Merge pull request #5 from sbslee/0.6.0-dev

0.6.0 dev

CHANGELOG.md

@@ -1,5 +1,9 @@
 # CHANGELOG
 
+## 0.6.0 (2023-06-12)
+* Enable users to chat with .csv documents.
+* Enable users to customize chat settings (e.g. font size and background color).
+
 ## 0.5.0 (2023-06-08)
 * Enable users to ask follow-up questions when chatting with documents.
 * Enable users to customize chatbot parameters by uploading a configuration file.

README.md

@@ -2,30 +2,71 @@
 
 [![PyPI](https://badge.fury.io/py/kanu.svg)](https://badge.fury.io/py/kanu)
 
-KANU is a minimalistic Python-based GUI for various chatbots.
+Welcome to KANU, a minimalistic Python-based GUI for various chatbots.
 
-## Chatbots
+There are currently two chatbots available in KANU:
 
-### ChatGPT
+- [ChatGPT](#chatgpt) harnesses the power of ChatGPT, bringing it directly to your local computer
+- [DocGPT](#docgpt) allows you to effortlessly interact with your documents and ask questions about them
 
-![Alt Text](https://raw.githubusercontent.com/sbslee/kanu/main/images/chatgpt.gif)
+Other features of KANU inclde:
 
-### DocGPT
-
-![Alt Text](https://raw.githubusercontent.com/sbslee/kanu/main/images/docgpt.gif)
+- Customize chat settings (e.g. font size and background color)
+- Customize chatbot parameters (e.g. prompt, temperature, and chunk size) by directly using the GUI or uploading a configuration file
 
 ## Installation
 
+The recommended way is via pip:
+
 ```
 $ pip install kanu
 ```
 
+KANU requires a different set of dependencies for each chatbot. You can find the dependencies specific to each chatbot in the [Chatbots](#chatbots) section.
+
 ## Running
 
 ```
 $ kanu
 ```
 
+<a id="chatbots"></a>
+## Chatbots
+
+<a id="chatgpt"></a>
+### ChatGPT
+
+![Alt Text](https://raw.githubusercontent.com/sbslee/kanu/main/images/chatgpt.gif)
+
+The following packages are required to run ChatGPT:
+
+```
+openai # Required.
+```
+
+<a id="docgpt"></a>
+### DocGPT
+
+![Alt Text](https://raw.githubusercontent.com/sbslee/kanu/main/images/docgpt.gif)
+
+The following document formats are supported by DocGPT:
+
+- .txt
+- .pdf
+- .doc and .docx
+- .csv
+
+The following packages are required to run DocGPT:
+
+```
+langchain    # Required.
+chromadb     # Required. 
+tiktoken     # Required.
+pdfminer.six # Optional. Only required for .pdf documents.
+unstructured # Optional. Only required for .doc and .docx documents.
+tabulate     # Optional. Only required for .doc and .docx documents.
+```
+
 ## Changelog
 
 See the [CHANGELOG.md](https://github.com/sbslee/kanu/blob/main/CHANGELOG.md) file for details.

kanu/__main__.py

@@ -84,14 +84,20 @@ def config_chatgpt(self):
 
     def parse_chatgpt_config(self):
         config = configparser.ConfigParser()
-        config.read(filedialog.askopenfilename())
+        file_path = filedialog.askopenfilename()
+        if not file_path:
+            return
+        config.read(file_path)
         self.deploy_agent("ChatGPT", config["USER"]["openai_key"], config["DEFAULT"]["model"], float(config["DEFAULT"]["temperature"]), config["DEFAULT"]["prompt"])
 
     def template_chatgpt_config(self):
+        file_path = filedialog.asksaveasfilename()
+        if not file_path:
+            return
         config = configparser.ConfigParser()
         config["DEFAULT"] = {"model": "gpt-3.5-turbo", "temperature": "0.5", "prompt": CHATGPT_PROMPT}
         config["USER"] = {"openai_key": ""}
-        with open(filedialog.asksaveasfilename(), "w") as f:
+        with open(file_path, "w") as f:
             config.write(f)
 
     def config_docgpt(self):
@@ -144,21 +150,27 @@ def config_docgpt(self):
         l.grid(row=18, column=0, columnspan=2)
         e = tk.Entry(self.container)
         e.grid(row=19, column=0, columnspan=2)
-        b = tk.Button(self.container, text="Submit", command=lambda: self.deploy_agent("DocGPT", e.get(), self.model.get(), self.prompt.get("1.0", "end-1c"), self.temperature.get()))
+        b = tk.Button(self.container, text="Submit", command=lambda: self.deploy_agent("DocGPT", e.get(), self.model.get(), self.prompt.get("1.0", "end-1c"), self.temperature.get(), 1000, 50))
         b.grid(row=20, column=0)
         b = tk.Button(self.container, text="Go back", command=lambda: self.homepage())
         b.grid(row=20, column=1)
 
     def parse_docgpt_config(self):
         config = configparser.ConfigParser()
-        config.read(filedialog.askopenfilename())
-        self.deploy_agent("DocGPT", config["USER"]["openai_key"], config["DEFAULT"]["model"], float(config["DEFAULT"]["temperature"]), config["DEFAULT"]["prompt"])
+        file_path = filedialog.askopenfilename()
+        if not file_path:
+            return
+        config.read(file_path)
+        self.deploy_agent("DocGPT", config["USER"]["openai_key"], config["DEFAULT"]["model"], float(config["DEFAULT"]["temperature"]), config["DEFAULT"]["prompt"], config["DEFAULT"]["chunk_size"], config["DEFAULT"]["chunk_overlap"])
 
     def template_docgpt_config(self):
+        file_path = filedialog.asksaveasfilename()
+        if not file_path:
+            return
         config = configparser.ConfigParser()
-        config["DEFAULT"] = {"model": "gpt-3.5-turbo", "temperature": "0.5", "prompt": DOCGPT_PROMPT}
+        config["DEFAULT"] = {"model": "gpt-3.5-turbo", "temperature": "0.5", "prompt": DOCGPT_PROMPT, "chunk_size": 1000, "chunk_overlap": 50}
         config["USER"] = {"openai_key": ""}
-        with open(filedialog.asksaveasfilename(), "w") as f:
+        with open(file_path, "w") as f:
             config.write(f)
 
     def deploy_agent(self, agent, *args, **kwargs):

kanu/chatgpt.py

@@ -2,31 +2,38 @@
 
 import openai
 
+from .utils import Settings
+
 class ChatGPT:
     def __init__(self, kanu, openai_key, model, temperature, prompt):
         self.kanu = kanu
         self.model = model
         self.temperature = temperature
         self.prompt = prompt
         openai.api_key = openai_key
+        self.settings = Settings(self)
 
     def run(self):
         self.kanu.container.pack_forget()
         self.kanu.container = tk.Frame(self.kanu.root)
         self.kanu.container.pack()
         l = tk.Label(self.kanu.container, text="ChatGPT")
-        l.grid(row=0, column=0, columnspan=3)
+        l.grid(row=0, column=0, columnspan=4)
         self.session = tk.Text(self.kanu.container, width=70, height=20)
-        self.session.grid(row=1, column=0, columnspan=3)
-        e = tk.Entry(self.kanu.container, width=54)
-        e.grid(row=2, column=0, columnspan=3)
+        self.session.grid(row=1, column=0, columnspan=4)
+        self.session.tag_config("user", **self.settings.get_user_kwargs())
+        self.session.tag_config("bot", **self.settings.get_bot_kwargs())
+        user_input = tk.Entry(self.kanu.container, width=54)
+        user_input.grid(row=2, column=0, columnspan=4)
         self.messages = []
-        b = tk.Button(self.kanu.container, text="Send", command=lambda: self.send_message(e))
+        b = tk.Button(self.kanu.container, text="Send", command=lambda: self.send_message(user_input))
         b.grid(row=3, column=0)
         b = tk.Button(self.kanu.container, text="Clear", command=lambda: self.clear_session())
         b.grid(row=3, column=1)
         b = tk.Button(self.kanu.container, text="Go back", command=lambda: self.kanu.config_chatgpt())
         b.grid(row=3, column=2)
+        b = tk.Button(self.kanu.container, text="Settings", command=lambda: self.settings.page())
+        b.grid(row=3, column=3)
 
     def send_message(self, entry):
         if not self.messages:
@@ -39,8 +46,8 @@ def send_message(self, entry):
         )
         response = bot_response["choices"][0]["message"]["content"]
         self.messages += [{"role": "assistant", "content": response}]
-        self.session.insert(tk.END, "You: " + entry.get() + "\n")
-        self.session.insert(tk.END, f"Bot: " + response + "\n")
+        self.session.insert(tk.END, "You: " + entry.get() + "\n", "user")
+        self.session.insert(tk.END, f"Bot: " + response + "\n", "bot")
         entry.delete(0, tk.END)
 
     def clear_session(self):

kanu/docgpt.py

@@ -13,25 +13,30 @@
 from langchain.document_loaders import (
     TextLoader,
     PDFMinerLoader,
-    UnstructuredWordDocumentLoader
+    UnstructuredWordDocumentLoader,
+    CSVLoader,
 )
 
-from .utils import Tooltip
+from .utils import Tooltip, Settings
 
 DOCUMENT_LOADERS = {
     ".txt": (TextLoader, {"encoding": "utf8"}),
     ".pdf": (PDFMinerLoader, {}),
     ".doc": (UnstructuredWordDocumentLoader, {}),
     ".docx": (UnstructuredWordDocumentLoader, {}),
+    ".csv": (CSVLoader, {}),
 }
 
 class DocGPT:
-    def __init__(self, kanu, openai_key, model, temperature, prompt):
+    def __init__(self, kanu, openai_key, model, temperature, prompt, default_chunk_size, default_chunk_overlap):
         self.kanu = kanu
         self.model = model
         self.temperature = temperature
         self.prompt = prompt
+        self.default_chunk_size = default_chunk_size
+        self.default_chunk_overlap = default_chunk_overlap
         os.environ["OPENAI_API_KEY"] = openai_key
+        self.settings = Settings(self)
 
     def run(self):
         self.kanu.container.pack_forget()
@@ -62,13 +67,13 @@ def run(self):
         l = tk.Label(self.kanu.container, text="Chunk size ⓘ:")
         Tooltip(l, "The maximum number of characters in each chunk.")
         l.grid(row=5, column=0)
-        self.chunk_size = tk.IntVar(self.kanu.container, value=1000)
+        self.chunk_size = tk.IntVar(self.kanu.container, value=self.default_chunk_size)
         e = tk.Entry(self.kanu.container, textvariable=self.chunk_size)
         e.grid(row=5, column=1, columnspan=2)
         l = tk.Label(self.kanu.container, text="Chunk overlap ⓘ:")
         Tooltip(l, "The number of overlapping characters between adjacent chunks.")
         l.grid(row=6, column=0)
-        self.chunk_overlap = tk.IntVar(self.kanu.container, value=50)
+        self.chunk_overlap = tk.IntVar(self.kanu.container, value=self.default_chunk_overlap)
         e = tk.Entry(self.kanu.container, textvariable=self.chunk_overlap)
         e.grid(row=6, column=1, columnspan=2)
         self.option1_button = tk.Button(self.kanu.container, text="Go with Option 1", command=self.go_with_option1)
@@ -101,22 +106,26 @@ def query(self):
         self.kanu.container = tk.Frame(self.kanu.root)
         self.kanu.container.pack()
         l = tk.Label(self.kanu.container, text="DocGPT")
-        l.grid(row=0, column=0, columnspan=3)    
+        l.grid(row=0, column=0, columnspan=4)    
         self.session = tk.Text(self.kanu.container, width=70, height=20)
-        self.session.grid(row=1, column=0, columnspan=3)
-        e = tk.Entry(self.kanu.container, width=54)
-        e.grid(row=2, column=0, columnspan=3)
-        b = tk.Button(self.kanu.container, text="Send", command=lambda: self.send_message(e))
+        self.session.grid(row=1, column=0, columnspan=4)
+        self.session.tag_config("user", **self.settings.get_user_kwargs())
+        self.session.tag_config("bot", **self.settings.get_bot_kwargs())
+        user_input = tk.Entry(self.kanu.container, width=54)
+        user_input.grid(row=2, column=0, columnspan=4)
+        b = tk.Button(self.kanu.container, text="Send", command=lambda: self.send_message(user_input))
         b.grid(row=3, column=0)
         b = tk.Button(self.kanu.container, text="Clear", command=lambda: self.clear_session())
         b.grid(row=3, column=1)
         b = tk.Button(self.kanu.container, text="Go back", command=lambda: self.run())
         b.grid(row=3, column=2)
+        b = tk.Button(self.kanu.container, text="Settings", command=lambda: self.settings.page())
+        b.grid(row=3, column=3)
 
     def send_message(self, entry):
-        self.session.insert(tk.END, "You: " + entry.get() + "\n")
+        self.session.insert(tk.END, "You: " + entry.get() + "\n", "user")
         response = self.qa(entry.get())["answer"]
-        self.session.insert(tk.END, "Bot: " + response + "\n")
+        self.session.insert(tk.END, "Bot: " + response + "\n", "bot")
         entry.delete(0, tk.END)
 
     def go_with_option1(self):
@@ -129,12 +138,12 @@ def go_with_option1(self):
                     continue
                 loader_class, loader_kwargs = DOCUMENT_LOADERS[file_ext]
                 loader = loader_class(file_path, **loader_kwargs)
-                document = loader.load()[0]
-                documents.append(document)
+                document = loader.load()
+                documents.extend(document)
         text_splitter = RecursiveCharacterTextSplitter(chunk_size=self.chunk_size.get(), chunk_overlap=self.chunk_overlap.get())
         texts = text_splitter.split_documents(documents)
         db = Chroma.from_documents(texts, OpenAIEmbeddings(), persist_directory=self.database_directory)
-        db.add_documents(texts)    
+        db.add_documents(texts)
         db.persist()
         db = None
         self.query()