Rulga's picture
message till history
ec0b4e0
raw
history blame
17.5 kB
"""
Module for managing dataset on Hugging Face Hub
"""
import os
import json
import tempfile
from typing import Tuple, List, Dict, Any, Optional, Union
from datetime import datetime
from huggingface_hub import HfApi, HfFolder
from langchain_community.vectorstores import FAISS
from config.settings import (
VECTOR_STORE_PATH,
HF_TOKEN,
EMBEDDING_MODEL,
DATASET_ID,
CHAT_HISTORY_PATH,
DATASET_CHAT_HISTORY_PATH,
DATASET_VECTOR_STORE_PATH,
DATASET_FINE_TUNED_PATH,
DATASET_ANNOTATIONS_PATH
)
from langchain_huggingface import HuggingFaceEmbeddings
import logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
class DatasetManager:
def __init__(self, dataset_name: Optional[str] = None, token: Optional[str] = None):
self.dataset_name = dataset_name or DATASET_ID
self.token = token if token else HF_TOKEN
self.api = HfApi(token=self.token)
# Use paths from settings
self.vector_store_path = DATASET_VECTOR_STORE_PATH
self.chat_history_path = DATASET_CHAT_HISTORY_PATH
self.fine_tuned_path = DATASET_FINE_TUNED_PATH
self.annotations_path = DATASET_ANNOTATIONS_PATH
def init_dataset_structure(self) -> Tuple[bool, str]:
"""
Initialize dataset structure with required directories
Returns:
(success, message)
"""
try:
# Check if repository exists
try:
self.api.repo_info(repo_id=self.dataset_name, repo_type="dataset")
except Exception:
# Create repository if it doesn't exist
self.api.create_repo(repo_id=self.dataset_name, repo_type="dataset", private=True)
# Create empty .gitkeep files to maintain structure
directories = ["vector_store", "chat_history", "documents"]
for directory in directories:
with tempfile.NamedTemporaryFile(delete=False) as temp:
temp_path = temp.name
try:
self.api.upload_file(
path_or_fileobj=temp_path,
path_in_repo=f"{directory}/.gitkeep",
repo_id=self.dataset_name,
repo_type="dataset"
)
finally:
if os.path.exists(temp_path):
os.remove(temp_path)
return True, "Dataset structure initialized successfully"
except Exception as e:
return False, f"Error initializing dataset structure: {str(e)}"
def upload_vector_store(self, vector_store: FAISS) -> Tuple[bool, str]:
"""
Upload vector store to dataset
Args:
vector_store: FAISS vector store to upload
Returns:
(success, message)
"""
try:
with tempfile.TemporaryDirectory() as temp_dir:
# Save vector store to temporary directory
vector_store.save_local(folder_path=temp_dir)
index_path = os.path.join(temp_dir, "index.faiss")
config_path = os.path.join(temp_dir, "index.pkl")
# Add debug logging
print(f"Debug - Checking files before upload:")
print(f"index.faiss exists: {os.path.exists(index_path)}, size: {os.path.getsize(index_path) if os.path.exists(index_path) else 0} bytes")
print(f"index.pkl exists: {os.path.exists(config_path)}, size: {os.path.getsize(config_path) if os.path.exists(config_path) else 0} bytes")
if not os.path.exists(index_path) or not os.path.exists(config_path):
return False, "Vector store files not created"
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
# First save old files to archive if they exist
try:
# Check for existing files
self.api.hf_hub_download(
repo_id=self.dataset_name,
filename="vector_store/index.faiss",
repo_type="dataset"
)
# If file exists, create archive copy
self.api.upload_file(
path_or_fileobj=index_path,
path_in_repo=f"vector_store/archive/index_{timestamp}.faiss",
repo_id=self.dataset_name,
repo_type="dataset"
)
self.api.upload_file(
path_or_fileobj=config_path,
path_in_repo=f"vector_store/archive/index_{timestamp}.pkl",
repo_id=self.dataset_name,
repo_type="dataset"
)
except Exception:
# If no files exist, create archive directory
with tempfile.NamedTemporaryFile(delete=False) as temp:
temp_path = temp.name
try:
self.api.upload_file(
path_or_fileobj=temp_path,
path_in_repo="vector_store/archive/.gitkeep",
repo_id=self.dataset_name,
repo_type="dataset"
)
finally:
if os.path.exists(temp_path):
os.remove(temp_path)
# Upload current files
self.api.upload_file(
path_or_fileobj=index_path,
path_in_repo="vector_store/index.faiss",
repo_id=self.dataset_name,
repo_type="dataset"
)
self.api.upload_file(
path_or_fileobj=config_path,
path_in_repo="vector_store/index.pkl",
repo_id=self.dataset_name,
repo_type="dataset"
)
# Update metadata about last update
metadata = {
"last_update": timestamp,
"version": "1.0"
}
with tempfile.NamedTemporaryFile(mode="w+", suffix=".json", delete=False) as temp:
json.dump(metadata, temp, ensure_ascii=False, indent=2)
temp_name = temp.name
try:
self.api.upload_file(
path_or_fileobj=temp_name,
path_in_repo="vector_store/metadata.json",
repo_id=self.dataset_name,
repo_type="dataset"
)
finally:
if os.path.exists(temp_name):
os.remove(temp_name)
return True, "Vector store uploaded successfully"
except Exception as e:
return False, f"Error uploading vector store: {str(e)}"
def download_vector_store(self) -> Tuple[bool, Union[FAISS, str]]:
"""Download vector store from dataset"""
try:
with tempfile.TemporaryDirectory() as temp_dir:
print(f"Downloading to temporary directory: {temp_dir}")
# Download files to temporary directory
try:
index_path = self.api.hf_hub_download(
repo_id=self.dataset_name,
filename="vector_store/index.faiss",
repo_type="dataset",
local_dir=temp_dir
)
print(f"Downloaded index.faiss to: {index_path}")
config_path = self.api.hf_hub_download(
repo_id=self.dataset_name,
filename="vector_store/index.pkl",
repo_type="dataset",
local_dir=temp_dir
)
print(f"Downloaded index.pkl to: {config_path}")
# Verify files exist
if not os.path.exists(index_path) or not os.path.exists(config_path):
return False, f"Downloaded files not found at {temp_dir}"
# Load vector store from temporary directory
embeddings = HuggingFaceEmbeddings(
model_name=EMBEDDING_MODEL,
model_kwargs={'device': 'cpu'}
)
# Use the directory containing the files
store_dir = os.path.dirname(index_path)
print(f"Loading vector store from: {store_dir}")
vector_store = FAISS.load_local(
store_dir,
embeddings,
allow_dangerous_deserialization=True
)
return True, vector_store
except Exception as e:
return False, f"Failed to download vector store: {str(e)}"
except Exception as e:
return False, f"Error downloading vector store: {str(e)}"
def save_chat_history(self, conversation_id: str, messages: List[Dict[str, str]]) -> Tuple[bool, str]:
try:
timestamp = datetime.now().isoformat()
filename = f"{self.chat_history_path}/{conversation_id}_{datetime.now().strftime('%Y%m%d-%H%M%S')}.json"
chat_data = {
"conversation_id": conversation_id,
"timestamp": timestamp,
"history": messages # Changed from 'messages' to 'history'
}
if not self._validate_chat_structure(chat_data):
return False, "Invalid chat history structure"
with tempfile.NamedTemporaryFile(mode="w+", suffix=".json", delete=False, encoding="utf-8") as temp:
json.dump(chat_data, temp, ensure_ascii=False, indent=2)
temp.flush()
return True, "Chat history saved successfully"
except Exception as e:
return False, f"Error saving chat history: {str(e)}"
def _validate_chat_structure(self, chat_data: Dict) -> bool:
required_fields = {"conversation_id", "timestamp", "history"}
if not all(field in chat_data for field in required_fields):
return False
if not isinstance(chat_data["history"], list):
return False
for message in chat_data["history"]:
if not all(field in message for field in ["role", "content", "timestamp"]):
return False
return True
def get_chat_history(self, conversation_id: Optional[str] = None) -> Tuple[bool, Any]:
try:
logger.info(f"Attempting to get chat history from dataset {self.dataset_name}")
# Get all files from repository
files = self.api.list_repo_files(
repo_id=self.dataset_name,
repo_type="dataset"
)
# Filter only files from chat_history directory using settings
chat_files = [f for f in files if f.startswith(f"{CHAT_HISTORY_PATH}/")]
logger.info(f"Found {len(chat_files)} files in {CHAT_HISTORY_PATH}")
if conversation_id:
chat_files = [f for f in chat_files if conversation_id in f]
if not chat_files:
logger.warning("No chat history files found")
return True, []
chat_histories = []
with tempfile.TemporaryDirectory() as temp_dir:
for file in chat_files:
if file.endswith(".gitkeep"):
continue
try:
local_file = self.api.hf_hub_download(
repo_id=self.dataset_name,
filename=file,
repo_type="dataset",
local_dir=temp_dir
)
with open(local_file, "r", encoding="utf-8") as f:
chat_data = json.load(f)
logger.debug(f"Loaded chat data: {chat_data}") # Debug log
if not isinstance(chat_data, dict):
logger.error(f"Chat data is not a dictionary in {file}")
continue
# Get messages from either 'messages' or 'history' key
messages = None
if "messages" in chat_data:
messages = chat_data["messages"]
elif "history" in chat_data:
messages = chat_data["history"]
if not messages:
logger.error(f"No messages found in {file}")
continue
if not isinstance(messages, list):
logger.error(f"Messages is not a list in {file}")
continue
# Create standardized format
standardized_data = {
"conversation_id": chat_data.get("conversation_id", "unknown"),
"timestamp": chat_data.get("timestamp", datetime.now().isoformat()),
"messages": messages
}
chat_histories.append(standardized_data)
logger.info(f"Successfully loaded chat data from {file}")
except json.JSONDecodeError as e:
logger.error(f"Invalid JSON in file {file}: {str(e)}")
continue
except Exception as e:
logger.error(f"Error processing file {file}: {e}")
continue
if not chat_histories:
logger.warning("No valid chat histories found")
else:
logger.info(f"Successfully loaded {len(chat_histories)} chat histories")
return True, chat_histories
except Exception as e:
logger.error(f"Error getting chat history: {str(e)}")
return False, str(e)
def upload_document(self, file_path: str, document_id: Optional[str] = None) -> Tuple[bool, str]:
"""
Upload document to the dataset
Args:
file_path: Path to the document file
document_id: Document identifier (if None, uses filename)
Returns:
(success, message)
"""
try:
if not os.path.exists(file_path):
return False, f"File not found: {file_path}"
# Use filename as document_id if not specified
if document_id is None:
document_id = os.path.basename(file_path)
# Add timestamp to filename
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
filename = f"documents/{document_id}_{timestamp}{os.path.splitext(file_path)[1]}"
# Upload file
self.api.upload_file(
path_or_fileobj=file_path,
path_in_repo=filename,
repo_id=self.dataset_name,
repo_type="dataset"
)
return True, f"Document uploaded successfully: {filename}"
except Exception as e:
return False, f"Error uploading document: {str(e)}"
def test_dataset_connection(token: Optional[str] = None) -> Tuple[bool, str]:
"""
Test function to check dataset connection
Args:
token: Hugging Face Hub access token
Returns:
(success, message)
"""
try:
manager = DatasetManager(token=token)
success, message = manager.init_dataset_structure()
if not success:
return False, message
print(f"Initialization test: {message}")
return True, "Dataset connection is working"
except Exception as e:
return False, f"Dataset connection error: {str(e)}"
if __name__ == "__main__":
# Test connection
success, message = test_dataset_connection()
print(message)