RouterRetriever with LLMSingleSelector. The problem is that the selector prompt in SelectorOutputParser isn't in the format that I want. I have prompt template like this:SINGLE_SELECTOR_PROMPT_TEMPLATE = (
"### System:\n"
"You are a helpful assistant. "
"Using only the choices above and not prior knowledge, return "
"the choice that is most relevant to the question: '{query_str}'\n"
"### User:\n"
"Some choices are given below. It is provided in a numbered list "
"(1 to {num_choices}), "
"where each item in the list corresponds to a summary.\n"
"---------------------\n"
"{context_list}"
"\n---------------------\n"
"### Assistant:\n"
)
...
retriever = RouterRetriever(
selector=LLMSingleSelector.from_defaults(
prompt_template_str=SINGLE_SELECTOR_PROMPT_TEMPLATE),
retriever_tools=[vs_tool, summary_tool]
)SelectionOutputParser.parse automatically append a FORMAT_STR like this prompt_template + "\n\n" + _escape_curly_braces(FORMAT_STR) which results in: ### System:
You are a helpful assistant. Using only the choices above and not prior knowledge, return the choice that is most relevant to the question: 'Summarize the uploaded document'
### User:
Some choices are given below. It is provided in a numbered list (1 to 2), where each item in the list corresponds to a summary.
---------------------
(1) Useful for retrieving specific context from uploaded documents.
(2) Useful to retrieve all context from uploaded documents and summary tasks. Don't use if the question only requires more specific context.
---------------------
### Assistant: ====> This phrase is in the wrong position.
The output should be ONLY JSON formatted as a JSON instance.
Here is an example:
[
{{
choice: 1,
reason: "<insert reason for choice>"
}},
...
]### Assistant: to right after the FORMAT_STR?persist_dir with the same storage context. But I can't have the same storage context because users don't upload documents as the same time => indices will be override when new document arrives.{s3_bucket}/{doc_id}, maybe it will work. Although, in loading state, I have to use a loop to load all documents' index.SubQuestionQueryEngine , is there anything with the same function for ContextChatEngine? Thanks in advance.index_store.json, docstore.json, graph_store.json do when I persist the index into my file system? And also what is the purpose of them when I later using the persisted index to query? I don't see so much information contained in them. I'm using PostgreSQL VectorStoreIndex with S3 File System. Thanks in advance.User postgresql table that populates a relationship with the table created by PGVectorStore to mark the owner of documents in the vector store. Something like this:class User(Base):
__tablename__ = "users"
id = Column(Integer, primary_key=True, index=True)
email = Column(String, unique=True, index=True)
hashed_password = Column(String)
files = relationship("File", back_populates="owner")sqlalchemy model but don't know how to make PGVectorStore uses this instead of the default model.class File(Base):
__tablename__ = "data_pg_vector_store"
id = Column(Integer, primary_key=True, index=True)
text = Column(String, index=True)
metadata_ = Column(JSON, index=True)
node_id = Column(String, unique=True, index=True)
embedding = Column(Vector(1024))
owner_id = Column(Integer, ForeignKey("users.id"))
owner = relationship("User", back_populates="files")text_store = LanceDBVectorStore(uri="lancedb", table_name="text_collection")
image_store = LanceDBVectorStore(uri="lancedb", table_name="image_collection")
storage_context = StorageContext.from_defaults(
vector_store=text_store, image_store=image_store
)
index = MultiModalVectorStoreIndex.from_vector_store(
vector_store=text_store,
embed_model=HuggingFaceEmbedding(model_name="BAAI/bge-large-en-v1.5"), # dim = 1024
image_vector_store=image_store,
image_embed_model=ClipEmbedding(model_name="ViT-L/14") # dim = 768
)
retriever_engine = index.as_retriever(
similarity_top_k=2, image_similarity_top_k=2
)ValueError: Query vector size 1024 does not match index column size 768use_async=True return errors. Am I doing anything wrong? Here is the example error when I use SubQuestionQueryEngine:File "/opt/anaconda3/envs/talking-resume/lib/python3.11/site-packages/llama_index/core/async_utils.py", line 49, in run_async_tasks
outputs: List[Any] = asyncio.run(_gather())
^^^^^^^^^^^^^^^^^^^^^^
File "/opt/anaconda3/envs/talking-resume/lib/python3.11/asyncio/runners.py", line 186, in run
raise RuntimeError(
RuntimeError: asyncio.run() cannot be called from a running event loop
/opt/anaconda3/envs/talking-resume/lib/python3.11/site-packages/uvicorn/protocols/http/httptools_impl.py:-1: RuntimeWarning: coroutine 'run_async_tasks.<locals>._gather' was never awaited
RuntimeWarning: Enable tracemalloc to get the object allocation traceback
/opt/anaconda3/envs/talking-resume/lib/python3.11/site-packages/uvicorn/protocols/http/httptools_impl.py:-1: RuntimeWarning: coroutine 'SubQuestionQueryEngine._aquery_subq' was never awaited
RuntimeWarning: Enable tracemalloc to get the object allocation traceback