Spaces:
Runtime error
Runtime error
Migrate to llama-index
Browse files- app/main.py +13 -11
- app/rag.py +81 -51
- requirements.txt +4 -5
- start_service.sh +3 -0
app/main.py
CHANGED
|
@@ -38,17 +38,19 @@ def upload(files: list[UploadFile]):
|
|
| 38 |
session_assistant.clear()
|
| 39 |
session_messages = []
|
| 40 |
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
|
| 51 |
-
|
|
|
|
|
|
|
| 52 |
|
| 53 |
return "Files inserted!"
|
| 54 |
|
|
|
|
| 38 |
session_assistant.clear()
|
| 39 |
session_messages = []
|
| 40 |
|
| 41 |
+
try:
|
| 42 |
+
for file in files:
|
| 43 |
+
path = f"files/{file.filename}"
|
| 44 |
+
try:
|
| 45 |
+
suffix = Path(file.filename).suffix
|
| 46 |
+
with NamedTemporaryFile(delete=False, suffix=suffix) as tmp:
|
| 47 |
+
shutil.copyfileobj(file.file, tmp)
|
| 48 |
+
|
| 49 |
+
finally:
|
| 50 |
+
file.file.close()
|
| 51 |
+
finally:
|
| 52 |
+
session_assistant.ingest("files/")
|
| 53 |
+
os.remove("files/")
|
| 54 |
|
| 55 |
return "Files inserted!"
|
| 56 |
|
app/rag.py
CHANGED
|
@@ -1,67 +1,97 @@
|
|
| 1 |
-
from
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
from
|
| 8 |
-
from
|
| 9 |
-
from
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
|
| 11 |
|
| 12 |
class ChatPDF:
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
|
| 17 |
def __init__(self):
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
keep_alive=-1,
|
| 21 |
-
temperature=0,
|
| 22 |
-
num_predict=512,
|
| 23 |
-
repeat_penalty=1.3,
|
| 24 |
-
repeat_last_n=-1
|
| 25 |
-
)
|
| 26 |
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
)
|
|
|
|
| 38 |
|
| 39 |
-
def ingest(self,
|
| 40 |
-
docs =
|
| 41 |
-
chunks = self.text_splitter.split_documents(docs)
|
| 42 |
-
chunks = filter_complex_metadata(chunks)
|
| 43 |
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 51 |
)
|
| 52 |
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 57 |
|
| 58 |
def ask(self, query: str):
|
| 59 |
-
if not self.
|
| 60 |
return "Please, add a PDF document first."
|
| 61 |
|
| 62 |
-
|
|
|
|
|
|
|
|
|
|
| 63 |
|
| 64 |
def clear(self):
|
| 65 |
-
self.
|
| 66 |
-
self.
|
| 67 |
-
self.
|
|
|
|
| 1 |
+
from llama_index.core import (
|
| 2 |
+
SimpleDirectoryReader,
|
| 3 |
+
VectorStoreIndex,
|
| 4 |
+
StorageContext,
|
| 5 |
+
Settings,
|
| 6 |
+
get_response_synthesizer)
|
| 7 |
+
from llama_index.core.query_engine import RetrieverQueryEngine, TransformQueryEngine
|
| 8 |
+
from llama_index.core.node_parser import SentenceSplitter
|
| 9 |
+
from llama_index.core.schema import TextNode, MetadataMode
|
| 10 |
+
from llama_index.vector_stores.qdrant import QdrantVectorStore
|
| 11 |
+
from llama_index.embeddings.ollama import OllamaEmbedding
|
| 12 |
+
from llama_index.llms.ollama import Ollama
|
| 13 |
+
from llama_index.core.retrievers import VectorIndexRetriever
|
| 14 |
+
from llama_index.core.indices.query.query_transform import HyDEQueryTransform
|
| 15 |
+
import qdrant_client
|
| 16 |
+
import logging
|
| 17 |
|
| 18 |
|
| 19 |
class ChatPDF:
|
| 20 |
+
text_chunks = []
|
| 21 |
+
doc_ids = []
|
| 22 |
+
nodes = []
|
| 23 |
|
| 24 |
def __init__(self):
|
| 25 |
+
logging.basicConfig(level=logging.INFO)
|
| 26 |
+
logger = logging.getLogger(__name__)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 27 |
|
| 28 |
+
text_parser = SentenceSplitter(chunk_size=512, chunk_overlap=100)
|
| 29 |
+
|
| 30 |
+
logger.info("initializing the vector store related objects")
|
| 31 |
+
client = qdrant_client.QdrantClient(host="localhost", port=6333)
|
| 32 |
+
vector_store = QdrantVectorStore(client=client, collection_name="rag_documents")
|
| 33 |
+
|
| 34 |
+
logger.info("initializing the OllamaEmbedding")
|
| 35 |
+
embed_model = OllamaEmbedding(model_name='mxbai-embed-large', request_timeout=1000000)
|
| 36 |
+
logger.info("initializing the global settings")
|
| 37 |
+
Settings.embed_model = embed_model
|
| 38 |
+
Settings.llm = Ollama(model="qwen:1.8b", request_timeout=1000000)
|
| 39 |
+
Settings.transformations = [text_parser]
|
| 40 |
|
| 41 |
+
def ingest(self, dir_path: str):
|
| 42 |
+
docs = SimpleDirectoryReader(input_dir=dir_path).load_data()
|
|
|
|
|
|
|
| 43 |
|
| 44 |
+
logger.info("enumerating docs")
|
| 45 |
+
for doc_idx, doc in enumerate(docs):
|
| 46 |
+
curr_text_chunks = text_parser.split_text(doc.text)
|
| 47 |
+
text_chunks.extend(curr_text_chunks)
|
| 48 |
+
doc_ids.extend([doc_idx] * len(curr_text_chunks))
|
| 49 |
+
|
| 50 |
+
logger.info("enumerating text_chunks")
|
| 51 |
+
for idx, text_chunk in enumerate(text_chunks):
|
| 52 |
+
node = TextNode(text=text_chunk)
|
| 53 |
+
src_doc = docs[doc_ids[idx]]
|
| 54 |
+
node.metadata = src_doc.metadata
|
| 55 |
+
nodes.append(node)
|
| 56 |
+
|
| 57 |
+
logger.info("enumerating nodes")
|
| 58 |
+
for node in nodes:
|
| 59 |
+
node_embedding = embed_model.get_text_embedding(
|
| 60 |
+
node.get_content(metadata_mode=MetadataMode.ALL)
|
| 61 |
+
)
|
| 62 |
+
node.embedding = node_embedding
|
| 63 |
+
|
| 64 |
+
logger.info("initializing the storage context")
|
| 65 |
+
storage_context = StorageContext.from_defaults(vector_store=vector_store)
|
| 66 |
+
logger.info("indexing the nodes in VectorStoreIndex")
|
| 67 |
+
index = VectorStoreIndex(
|
| 68 |
+
nodes=nodes,
|
| 69 |
+
storage_context=storage_context,
|
| 70 |
+
transformations=Settings.transformations,
|
| 71 |
)
|
| 72 |
|
| 73 |
+
logger.info("initializing the VectorIndexRetriever with top_k as 5")
|
| 74 |
+
vector_retriever = VectorIndexRetriever(index=index, similarity_top_k=5)
|
| 75 |
+
response_synthesizer = get_response_synthesizer()
|
| 76 |
+
logger.info("creating the RetrieverQueryEngine instance")
|
| 77 |
+
vector_query_engine = RetrieverQueryEngine(
|
| 78 |
+
retriever=vector_retriever,
|
| 79 |
+
response_synthesizer=response_synthesizer,
|
| 80 |
+
)
|
| 81 |
+
logger.info("creating the HyDEQueryTransform instance")
|
| 82 |
+
hyde = HyDEQueryTransform(include_original=True)
|
| 83 |
+
self.hyde_query_engine = TransformQueryEngine(vector_query_engine, hyde)
|
| 84 |
|
| 85 |
def ask(self, query: str):
|
| 86 |
+
if not self.hyde_query_engine:
|
| 87 |
return "Please, add a PDF document first."
|
| 88 |
|
| 89 |
+
logger.info("retrieving the response to the query")
|
| 90 |
+
response = self.hyde_query_engine.query(str_or_query_bundle=query)
|
| 91 |
+
print(response)
|
| 92 |
+
return response
|
| 93 |
|
| 94 |
def clear(self):
|
| 95 |
+
self.text_chunks = []
|
| 96 |
+
self.doc_ids = []
|
| 97 |
+
self.nodes = []
|
requirements.txt
CHANGED
|
@@ -1,6 +1,5 @@
|
|
| 1 |
fastapi
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
chromadb
|
|
|
|
| 1 |
fastapi
|
| 2 |
+
llama-index
|
| 3 |
+
llama-index-vector-stores-qdrant
|
| 4 |
+
llama-index-embeddings-ollama
|
| 5 |
+
llama-index-llms-ollama
|
|
|
start_service.sh
CHANGED
|
@@ -6,6 +6,9 @@ ollama serve &
|
|
| 6 |
# Wait for Ollama to start
|
| 7 |
sleep 5
|
| 8 |
|
|
|
|
|
|
|
|
|
|
| 9 |
# Pull and run <YOUR_MODEL_NAME>
|
| 10 |
ollama pull qwen:1.8b
|
| 11 |
|
|
|
|
| 6 |
# Wait for Ollama to start
|
| 7 |
sleep 5
|
| 8 |
|
| 9 |
+
#
|
| 10 |
+
ollama pull mxbai-embed-large
|
| 11 |
+
|
| 12 |
# Pull and run <YOUR_MODEL_NAME>
|
| 13 |
ollama pull qwen:1.8b
|
| 14 |
|