Error searching for documents: list indices must be integers or slices, not str

General
,
1

I am using local pdf to be added to my chromadb vector store and the pdf is successfully added to the vs but there is some issue in search.

from agno.agent import Agent
from agno.vectordb.chroma import ChromaDb 
from agno.models.openai import OpenAIChat
from agno.embedder.openai import OpenAIEmbedder
from agno.knowledge.pdf import PDFKnowledgeBase
from agno.document.reader.pdf_reader import PDFReader
from dotenv import load_dotenv
import os

load_dotenv()


if __name__ == "__main__":
    vector_db = ChromaDb(collection="recipes",
                     path="tmp\chromadb",
                     persistent_client=True,
                     embedder=OpenAIEmbedder(api_key=os.environ["OPENAI_API_KEY"]))
    
    knowledge_base = PDFKnowledgeBase(
                            path="path/to/my/pdf",
                            vector_db=vector_db,
                            reader=PDFReader(chunk=True),
                        )
    
    knowledge_base.load(recreate=True)
    agent = Agent(model=OpenAIChat(id="gpt-4o",api_key=os.environ["OPENAI_API_KEY"]),
                  knowledge=knowledge_base,
                  show_tool_calls=True)
    agent.print_response(f"Instruction : Based on the context answer the question below.Answer in detail with facts from context.\nQuestion : What is the document about?", markdown=True)
2

Hi @shivamSingh

Thanks for reaching out and for using Agno! I’ve looped in the right engineers to help with your question. We usually respond within 24 hours, but if this is urgent, just let us know, and we’ll do our best to prioritize it.
Appreciate your patience—we’ll get back to you soon! :smile:

1 Like
3

I traced the issue back to

agno>vectordb>chroma>chromadb.py>ChromaDb>search

    search_results: List[Document] = []
    ids = result.get("ids", [[]])[0]
    metadata = result.get("metadatas", [[]])[0]  # type: ignore // The issue is here since i have no metadata this returns [None]
    documents = result.get("documents", [[]])[0]  # type: ignore
    embeddings = result.get("embeddings")
    distances = result.get("distances", [[]])[0]  # type: ignore
    uris = result.get("uris")
    data = result.get("data")
    metadata["distances"] = distances  # type: ignore // then we try to add distance into [None] which is not valid
    metadata["uris"] = uris  # type: ignore
    metadata["data"] = data  # type: ignore

i fixed this by updating :
metadata = result.get("metadatas", [[]])[0] --> metadata = {}
This is what got it working for my case. But if possible do let me know about the more proper solution.

4

Hey @shivamSingh! You are on the correct path. ChromaDb needs to be fixed and I will be on it soon enough. Hopefully in the upcoming releases it will be done

1 Like
5

Thanks for replying, I would like to add that
metadata = result.get("metadatas", [[]])[0] returns a list of dicts instead of a dict hence doing metadata["distances"] = distances gives : Error searching for documents: list indices must be integers or slices, not str : which is not the real cause.
Again Thanks for the response will use my local patch until new updates are pushed. :grin:

6

Hey @shivamSingh! In our recent release, we fixed chromaDB. Please try it out and let me know if you still are facing the same issue