Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,53 +1,11 @@
|
|
| 1 |
import os
|
| 2 |
-
from typing import Any, List, Mapping, Optional, Dict
|
| 3 |
import chainlit as cl
|
| 4 |
from langchain_community.embeddings import HuggingFaceEmbeddings
|
| 5 |
from langchain.prompts import ChatPromptTemplate
|
| 6 |
from langchain_core.output_parsers import StrOutputParser
|
| 7 |
from langchain_core.runnables import RunnablePassthrough
|
| 8 |
from langchain_community.vectorstores import Chroma
|
| 9 |
-
from langchain.callbacks.manager import CallbackManagerForLLMRun
|
| 10 |
-
from langchain.llms.base import LLM
|
| 11 |
from llama_cpp import Llama
|
| 12 |
-
from pydantic import Field, BaseModel
|
| 13 |
-
|
| 14 |
-
class LlamaCppLLM(LLM, BaseModel):
|
| 15 |
-
"""Custom LangChain wrapper for llama.cpp"""
|
| 16 |
-
|
| 17 |
-
client: Any = Field(default=None, exclude=True)
|
| 18 |
-
model_path: str = Field(..., description="Path to the model file")
|
| 19 |
-
n_ctx: int = Field(default=2048, description="Context window size")
|
| 20 |
-
n_threads: int = Field(default=4, description="Number of CPU threads")
|
| 21 |
-
n_gpu_layers: int = Field(default=0, description="Number of GPU layers")
|
| 22 |
-
|
| 23 |
-
def __init__(self, **kwargs):
|
| 24 |
-
super().__init__(**kwargs)
|
| 25 |
-
self.client = Llama(
|
| 26 |
-
model_path=self.model_path,
|
| 27 |
-
n_ctx=self.n_ctx,
|
| 28 |
-
n_threads=self.n_threads,
|
| 29 |
-
n_gpu_layers=self.n_gpu_layers
|
| 30 |
-
)
|
| 31 |
-
|
| 32 |
-
@property
|
| 33 |
-
def _llm_type(self) -> str:
|
| 34 |
-
return "llama.cpp"
|
| 35 |
-
|
| 36 |
-
def _call(
|
| 37 |
-
self,
|
| 38 |
-
prompt: str,
|
| 39 |
-
stop: Optional[List[str]] = None,
|
| 40 |
-
run_manager: Optional[CallbackManagerForLLMRun] = None,
|
| 41 |
-
**kwargs: Any,
|
| 42 |
-
) -> str:
|
| 43 |
-
if not self.client:
|
| 44 |
-
raise RuntimeError("Model not initialized")
|
| 45 |
-
|
| 46 |
-
response = self.client.create_chat_completion(
|
| 47 |
-
messages=[{"role": "user", "content": prompt}],
|
| 48 |
-
**kwargs
|
| 49 |
-
)
|
| 50 |
-
return response["choices"][0]["message"]["content"]
|
| 51 |
|
| 52 |
# Initialize the embedding model
|
| 53 |
embeddings = HuggingFaceEmbeddings(
|
|
@@ -58,15 +16,15 @@ embeddings = HuggingFaceEmbeddings(
|
|
| 58 |
|
| 59 |
# Load the existing Chroma vector store
|
| 60 |
persist_directory = os.path.join(os.path.dirname(__file__), 'mydb')
|
| 61 |
-
vectorstore = Chroma(
|
|
|
|
|
|
|
|
|
|
| 62 |
|
| 63 |
-
# Initialize the
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
n_ctx=2048,
|
| 68 |
-
n_threads=4,
|
| 69 |
-
n_gpu_layers=0
|
| 70 |
)
|
| 71 |
|
| 72 |
# Create the RAG prompt template
|
|
@@ -84,39 +42,39 @@ prompt = ChatPromptTemplate.from_template(template)
|
|
| 84 |
|
| 85 |
@cl.on_chat_start
|
| 86 |
async def start():
|
| 87 |
-
# Send initial message
|
| 88 |
await cl.Message(
|
| 89 |
content="Hi! I'm ready to answer your questions based on the stored documents. What would you like to know?"
|
| 90 |
).send()
|
| 91 |
|
| 92 |
@cl.on_message
|
| 93 |
async def main(message: cl.Message):
|
| 94 |
-
# Create a loading message
|
| 95 |
msg = cl.Message(content="")
|
| 96 |
await msg.send()
|
| 97 |
|
| 98 |
-
# Start typing effect
|
| 99 |
async with cl.Step(name="Searching documents..."):
|
| 100 |
try:
|
| 101 |
-
# Search the vector store
|
| 102 |
retriever = vectorstore.as_retriever(search_kwargs={"k": 3})
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 103 |
|
| 104 |
-
#
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
|
|
|
|
|
|
| 110 |
)
|
|
|
|
| 111 |
|
| 112 |
-
# Execute the chain
|
| 113 |
-
response = await cl.make_async(rag_chain)(message.content)
|
| 114 |
-
|
| 115 |
# Update loading message with response
|
| 116 |
-
await msg.update(content=
|
| 117 |
-
|
| 118 |
# Show source documents
|
| 119 |
-
docs = retriever.get_relevant_documents(message.content)
|
| 120 |
elements = []
|
| 121 |
for i, doc in enumerate(docs):
|
| 122 |
source_name = f"Source {i+1}"
|
|
@@ -133,4 +91,4 @@ async def main(message: cl.Message):
|
|
| 133 |
await msg.update(content=error_msg)
|
| 134 |
|
| 135 |
if __name__ == '__main__':
|
| 136 |
-
cl.
|
|
|
|
| 1 |
import os
|
|
|
|
| 2 |
import chainlit as cl
|
| 3 |
from langchain_community.embeddings import HuggingFaceEmbeddings
|
| 4 |
from langchain.prompts import ChatPromptTemplate
|
| 5 |
from langchain_core.output_parsers import StrOutputParser
|
| 6 |
from langchain_core.runnables import RunnablePassthrough
|
| 7 |
from langchain_community.vectorstores import Chroma
|
|
|
|
|
|
|
| 8 |
from llama_cpp import Llama
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
|
| 10 |
# Initialize the embedding model
|
| 11 |
embeddings = HuggingFaceEmbeddings(
|
|
|
|
| 16 |
|
| 17 |
# Load the existing Chroma vector store
|
| 18 |
persist_directory = os.path.join(os.path.dirname(__file__), 'mydb')
|
| 19 |
+
vectorstore = Chroma(
|
| 20 |
+
persist_directory=persist_directory,
|
| 21 |
+
embedding_function=embeddings
|
| 22 |
+
)
|
| 23 |
|
| 24 |
+
# Initialize the Llama model using from_pretrained
|
| 25 |
+
llm = Llama.from_pretrained(
|
| 26 |
+
repo_id="bartowski/Llama-3.2-1B-Instruct-GGUF",
|
| 27 |
+
filename="Llama-3.2-1B-Instruct-Q8_0.gguf",
|
|
|
|
|
|
|
|
|
|
| 28 |
)
|
| 29 |
|
| 30 |
# Create the RAG prompt template
|
|
|
|
| 42 |
|
| 43 |
@cl.on_chat_start
|
| 44 |
async def start():
|
|
|
|
| 45 |
await cl.Message(
|
| 46 |
content="Hi! I'm ready to answer your questions based on the stored documents. What would you like to know?"
|
| 47 |
).send()
|
| 48 |
|
| 49 |
@cl.on_message
|
| 50 |
async def main(message: cl.Message):
|
|
|
|
| 51 |
msg = cl.Message(content="")
|
| 52 |
await msg.send()
|
| 53 |
|
|
|
|
| 54 |
async with cl.Step(name="Searching documents..."):
|
| 55 |
try:
|
|
|
|
| 56 |
retriever = vectorstore.as_retriever(search_kwargs={"k": 3})
|
| 57 |
+
docs = retriever.get_relevant_documents(message.content)
|
| 58 |
+
context = "\n\n".join([doc.page_content for doc in docs])
|
| 59 |
+
|
| 60 |
+
# Format the prompt
|
| 61 |
+
final_prompt = prompt.format(context=context, question=message.content)
|
| 62 |
|
| 63 |
+
# Generate response using the Llama model
|
| 64 |
+
response = llm.create_chat_completion(
|
| 65 |
+
messages=[
|
| 66 |
+
{
|
| 67 |
+
"role": "user",
|
| 68 |
+
"content": final_prompt
|
| 69 |
+
}
|
| 70 |
+
]
|
| 71 |
)
|
| 72 |
+
assistant_reply = response['choices'][0]['message']['content']
|
| 73 |
|
|
|
|
|
|
|
|
|
|
| 74 |
# Update loading message with response
|
| 75 |
+
await msg.update(content=assistant_reply)
|
| 76 |
+
|
| 77 |
# Show source documents
|
|
|
|
| 78 |
elements = []
|
| 79 |
for i, doc in enumerate(docs):
|
| 80 |
source_name = f"Source {i+1}"
|
|
|
|
| 91 |
await msg.update(content=error_msg)
|
| 92 |
|
| 93 |
if __name__ == '__main__':
|
| 94 |
+
cl.run()
|