initial commit

.gitignore

@@ -0,0 +1,161 @@
+models/**
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/#use-with-ide
+.pdm.toml
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2023 Karim Lalani
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -0,0 +1,46 @@
+# Streamlit + Langchain + LLama.cpp w/ Mistral + Conversational Memory
+
+Run your own AI Chatbot locally without a GPU.
+
+To make that possible, we use the [Mistral 7b](https://mistral.ai/news/announcing-mistral-7b/) model.  
+However, you can use any quantized model that is supported by [llama.cpp](https://github.com/ggerganov/llama.cpp).
+
+This model will chatbot will allow you to define it's personality and respond to the questions accordingly.  
+This example remembers the chat history allowing you to ask follow up questions.
+
+# TL;DR instructions
+
+1. Install llama-cpp-python
+2. Install langchain
+3. Install streamlit
+4. Download Mistral from HuggingFace from TheBloke's repo: mistral-7b-instruct-v0.1.Q4_0.gguf
+5. Place model file in the `models` subfolder
+6. Run streamlit
+
+# Step by Step instructions
+
+The setup assumes you have `python` already installed and `venv` module available.
+
+1. Download the code or clone the repository.
+2. Inside the root folder of the repository, initialize a python virtual environment:
+```bash
+python -m venv venv
+```
+3. Activate the python envitonment:
+```bash
+source venv/bin/activate
+```
+4. Install required packages (`langchain`, `llama.cpp`, and `streamlit`):
+```bash
+pip install -r requirements.txt
+```
+5. Create a subdirectory to place the models in:
+```bash
+mkdir -p models
+```
+6. Download the `Mistral7b` quantized model from `huggingface` from the following link:
+[mistral-7b-instruct-v0.1.Q4_0.gguf](https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF/resolve/main/mistral-7b-instruct-v0.1.Q4_0.gguf)
+7. Start `streamlit`:
+```bash
+streamlit run main.py
+```

main.py

@@ -0,0 +1,140 @@
+import streamlit as st
+from langchain.llms import LlamaCpp
+from langchain.schema import SystemMessage
+from langchain.prompts import ChatPromptTemplate, HumanMessagePromptTemplate, MessagesPlaceholder, PromptTemplate
+from langchain.chains import LLMChain
+from langchain.memory import ConversationBufferMemory
+# from langchain.callbacks.manager import CallbackManager
+from langchain.callbacks.base import BaseCallbackHandler
+import json
+
+
+# StreamHandler to intercept streaming output from the LLM.
+# This makes it appear that the Language Model is "typing"
+# in realtime.
+class StreamHandler(BaseCallbackHandler):
+    def __init__(self, container, initial_text=""):
+        self.container = container
+        self.text = initial_text
+
+    def on_llm_new_token(self, token: str, **kwargs) -> None:
+        self.text += token
+        self.container.markdown(self.text)
+
+
+@st.cache_resource
+def create_chain(system_prompt):
+    # A stream handler to direct streaming output on the chat screen.
+    # This will need to be handled somewhat differently.
+    # But it demonstrates what potential it carries.
+    # stream_handler = StreamHandler(st.empty())
+
+    # Callback manager is a way to intercept streaming output from the
+    # LLM and take some action on it. Here we are giving it our custom
+    # stream handler to make it appear as if the LLM is typing the
+    # responses in real time.
+    # callback_manager = CallbackManager([stream_handler])
+
+    llm = LlamaCpp(
+            model_path="models/mistral-7b-instruct-v0.1.Q4_0.gguf",
+            temperature=0,
+            max_tokens=512,
+            top_p=1,
+            # callback_manager=callback_manager,
+            verbose=False,
+            streaming=True,
+            )
+
+    # Template you will use to structure your user input into before converting
+    # into a prompt. Here, my template first injects the personality I wish to give
+    # to the LLM before in the form of system_prompt pushing the actual prompt from the user.
+    # Then we'll inject the chat history followed by the user prompt and a placeholder token
+    # for the LLM to complete.
+    template = """
+    {}
+
+    {}
+
+    Human: {}
+    AI: 
+    """.format(system_prompt, "{chat_history}","{human_input}")
+
+    # We create a prompt from the template so we can use it with langchain
+    # prompt = ChatPromptTemplate.from_messages([
+    #     SystemMessage(content=system_prompt),
+    #     MessagesPlaceholder(variable_name="chat_history"),
+    #     HumanMessagePromptTemplate.from_template("{human_input}")
+    #     ])
+    prompt = PromptTemplate(input_variables=["chat_history","human_input"], template=template)
+
+    # Conversation buffer memory will keep track of the conversation in the memory
+    memory = ConversationBufferMemory(memory_key="chat_history")
+
+    # We create an llm chain with our llm with prompt and memory
+    llm_chain = LLMChain(prompt=prompt, llm=llm, memory=memory, verbose=True)
+
+    return llm_chain
+
+
+# Set the webpage title
+st.set_page_config(
+    page_title="Your own Chat!"
+)
+
+# Create a header element
+st.header("Your own Chat!")
+
+# This sets the LLM's personality for each prompt.
+# The initial personality privided is basic.
+# Try something interesting and notice how the LLM responses are affected.
+system_prompt = st.text_area(
+    label="System Prompt",
+    value="You are a helpful AI assistant who answers questions in short sentences.",
+    key="system_prompt")
+
+# Create llm chain to use for our chat bot.
+llm_chain = create_chain(system_prompt)
+
+# We store the conversation in the session state.
+# This will be used to render the chat conversation.
+# We initialize it with the first message we want to be greeted with.
+if "messages" not in st.session_state:
+    st.session_state.messages = [
+        {"role": "assistant", "content": "How may I help you today?"}
+    ]
+
+if "current_response" not in st.session_state:
+    st.session_state.current_response = ""
+
+# We loop through each message in the session state and render it as
+# a chat message.
+for message in st.session_state.messages:
+    with st.chat_message(message["role"]):
+        st.markdown(message["content"])
+
+# We take questions/instructions from the chat input to pass to the LLM
+if user_prompt := st.chat_input("Your message here", key="user_input"):
+
+    # Add our input to the session state
+    st.session_state.messages.append(
+        {"role": "user", "content": user_prompt}
+    )
+
+    # Add our input to the chat window
+    with st.chat_message("user"):
+        st.markdown(user_prompt)
+
+    # Pass our input to the llm chain and capture the final responses.
+    # It is worth noting that the Stream Handler is already receiving the
+    # streaming response as the llm is generating. We get our response
+    # here once the llm has finished generating the complete response.
+    response = llm_chain.predict(human_input=user_prompt)
+
+    # Add the response to the session state
+    st.session_state.messages.append(
+        {"role": "assistant", "content": response}
+    )
+
+    # Add the response to the chat window
+    with st.chat_message("assistant"):
+        st.markdown(response)

requirements.txt

@@ -0,0 +1,3 @@
+langchain==0.0.321
+llama_cpp_python==0.2.11
+streamlit==1.27.2