Upload 14 files
Browse files- .gitattributes +1 -0
- app.py +66 -0
- models/codellama-7b.Q4_K_M.gguf +3 -0
- repochat/__init__.py +0 -0
- repochat/__pycache__/__init__.cpython-312.pyc +0 -0
- repochat/__pycache__/chain.cpython-312.pyc +0 -0
- repochat/__pycache__/db.cpython-312.pyc +0 -0
- repochat/__pycache__/git.cpython-312.pyc +0 -0
- repochat/__pycache__/models.cpython-312.pyc +0 -0
- repochat/__pycache__/utils.cpython-312.pyc +0 -0
- repochat/chain.py +40 -0
- repochat/db.py +48 -0
- repochat/git.py +32 -0
- repochat/models.py +23 -0
- repochat/utils.py +66 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
models/codellama-7b.Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
|
app.py
ADDED
|
@@ -0,0 +1,66 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import time
|
| 2 |
+
import streamlit as st
|
| 3 |
+
|
| 4 |
+
from repochat.utils import init_session_state
|
| 5 |
+
from repochat.git import git_form
|
| 6 |
+
from repochat.db import vector_db, load_to_db
|
| 7 |
+
from repochat.models import hf_embeddings, code_llama
|
| 8 |
+
from repochat.chain import response_chain
|
| 9 |
+
|
| 10 |
+
init_session_state()
|
| 11 |
+
|
| 12 |
+
st.set_page_config(
|
| 13 |
+
page_title="RepoChat",
|
| 14 |
+
page_icon="💻",
|
| 15 |
+
initial_sidebar_state="expanded",
|
| 16 |
+
menu_items={
|
| 17 |
+
'Report a bug': "https://github.com/pnkvalavala/repochat/issues",
|
| 18 |
+
'About': "No need to worry if you can't understand GitHub code or repositories anymore! Introducing RepoChat, where you can effortlessly chat and discuss all things related to GitHub repositories."
|
| 19 |
+
}
|
| 20 |
+
)
|
| 21 |
+
|
| 22 |
+
st.markdown(
|
| 23 |
+
"<h1 style='text-align: center;'>RepoChat</h1>",
|
| 24 |
+
unsafe_allow_html=True
|
| 25 |
+
)
|
| 26 |
+
|
| 27 |
+
try:
|
| 28 |
+
st.session_state["db_name"], st.session_state['git_form'] = git_form(st.session_state['repo_path'])
|
| 29 |
+
|
| 30 |
+
if st.session_state['git_form']:
|
| 31 |
+
with st.spinner('Loading the contents to database. This may take some time...'):
|
| 32 |
+
st.session_state["chroma_db"] = vector_db(
|
| 33 |
+
hf_embeddings(),
|
| 34 |
+
load_to_db(st.session_state['repo_path'])
|
| 35 |
+
)
|
| 36 |
+
with st.spinner('Loading model to memory'):
|
| 37 |
+
st.session_state["qa"] = response_chain(
|
| 38 |
+
db=st.session_state["chroma_db"],
|
| 39 |
+
llm=code_llama()
|
| 40 |
+
)
|
| 41 |
+
|
| 42 |
+
st.session_state["db_loaded"] = True
|
| 43 |
+
except TypeError:
|
| 44 |
+
pass
|
| 45 |
+
|
| 46 |
+
if st.session_state["db_loaded"]:
|
| 47 |
+
for message in st.session_state["messages"]:
|
| 48 |
+
with st.chat_message(message["role"]):
|
| 49 |
+
st.markdown(message["content"])
|
| 50 |
+
|
| 51 |
+
if prompt := st.chat_input("Enter your query"):
|
| 52 |
+
st.session_state["messages"].append({"role": "user", "content": prompt})
|
| 53 |
+
with st.chat_message("user"):
|
| 54 |
+
st.markdown(prompt)
|
| 55 |
+
|
| 56 |
+
with st.chat_message("assistant"):
|
| 57 |
+
message_placeholder = st.empty()
|
| 58 |
+
full_response = ""
|
| 59 |
+
with st.spinner("Generating response..."):
|
| 60 |
+
result = st.session_state["qa"](prompt)
|
| 61 |
+
for chunk in result['answer'].split():
|
| 62 |
+
full_response += chunk + " "
|
| 63 |
+
time.sleep(0.05)
|
| 64 |
+
message_placeholder.markdown(full_response + "▌")
|
| 65 |
+
message_placeholder.markdown(full_response)
|
| 66 |
+
st.session_state["messages"].append({"role": "assistant", "content": full_response})
|
models/codellama-7b.Q4_K_M.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c3866cb985de17d398c2d7e85bb8a9c717a76bb997dce0e72a0cb169063e6d17
|
| 3 |
+
size 4081095360
|
repochat/__init__.py
ADDED
|
File without changes
|
repochat/__pycache__/__init__.cpython-312.pyc
ADDED
|
Binary file (162 Bytes). View file
|
|
|
repochat/__pycache__/chain.cpython-312.pyc
ADDED
|
Binary file (1.43 kB). View file
|
|
|
repochat/__pycache__/db.cpython-312.pyc
ADDED
|
Binary file (2.45 kB). View file
|
|
|
repochat/__pycache__/git.cpython-312.pyc
ADDED
|
Binary file (2.35 kB). View file
|
|
|
repochat/__pycache__/models.cpython-312.pyc
ADDED
|
Binary file (1.05 kB). View file
|
|
|
repochat/__pycache__/utils.cpython-312.pyc
ADDED
|
Binary file (3.28 kB). View file
|
|
|
repochat/chain.py
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from langchain.prompts import PromptTemplate
|
| 2 |
+
from langchain.chains.conversation.memory import ConversationBufferMemory
|
| 3 |
+
from langchain.chains import ConversationalRetrievalChain
|
| 4 |
+
|
| 5 |
+
from .utils import model_prompt, custom_que_prompt
|
| 6 |
+
|
| 7 |
+
def response_chain(db, llm):
|
| 8 |
+
retriever = db.as_retriever()
|
| 9 |
+
search_kwargs = {
|
| 10 |
+
"k": 3,
|
| 11 |
+
"distance_metric": "cos"
|
| 12 |
+
}
|
| 13 |
+
|
| 14 |
+
retriever.search_kwargs.update(search_kwargs)
|
| 15 |
+
|
| 16 |
+
memory = ConversationBufferMemory(
|
| 17 |
+
memory_key='chat_history',
|
| 18 |
+
return_messages=True
|
| 19 |
+
)
|
| 20 |
+
|
| 21 |
+
model_template = model_prompt()
|
| 22 |
+
QA_CHAIN_PROMPT = PromptTemplate(
|
| 23 |
+
input_variables=["context", "question"],
|
| 24 |
+
template=model_template
|
| 25 |
+
)
|
| 26 |
+
question_prompt = PromptTemplate.from_template(custom_que_prompt())
|
| 27 |
+
|
| 28 |
+
qa = ConversationalRetrievalChain.from_llm(
|
| 29 |
+
llm=llm,
|
| 30 |
+
retriever=retriever,
|
| 31 |
+
memory=memory,
|
| 32 |
+
chain_type='stuff',
|
| 33 |
+
verbose=True,
|
| 34 |
+
combine_docs_chain_kwargs={
|
| 35 |
+
'prompt': QA_CHAIN_PROMPT
|
| 36 |
+
},
|
| 37 |
+
condense_question_prompt=question_prompt
|
| 38 |
+
)
|
| 39 |
+
|
| 40 |
+
return qa
|
repochat/db.py
ADDED
|
@@ -0,0 +1,48 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import shutil
|
| 3 |
+
import streamlit as st
|
| 4 |
+
from langchain.vectorstores import Chroma
|
| 5 |
+
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
| 6 |
+
from langchain.document_loaders import NotebookLoader, TextLoader
|
| 7 |
+
|
| 8 |
+
def vector_db(embeddings, code):
|
| 9 |
+
collection_name = "db_collection"
|
| 10 |
+
local_directory = "db_" + st.session_state["db_name"]
|
| 11 |
+
persist_directory = os.path.join(os.getcwd(), local_directory)
|
| 12 |
+
|
| 13 |
+
if os.path.exists(persist_directory):
|
| 14 |
+
shutil.rmtree(persist_directory)
|
| 15 |
+
|
| 16 |
+
vec_db = Chroma.from_documents(
|
| 17 |
+
documents=code,
|
| 18 |
+
embedding=embeddings,
|
| 19 |
+
collection_name=collection_name,
|
| 20 |
+
persist_directory=persist_directory
|
| 21 |
+
)
|
| 22 |
+
|
| 23 |
+
vec_db.persist()
|
| 24 |
+
|
| 25 |
+
return vec_db
|
| 26 |
+
|
| 27 |
+
def load_to_db(repo_path):
|
| 28 |
+
docs = []
|
| 29 |
+
for root, dirs, files in os.walk(repo_path):
|
| 30 |
+
dirs[:] = [d for d in dirs if not d.startswith('.')]
|
| 31 |
+
for filename in files:
|
| 32 |
+
if filename.startswith('.'):
|
| 33 |
+
continue
|
| 34 |
+
if filename == 'package-lock.json':
|
| 35 |
+
continue
|
| 36 |
+
file_path = os.path.join(root, filename)
|
| 37 |
+
try:
|
| 38 |
+
if file_path.endswith('.ipynb'):
|
| 39 |
+
loader = NotebookLoader(file_path)
|
| 40 |
+
else:
|
| 41 |
+
loader = TextLoader(file_path, encoding="utf-8")
|
| 42 |
+
docs.extend(loader.load_and_split())
|
| 43 |
+
except Exception as e:
|
| 44 |
+
pass
|
| 45 |
+
|
| 46 |
+
code_splitter = RecursiveCharacterTextSplitter(chunk_size=2000, chunk_overlap=200)
|
| 47 |
+
code = code_splitter.split_documents(docs)
|
| 48 |
+
return code
|
repochat/git.py
ADDED
|
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import requests
|
| 2 |
+
import streamlit as st
|
| 3 |
+
|
| 4 |
+
from .utils import url_name, clone_repo
|
| 5 |
+
|
| 6 |
+
def git_form(repo_path):
|
| 7 |
+
with st.sidebar:
|
| 8 |
+
st.title("GitHub Link")
|
| 9 |
+
with st.form("git"):
|
| 10 |
+
git_url = st.text_input("Enter GitHub Repository Link")
|
| 11 |
+
submit_git = st.form_submit_button("Submit")
|
| 12 |
+
if submit_git:
|
| 13 |
+
with st.spinner("Checking GitHub URL"):
|
| 14 |
+
if not(git_url):
|
| 15 |
+
st.warning("Enter GitHub URL")
|
| 16 |
+
st.stop()
|
| 17 |
+
try:
|
| 18 |
+
response = requests.get(git_url)
|
| 19 |
+
if response.status_code == 200 and url_name(git_url):
|
| 20 |
+
st.success("GitHub Link loaded successfully!")
|
| 21 |
+
db_name = url_name(git_url)
|
| 22 |
+
else:
|
| 23 |
+
st.error("Enter Valid GitHub Repo")
|
| 24 |
+
st.stop()
|
| 25 |
+
except requests.exceptions.MissingSchema:
|
| 26 |
+
st.error("Invalid URL. Please include the scheme (e.g., https://)")
|
| 27 |
+
st.stop()
|
| 28 |
+
|
| 29 |
+
with st.spinner(f"Cloning {db_name} Repository"):
|
| 30 |
+
clone_repo(git_url, repo_path)
|
| 31 |
+
st.success("Cloned successfully!")
|
| 32 |
+
return db_name, 1
|
repochat/models.py
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from langchain.llms import LlamaCpp
|
| 2 |
+
from langchain.embeddings import HuggingFaceEmbeddings
|
| 3 |
+
from langchain.callbacks.manager import CallbackManager
|
| 4 |
+
from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
|
| 5 |
+
|
| 6 |
+
def hf_embeddings():
|
| 7 |
+
return HuggingFaceEmbeddings(
|
| 8 |
+
model_name = "sentence-transformers/all-mpnet-base-v2",
|
| 9 |
+
)
|
| 10 |
+
|
| 11 |
+
def code_llama():
|
| 12 |
+
callbackmanager = CallbackManager([StreamingStdOutCallbackHandler()])
|
| 13 |
+
llm = LlamaCpp(
|
| 14 |
+
model_path="./models/codellama-7b.Q4_K_M.gguf",
|
| 15 |
+
n_ctx=2048,
|
| 16 |
+
max_tokens=200,
|
| 17 |
+
n_gpu_layers=1,
|
| 18 |
+
f16_kv=True,
|
| 19 |
+
callback_manager=callbackmanager,
|
| 20 |
+
verbose=True,
|
| 21 |
+
use_mlock=True
|
| 22 |
+
)
|
| 23 |
+
return llm
|
repochat/utils.py
ADDED
|
@@ -0,0 +1,66 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import re
|
| 3 |
+
import shutil
|
| 4 |
+
import subprocess
|
| 5 |
+
import streamlit as st
|
| 6 |
+
|
| 7 |
+
def init_session_state():
|
| 8 |
+
SESSION_DEFAULTS = {
|
| 9 |
+
"messages": [],
|
| 10 |
+
"chroma_db": None,
|
| 11 |
+
"db_loaded": False,
|
| 12 |
+
"repo_path": './cloned_repo',
|
| 13 |
+
"git_form": False,
|
| 14 |
+
"qa": None,
|
| 15 |
+
"db_name": None
|
| 16 |
+
}
|
| 17 |
+
|
| 18 |
+
for keys, values in SESSION_DEFAULTS.items():
|
| 19 |
+
if keys not in st.session_state:
|
| 20 |
+
st.session_state[keys] = values
|
| 21 |
+
|
| 22 |
+
def url_name(url):
|
| 23 |
+
pattern = r"https?://github.com/([^/]+)/([^/]+)"
|
| 24 |
+
match = re.match(pattern, url)
|
| 25 |
+
if match:
|
| 26 |
+
owner = match.group(1)
|
| 27 |
+
repo = match.group(2)
|
| 28 |
+
return f"{owner}_{repo}"
|
| 29 |
+
else:
|
| 30 |
+
st.error("Enter valid GitHub URL")
|
| 31 |
+
st.stop()
|
| 32 |
+
|
| 33 |
+
def clone_repo(git_url, repo_path):
|
| 34 |
+
if os.path.exists(repo_path):
|
| 35 |
+
shutil.rmtree(repo_path)
|
| 36 |
+
|
| 37 |
+
git_url = git_url.replace(".git", "")
|
| 38 |
+
|
| 39 |
+
command = f'git clone {git_url}.git {repo_path} && rm -rf {repo_path}/.git'
|
| 40 |
+
subprocess.run(command, shell=True)
|
| 41 |
+
|
| 42 |
+
def prompt_format(system_prompt, instruction):
|
| 43 |
+
B_INST, E_INST = "[INST]", "[/INST]"
|
| 44 |
+
B_SYS, E_SYS = "<SYS>>\n", "\n<</SYS>>\n\n"
|
| 45 |
+
SYSTEM_PROMPT = B_SYS + system_prompt + E_SYS
|
| 46 |
+
prompt_template = B_INST + SYSTEM_PROMPT + instruction + E_INST
|
| 47 |
+
return prompt_template
|
| 48 |
+
|
| 49 |
+
def model_prompt():
|
| 50 |
+
system_prompt = """You are a helpful assistant, you have good knowledge in coding and you will use the provided context to answer user questions with detailed explanations.
|
| 51 |
+
Read the given context before answering questions and think step by step. If you can not answer a user question based on the provided context, inform the user. Do not use any other information for answering user"""
|
| 52 |
+
instruction = """
|
| 53 |
+
Context: {context}
|
| 54 |
+
User: {question}"""
|
| 55 |
+
return prompt_format(system_prompt, instruction)
|
| 56 |
+
|
| 57 |
+
def custom_que_prompt():
|
| 58 |
+
que_system_prompt = """Given the following conversation and a follow up question, rephrase the follow up question to be a standalone question and give only the standalone question as output in the tags <question> and </question>.
|
| 59 |
+
"""
|
| 60 |
+
|
| 61 |
+
instr_prompt = """Chat History:
|
| 62 |
+
{chat_history}
|
| 63 |
+
Follow Up Input: {question}
|
| 64 |
+
Standalone question:"""
|
| 65 |
+
|
| 66 |
+
return prompt_format(que_system_prompt, instr_prompt)
|