Help with Raptor implementation #15472
Unanswered
rahulkrprajapati
asked this question in
Q&A
Replies: 7 comments 6 replies
-
|
To modify your class to use Updated Constructor Codefrom llama_index.packs.raptor import RaptorPack
from llama_index.packs.raptor.base import SummaryModule
from llama_index.core import Document, MockEmbedding
from llama_index.core.llms import OpenAI
from llama_index.storage.docstore.redis import RedisDocumentStore
from llama_index.storage.vectorstore.redis import RedisVectorStore
from llama_index.storage.context import StorageContext
from llama_index.packs.raptor.base import RaptorRetriever
class YourClass:
def __init__(self, config, namespace, similarity_top_k, recency, similarity_top_k_reranker, verbose=False):
self.config = config
self.namespace = namespace
self.similarity_top_k = similarity_top_k
self.redis_client = redis.from_url(self.config.get("REDIS_VECTOR_STORE"))
self.docstore = RedisDocumentStore.from_redis_client(
redis_client=self.redis_client, namespace=self.namespace
)
self.vector_store = RedisVectorStore(
index_name=self.namespace,
redis_url=self.config.get("REDIS_VECTOR_STORE"),
metadata_fields=["namespace"],
)
self.storage_context = StorageContext.from_defaults(
docstore=self.docstore,
index_store=RedisIndexStore.from_redis_client(
redis_client=self.redis_client, namespace=self.namespace
),
vector_store=self.vector_store,
)
self.llm = OpenAI(
model=self.config.get("LLM_ANSWERING_MODEL"),
api_key=self.config.get("OPENAI_API_KEY"),
)
summary_prompt = "As a professional summarizer, create a concise and comprehensive summary of the provided text, be it an article, post, conversation, or passage with as much detail as possible."
self.summary_module = SummaryModule(
llm=self.llm, summary_prompt=summary_prompt, num_workers=16
)
self.raptor_pack = RaptorPack(
documents=[],
embed_model=MockEmbedding(embed_dim=1536),
llm=self.llm,
vector_store=self.vector_store,
similarity_top_k=self.similarity_top_k,
mode="collapsed",
summary_module=self.summary_module
)
try:
self.base_index = load_index_from_storage(self.storage_context)
print("[INFO] Index found at storage")
except ValueError as e:
print("[INFO] No index found at storage")
self.base_index = VectorStoreIndex(
nodes=[],
storage_context=self.storage_context,
store_nodes_override=True,
)
self.base_retriever = self.base_index.as_retriever(
similarity_top_k=self.similarity_top_k,
filters=MetadataFilters(
filters=[ExactMatchFilter(key="namespace", value=self.namespace)]
),
)
self.retriever = AutoMergingRetriever(
self.base_retriever, self.storage_context, verbose=verbose
)
self.recency_postprocessor = FixedRecencyPostprocessor(
top_k=recency, date_key="creation_date"
)
self.postprocessor = SentenceTransformerRerank(
model=self.config.get("RAG_RERANK_MODEL"),
top_n=similarity_top_k_reranker,
)
self.response_synthesizer = get_response_synthesizer(
response_mode=ResponseMode.COMPACT, llm=self.llm
)Updated Helper Functions def create_document(self, text, filename):
documents = []
for idx, page in text.items():
document = Document(text=page)
current_date = datetime.now().strftime("%Y-%m-%d")
document.metadata = {
"filename": filename,
"page_number": idx,
"creation_date": current_date,
"last_accessed_date": current_date,
"last_modified_date": current_date,
"namespace": self.namespace,
}
documents.append(document)
return documents
def return_nodes(self, docs):
nodes = self.node_parser.get_nodes_from_documents(docs)
return nodes
def add_nodes_to_doc_store(self, all_nodes):
try:
self.docstore.add_documents(nodes=all_nodes)
except Exception as e:
print(f"An error occurred when adding documents to doc store: {e}")
traceback.print_exc()
raise Exception(
f"An error occurred when adding documents to doc store: {e}"
)
def process_streamlit_fetch_query_results(self, query="", verbose=False):
try:
query_bundle = QueryBundle(query_str=query)
nodes = self.raptor_pack.run(query, mode="collapsed")
recency_nodes = self.recency_postprocessor.postprocess_nodes(
nodes, query_bundle=query_bundle
)
rerank_nodes = self.postprocessor.postprocess_nodes(
nodes=recency_nodes, query_bundle=query_bundle
)
return rerank_nodes
except Exception as e:
traceback.print_exc()
raise Exception(f"An error occurred retrieving: {e}")Explanation
This setup should allow you to use |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
@dosu use QueryFusionRetriever to merger raptor + bm25 in the generated code |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
To merge Updated Constructor Codefrom llama_index.packs.raptor import RaptorPack
from llama_index.packs.raptor.base import SummaryModule
from llama_index.core import Document, MockEmbedding
from llama_index.core.llms import OpenAI
from llama_index.storage.docstore.redis import RedisDocumentStore
from llama_index.storage.vectorstore.redis import RedisVectorStore
from llama_index.storage.context import StorageContext
from llama_index.retrievers.bm25 import BM25Retriever
from llama_index.core.retrievers import QueryFusionRetriever
class YourClass:
def __init__(self, config, namespace, similarity_top_k, recency, similarity_top_k_reranker, verbose=False):
self.config = config
self.namespace = namespace
self.similarity_top_k = similarity_top_k
self.redis_client = redis.from_url(self.config.get("REDIS_VECTOR_STORE"))
self.docstore = RedisDocumentStore.from_redis_client(
redis_client=self.redis_client, namespace=self.namespace
)
self.vector_store = RedisVectorStore(
index_name=self.namespace,
redis_url=self.config.get("REDIS_VECTOR_STORE"),
metadata_fields=["namespace"],
)
self.storage_context = StorageContext.from_defaults(
docstore=self.docstore,
index_store=RedisIndexStore.from_redis_client(
redis_client=self.redis_client, namespace=self.namespace
),
vector_store=self.vector_store,
)
self.llm = OpenAI(
model=self.config.get("LLM_ANSWERING_MODEL"),
api_key=self.config.get("OPENAI_API_KEY"),
)
summary_prompt = "As a professional summarizer, create a concise and comprehensive summary of the provided text, be it an article, post, conversation, or passage with as much detail as possible."
self.summary_module = SummaryModule(
llm=self.llm, summary_prompt=summary_prompt, num_workers=16
)
self.raptor_pack = RaptorPack(
documents=[],
embed_model=MockEmbedding(embed_dim=1536),
llm=self.llm,
vector_store=self.vector_store,
similarity_top_k=self.similarity_top_k,
mode="collapsed",
summary_module=self.summary_module
)
self.bm25_retriever = BM25Retriever.from_defaults(
docstore=self.docstore, similarity_top_k=self.similarity_top_k
)
self.fusion_retriever = QueryFusionRetriever(
[self.raptor_pack.fusion_retriever, self.bm25_retriever],
similarity_top_k=self.similarity_top_k,
num_queries=1, # set this to 1 to disable query generation
mode="reciprocal_rerank",
use_async=True,
verbose=verbose,
)
try:
self.base_index = load_index_from_storage(self.storage_context)
print("[INFO] Index found at storage")
except ValueError as e:
print("[INFO] No index found at storage")
self.base_index = VectorStoreIndex(
nodes=[],
storage_context=self.storage_context,
store_nodes_override=True,
)
self.base_retriever = self.base_index.as_retriever(
similarity_top_k=self.similarity_top_k,
filters=MetadataFilters(
filters=[ExactMatchFilter(key="namespace", value=self.namespace)]
),
)
self.retriever = AutoMergingRetriever(
self.base_retriever, self.storage_context, verbose=verbose
)
self.recency_postprocessor = FixedRecencyPostprocessor(
top_k=recency, date_key="creation_date"
)
self.postprocessor = SentenceTransformerRerank(
model=self.config.get("RAG_RERANK_MODEL"),
top_n=similarity_top_k_reranker,
)
self.response_synthesizer = get_response_synthesizer(
response_mode=ResponseMode.COMPACT, llm=self.llm
)Updated Helper Functions def create_document(self, text, filename):
documents = []
for idx, page in text.items():
document = Document(text=page)
current_date = datetime.now().strftime("%Y-%m-%d")
document.metadata = {
"filename": filename,
"page_number": idx,
"creation_date": current_date,
"last_accessed_date": current_date,
"last_modified_date": current_date,
"namespace": self.namespace,
}
documents.append(document)
return documents
def return_nodes(self, docs):
nodes = self.node_parser.get_nodes_from_documents(docs)
return nodes
def add_nodes_to_doc_store(self, all_nodes):
try:
self.docstore.add_documents(nodes=all_nodes)
except Exception as e:
print(f"An error occurred when adding documents to doc store: {e}")
traceback.print_exc()
raise Exception(
f"An error occurred when adding documents to doc store: {e}"
)
def process_streamlit_fetch_query_results(self, query="", verbose=False):
try:
query_bundle = QueryBundle(query_str=query)
retrived_nodes = self.fusion_retriever.retrieve(query_bundle)
recency_nodes = self.recency_postprocessor.postprocess_nodes(
retrived_nodes, query_bundle=query_bundle
)
rerank_nodes = self.postprocessor.postprocess_nodes(
nodes=recency_nodes, query_bundle=query_bundle
)
return rerank_nodes
except Exception as e:
traceback.print_exc()
raise Exception(f"An error occurred retrieving: {e}")Explanation
This setup should allow you to use |
Beta Was this translation helpful? Give feedback.
-
|
@dosu you forgot the HierarchicalNodeParser should it be used for adding the nodes? |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
Yes, the Updated Constructor Codefrom llama_index.packs.raptor import RaptorPack
from llama_index.packs.raptor.base import SummaryModule
from llama_index.core import Document, MockEmbedding
from llama_index.core.llms import OpenAI
from llama_index.storage.docstore.redis import RedisDocumentStore
from llama_index.storage.vectorstore.redis import RedisVectorStore
from llama_index.storage.context import StorageContext
from llama_index.retrievers.bm25 import BM25Retriever
from llama_index.core.retrievers import QueryFusionRetriever
from llama_index.core.node_parser.relational.hierarchical import HierarchicalNodeParser
from llama_index.core.node_parser.text.sentence import SentenceSplitter
class YourClass:
def __init__(self, config, namespace, similarity_top_k, recency, similarity_top_k_reranker, verbose=False):
self.config = config
self.namespace = namespace
self.similarity_top_k = similarity_top_k
self.redis_client = redis.from_url(self.config.get("REDIS_VECTOR_STORE"))
self.docstore = RedisDocumentStore.from_redis_client(
redis_client=self.redis_client, namespace=self.namespace
)
self.vector_store = RedisVectorStore(
index_name=self.namespace,
redis_url=self.config.get("REDIS_VECTOR_STORE"),
metadata_fields=["namespace"],
)
self.storage_context = StorageContext.from_defaults(
docstore=self.docstore,
index_store=RedisIndexStore.from_redis_client(
redis_client=self.redis_client, namespace=self.namespace
),
vector_store=self.vector_store,
)
self.llm = OpenAI(
model=self.config.get("LLM_ANSWERING_MODEL"),
api_key=self.config.get("OPENAI_API_KEY"),
)
summary_prompt = "As a professional summarizer, create a concise and comprehensive summary of the provided text, be it an article, post, conversation, or passage with as much detail as possible."
self.summary_module = SummaryModule(
llm=self.llm, summary_prompt=summary_prompt, num_workers=16
)
self.raptor_pack = RaptorPack(
documents=[],
embed_model=MockEmbedding(embed_dim=1536),
llm=self.llm,
vector_store=self.vector_store,
similarity_top_k=self.similarity_top_k,
mode="collapsed",
summary_module=self.summary_module
)
self.bm25_retriever = BM25Retriever.from_defaults(
docstore=self.docstore, similarity_top_k=self.similarity_top_k
)
self.fusion_retriever = QueryFusionRetriever(
[self.raptor_pack.fusion_retriever, self.bm25_retriever],
similarity_top_k=self.similarity_top_k,
num_queries=1, # set this to 1 to disable query generation
mode="reciprocal_rerank",
use_async=True,
verbose=verbose,
)
self.node_parser = HierarchicalNodeParser.from_defaults(
chunk_sizes=[2048, 512, 128],
node_parser_map={
"chunk_size_2048": SentenceSplitter(chunk_size=2048),
"chunk_size_512": SentenceSplitter(chunk_size=512),
"chunk_size_128": SentenceSplitter(chunk_size=128),
}
)
try:
self.base_index = load_index_from_storage(self.storage_context)
print("[INFO] Index found at storage")
except ValueError as e:
print("[INFO] No index found at storage")
self.base_index = VectorStoreIndex(
nodes=[],
storage_context=self.storage_context,
store_nodes_override=True,
)
self.base_retriever = self.base_index.as_retriever(
similarity_top_k=self.similarity_top_k,
filters=MetadataFilters(
filters=[ExactMatchFilter(key="namespace", value=self.namespace)]
),
)
self.retriever = AutoMergingRetriever(
self.base_retriever, self.storage_context, verbose=verbose
)
self.recency_postprocessor = FixedRecencyPostprocessor(
top_k=recency, date_key="creation_date"
)
self.postprocessor = SentenceTransformerRerank(
model=self.config.get("RAG_RERANK_MODEL"),
top_n=similarity_top_k_reranker,
)
self.response_synthesizer = get_response_synthesizer(
response_mode=ResponseMode.COMPACT, llm=self.llm
)Updated Helper Functions def create_document(self, text, filename):
documents = []
for idx, page in text.items():
document = Document(text=page)
current_date = datetime.now().strftime("%Y-%m-%d")
document.metadata = {
"filename": filename,
"page_number": idx,
"creation_date": current_date,
"last_accessed_date": current_date,
"last_modified_date": current_date,
"namespace": self.namespace,
}
documents.append(document)
return documents
def return_nodes(self, docs):
nodes = self.node_parser.get_nodes_from_documents(docs)
return nodes
def add_nodes_to_doc_store(self, all_nodes):
try:
self.docstore.add_documents(nodes=all_nodes)
except Exception as e:
print(f"An error occurred when adding documents to doc store: {e}")
traceback.print_exc()
raise Exception(
f"An error occurred when adding documents to doc store: {e}"
)
def process_streamlit_fetch_query_results(self, query="", verbose=False):
try:
query_bundle = QueryBundle(query_str=query)
retrived_nodes = self.fusion_retriever.retrieve(query_bundle)
recency_nodes = self.recency_postprocessor.postprocess_nodes(
retrived_nodes, query_bundle=query_bundle
)
rerank_nodes = self.postprocessor.postprocess_nodes(
nodes=recency_nodes, query_bundle=query_bundle
)
return rerank_nodes
except Exception as e:
traceback.print_exc()
raise Exception(f"An error occurred retrieving: {e}")Explanation
This setup ensures that the |
Beta Was this translation helpful? Give feedback.
-
|
@dosu raise Exception(f"An error occurred retrieving: {e}") |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
To resolve the error "No docs found on index 'test' with prefix 'llama_index/vector' and filters '*'", you need to ensure that the index on the field you are filtering is correctly set up. Specifically, you need to add an additional member to the "fields" list of the vector index, like so: { "type": "filter", "path": "text" }This setup is necessary for filtering within filters = MetadataFilters(
filters=[
MetadataFilter(
key="text",
value="How do we best augment LLMs with our own private data?",
operator=FilterOperator.NE,
)
]
)
query = VectorStoreQuery(
query_str="What are LLMs useful for?",
query_embedding=query_embedding,
similarity_top_k=2,
filters=filters,
)
responses_with_filter = vector_store.query(query=query)Ensure that the index on the field "text" is properly configured in your MongoDB collection to support this filter [1]. |
Beta Was this translation helpful? Give feedback.
-
|
@dosu in the code below define my metadata filter properly: self.raptor_pack = RaptorPack(
documents=[],
embed_model=Settings.embed_model,
llm=self.llm,
vector_store=self.vector_store,
similarity_top_k=self.similarity_top_k,
mode="collapsed",
summary_module=self.summary_module,
)
self.bm25_retriever = BM25Retriever.from_defaults(
docstore=self.docstore, similarity_top_k=self.similarity_top_k
)
self.fusion_retriever = QueryFusionRetriever(
[self.raptor_pack.retriever, self.bm25_retriever],
similarity_top_k=self.similarity_top_k,
num_queries=1, # set this to 1 to disable query generation
mode="reciprocal_rerank",
use_async=True,
verbose=verbose,
)
self.node_parser = HierarchicalNodeParser.from_defaults()
try:
self.base_index = load_index_from_storage(self.storage_context)
print("[INFO] Index found at storage")
except ValueError as e:
print("[INFO] No index found at storage")
self.base_index = VectorStoreIndex(
nodes=[],
storage_context=self.storage_context,
store_nodes_override=True,
)
self.base_retriever = self.base_index.as_retriever(
similarity_top_k=self.similarity_top_k,
filters=MetadataFilters(
filters=[
ExactMatchFilter(key="namespace", value=self.namespace)
]
),
)
self.retriever = AutoMergingRetriever(
self.base_retriever, self.storage_context, verbose=verbose
)
self.recency_postprocessor = FixedRecencyPostprocessor(
top_k=recency, date_key="creation_date"
)
self.postprocessor = SentenceTransformerRerank(
model=self.config.get("RAG_RERANK_MODEL"),
top_n=similarity_top_k_reranker,
)
self.response_synthesizer = get_response_synthesizer(
response_mode=ResponseMode.COMPACT, llm=self.llm
)the code fragment here does not take into account my metadata filter: self.fusion_retriever = QueryFusionRetriever(
[self.raptor_pack.retriever, self.bm25_retriever],
similarity_top_k=self.similarity_top_k,
num_queries=1, # set this to 1 to disable query generation
mode="reciprocal_rerank",
use_async=True,
verbose=verbose,
)the filter should be something like: filters=MetadataFilters(
filters=[
ExactMatchFilter(key="namespace", value=self.namespace)
]
),this is being used in the function as: def process_streamlit_fetch_query_results(self, query="", verbose=False):
try:
query_bundle = QueryBundle(query_str=query)
retrived_nodes = self.fusion_retriever.retrieve(query_bundle)
recency_nodes = self.recency_postprocessor.postprocess_nodes(
retrived_nodes, query_bundle=query_bundle
)
rerank_nodes = self.postprocessor.postprocess_nodes(
nodes=recency_nodes, query_bundle=query_bundle
)
px.active_session().url
return rerank_nodes
except Exception as e:
traceback.print_exc()
raise Exception(f"An error occurred retrieving: {e}") |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
To properly define and integrate a metadata filter for the Updated Constructor Codefrom llama_index.packs.raptor import RaptorPack
from llama_index.packs.raptor.base import SummaryModule
from llama_index.core import Document, MockEmbedding
from llama_index.core.llms import OpenAI
from llama_index.storage.docstore.redis import RedisDocumentStore
from llama_index.storage.vectorstore.redis import RedisVectorStore
from llama_index.storage.context import StorageContext
from llama_index.retrievers.bm25 import BM25Retriever
from llama_index.core.retrievers import QueryFusionRetriever
from llama_index.core.node_parser.relational.hierarchical import HierarchicalNodeParser
from llama_index.core.vector_stores import MetadataFilter, MetadataFilters, FilterOperator
class YourClass:
def __init__(self, config, namespace, similarity_top_k, recency, similarity_top_k_reranker, verbose=False):
self.config = config
self.namespace = namespace
self.similarity_top_k = similarity_top_k
self.redis_client = redis.from_url(self.config.get("REDIS_VECTOR_STORE"))
self.docstore = RedisDocumentStore.from_redis_client(
redis_client=self.redis_client, namespace=self.namespace
)
self.vector_store = RedisVectorStore(
index_name=self.namespace,
redis_url=self.config.get("REDIS_VECTOR_STORE"),
metadata_fields=["namespace"],
)
self.storage_context = StorageContext.from_defaults(
docstore=self.docstore,
index_store=RedisIndexStore.from_redis_client(
redis_client=self.redis_client, namespace=self.namespace
),
vector_store=self.vector_store,
)
self.llm = OpenAI(
model=self.config.get("LLM_ANSWERING_MODEL"),
api_key=self.config.get("OPENAI_API_KEY"),
)
summary_prompt = "As a professional summarizer, create a concise and comprehensive summary of the provided text, be it an article, post, conversation, or passage with as much detail as possible."
self.summary_module = SummaryModule(
llm=self.llm, summary_prompt=summary_prompt, num_workers=16
)
self.raptor_pack = RaptorPack(
documents=[],
embed_model=MockEmbedding(embed_dim=1536),
llm=self.llm,
vector_store=self.vector_store,
similarity_top_k=self.similarity_top_k,
mode="collapsed",
summary_module=self.summary_module
)
self.bm25_retriever = BM25Retriever.from_defaults(
docstore=self.docstore, similarity_top_k=self.similarity_top_k
)
filters = MetadataFilters(
filters=[
MetadataFilter(key="namespace", operator=FilterOperator.EQ, value=self.namespace)
]
)
self.fusion_retriever = QueryFusionRetriever(
[self.raptor_pack.retriever, self.bm25_retriever],
similarity_top_k=self.similarity_top_k,
num_queries=1, # set this to 1 to disable query generation
mode="reciprocal_rerank",
use_async=True,
verbose=verbose,
filters=filters
)
self.node_parser = HierarchicalNodeParser.from_defaults()
try:
self.base_index = load_index_from_storage(self.storage_context)
print("[INFO] Index found at storage")
except ValueError as e:
print("[INFO] No index found at storage")
self.base_index = VectorStoreIndex(
nodes=[],
storage_context=self.storage_context,
store_nodes_override=True,
)
self.base_retriever = self.base_index.as_retriever(
similarity_top_k=self.similarity_top_k,
filters=filters
)
self.retriever = AutoMergingRetriever(
self.base_retriever, self.storage_context, verbose=verbose
)
self.recency_postprocessor = FixedRecencyPostprocessor(
top_k=recency, date_key="creation_date"
)
self.postprocessor = SentenceTransformerRerank(
model=self.config.get("RAG_RERANK_MODEL"),
top_n=similarity_top_k_reranker,
)
self.response_synthesizer = get_response_synthesizer(
response_mode=ResponseMode.COMPACT, llm=self.llm
)Updated Helper Functions def create_document(self, text, filename):
documents = []
for idx, page in text.items():
document = Document(text=page)
current_date = datetime.now().strftime("%Y-%m-%d")
document.metadata = {
"filename": filename,
"page_number": idx,
"creation_date": current_date,
"last_accessed_date": current_date,
"last_modified_date": current_date,
"namespace": self.namespace,
}
documents.append(document)
return documents
def return_nodes(self, docs):
nodes = self.node_parser.get_nodes_from_documents(docs)
return nodes
def add_nodes_to_doc_store(self, all_nodes):
try:
self.docstore.add_documents(nodes=all_nodes)
except Exception as e:
print(f"An error occurred when adding documents to doc store: {e}")
traceback.print_exc()
raise Exception(
f"An error occurred when adding documents to doc store: {e}"
)
def process_streamlit_fetch_query_results(self, query="", verbose=False):
try:
query_bundle = QueryBundle(query_str=query)
retrived_nodes = self.fusion_retriever.retrieve(query_bundle)
recency_nodes = self.recency_postprocessor.postprocess_nodes(
retrived_nodes, query_bundle=query_bundle
)
rerank_nodes = self.postprocessor.postprocess_nodes(
nodes=recency_nodes, query_bundle=query_bundle
)
return rerank_nodes
except Exception as e:
traceback.print_exc()
raise Exception(f"An error occurred retrieving: {e}")Explanation
This setup ensures that the |
Beta Was this translation helpful? Give feedback.
-
|
@dosu TypeError: QueryFusionRetriever.init() got an unexpected keyword argument 'filters |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
To properly integrate metadata filters with Updated Constructor Codefrom llama_index.packs.raptor import RaptorPack
from llama_index.packs.raptor.base import SummaryModule
from llama_index.core import Document, MockEmbedding
from llama_index.core.llms import OpenAI
from llama_index.storage.docstore.redis import RedisDocumentStore
from llama_index.storage.vectorstore.redis import RedisVectorStore
from llama_index.storage.context import StorageContext
from llama_index.retrievers.bm25 import BM25Retriever
from llama_index.core.retrievers import QueryFusionRetriever
from llama_index.core.node_parser.relational.hierarchical import HierarchicalNodeParser
from llama_index.core.vector_stores import MetadataFilter, MetadataFilters, FilterOperator
class YourClass:
def __init__(self, config, namespace, similarity_top_k, recency, similarity_top_k_reranker, verbose=False):
self.config = config
self.namespace = namespace
self.similarity_top_k = similarity_top_k
self.redis_client = redis.from_url(self.config.get("REDIS_VECTOR_STORE"))
self.docstore = RedisDocumentStore.from_redis_client(
redis_client=self.redis_client, namespace=self.namespace
)
self.vector_store = RedisVectorStore(
index_name=self.namespace,
redis_url=self.config.get("REDIS_VECTOR_STORE"),
metadata_fields=["namespace"],
)
self.storage_context = StorageContext.from_defaults(
docstore=self.docstore,
index_store=RedisIndexStore.from_redis_client(
redis_client=self.redis_client, namespace=self.namespace
),
vector_store=self.vector_store,
)
self.llm = OpenAI(
model=self.config.get("LLM_ANSWERING_MODEL"),
api_key=self.config.get("OPENAI_API_KEY"),
)
summary_prompt = "As a professional summarizer, create a concise and comprehensive summary of the provided text, be it an article, post, conversation, or passage with as much detail as possible."
self.summary_module = SummaryModule(
llm=self.llm, summary_prompt=summary_prompt, num_workers=16
)
self.raptor_pack = RaptorPack(
documents=[],
embed_model=MockEmbedding(embed_dim=1536),
llm=self.llm,
vector_store=self.vector_store,
similarity_top_k=self.similarity_top_k,
mode="collapsed",
summary_module=self.summary_module
)
self.bm25_retriever = BM25Retriever.from_defaults(
docstore=self.docstore, similarity_top_k=self.similarity_top_k
)
filters = MetadataFilters(
filters=[
MetadataFilter(key="namespace", operator=FilterOperator.EQ, value=self.namespace)
]
)
self.raptor_retriever = self.raptor_pack.retriever
self.bm25_retriever_with_filters = self.bm25_retriever.as_retriever(filters=filters)
self.fusion_retriever = QueryFusionRetriever(
[self.raptor_retriever, self.bm25_retriever_with_filters],
similarity_top_k=self.similarity_top_k,
num_queries=1, # set this to 1 to disable query generation
mode="reciprocal_rerank",
use_async=True,
verbose=verbose
)
self.node_parser = HierarchicalNodeParser.from_defaults()
try:
self.base_index = load_index_from_storage(self.storage_context)
print("[INFO] Index found at storage")
except ValueError as e:
print("[INFO] No index found at storage")
self.base_index = VectorStoreIndex(
nodes=[],
storage_context=self.storage_context,
store_nodes_override=True,
)
self.base_retriever = self.base_index.as_retriever(
similarity_top_k=self.similarity_top_k,
filters=filters
)
self.retriever = AutoMergingRetriever(
self.base_retriever, self.storage_context, verbose=verbose
)
self.recency_postprocessor = FixedRecencyPostprocessor(
top_k=recency, date_key="creation_date"
)
self.postprocessor = SentenceTransformerRerank(
model=self.config.get("RAG_RERANK_MODEL"),
top_n=similarity_top_k_reranker,
)
self.response_synthesizer = get_response_synthesizer(
response_mode=ResponseMode.COMPACT, llm=self.llm
)Explanation
This setup ensures that the |
Beta Was this translation helpful? Give feedback.
-
|
@dosu this does not work. Can we do it here in the function? def process_streamlit_fetch_query_results(self, query="", verbose=False):
try:
query_bundle = QueryBundle(query_str=query)
retrived_nodes = self.fusion_retriever.retrieve(query_bundle)
recency_nodes = self.recency_postprocessor.postprocess_nodes(
retrived_nodes, query_bundle=query_bundle
)
rerank_nodes = self.postprocessor.postprocess_nodes(
nodes=recency_nodes, query_bundle=query_bundle
)
px.active_session().url
return rerank_nodes
except Exception as e:
traceback.print_exc()
raise Exception(f"An error occurred retrieving: {e}") |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
Yes, you can apply the metadata filter within the Updated
|
Beta Was this translation helpful? Give feedback.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
I have the following initialisation for my class:L
constructor code:
I also have the following helper functions:
How can I modify it to use raptor ?
Beta Was this translation helpful? Give feedback.
All reactions