("Maximum querys changed, reloading DB")
-          self.qa = load_db(self.loaded_file,"stuff", k_max, self.llm)
+          print("Maximum querys changed")
+          self.qa = q_a(self.db, "stuff", k_max, self.llm)
           self.k_value = k_max
 
-        result = self.qa({"question": query, "chat_history": self.chat_history})
+        if not recall_previous_messages:
+          self.clr_history()
+
+        try:
+          result = self.qa({"question": query, "chat_history": self.chat_history})
+        except:
+          self.default_falcon_model()
+          self.qa = q_a(self.db, "stuff", k_max, self.llm)
+          result = self.qa({"question": query, "chat_history": self.chat_history})
+        
         self.chat_history.extend([(query, result["answer"])])
         self.db_query = result["generated_question"]
         self.db_response = result["source_documents"]
@@ -152,9 +169,9 @@ class DocChat(param.Parameterized):
                   top_k=top_k,
                   repeat_penalty=repeat_penalty,
                   )
-              self.qa = load_db(self.loaded_file,"stuff", k, self.llm)
+              self.qa = q_a(self.db, "stuff", k, self.llm)
               self.k_value = k
-              return f"Loaded {file_}"
+              return f"Loaded {file_} [GPU INFERENCE]"
             except:
               return "No valid model"
         else:
@@ -172,17 +189,22 @@ class DocChat(param.Parameterized):
                   top_k=top_k,
                   repeat_penalty=repeat_penalty,
                   )
-              self.qa = load_db(self.loaded_file,"stuff", k, self.llm)
+              self.qa = q_a(self.db, "stuff", k, self.llm)
               self.k_value = k
-              return f"Loaded {file_}"
+              return f"Loaded {file_} [CPU INFERENCE SLOW]"
             except:
               return "No valid model"
 
     def default_falcon_model(self):
       self.llm = llm_api[0]
-      self.qa = load_db(self.loaded_file,"stuff", self.k_value, self.llm)
-      return "Loaded model Falcon 7B-instruct"
-
+      self.qa = q_a(self.db, "stuff", self.k_value, self.llm)
+      return "Loaded model Falcon 7B-instruct [API FAST INFERENCE]"
+
+    def openai_model(self, API_KEY):
+        self.llm = ChatOpenAI(temperature=0, openai_api_key=API_KEY, model_name='gpt-3.5-turbo')
+        self.qa = q_a(self.db, "stuff", self.k_value, self.llm)
+        API_KEY = ""
+        return "Loaded model OpenAI gpt-3.5-turbo [API FAST INFERENCE] | If there is no response from the API, Falcon 7B-instruct will be used."
 
     @param.depends('db_query ', )
     def get_lquest(self):

requirements.txt

@@ -11,3 +11,4 @@ huggingface_hub
 unstructured[local-inference]
 gradio==3.35.2
 param==1.13.0
+openai