Description

Using ChromaDB as Vector Database of Knowledge Base is always giving response - No Documents Found.
After calling the search function on Knowledge Base, It logs "Error searching for documents: list indices must be integers or slices, not str"

Steps to Reproduce

Create a Knowledge Base like this:

from agno.agent import AgentKnowledge
from agno.embedder.sentence_transformer import SentenceTransformerEmbedder
from agno.vectordb.chroma import ChromaDb

knowledge_base = AgentKnowledge(
    vector_db=ChromaDb(
        collection="collection_name",
        persistent_client=True,
        path="path/to/db",
        embedder=SentenceTransformerEmbedder(
            id="all-mpnet-base-v2" # I was using this embedding model.
        )
    ),
)

docs = knowledge_base.search(query="Some Query",num_documents=5)
for doc in docs:
    print(doc,"\n\n")

Expected Behavior

It should have Listed down 5 relevant documents from the Persistent ChromaDB.

Actual Behavior

It logged the Error saying "Error searching for documents: list indices must be integers or slices, not str" and returned empty list of Documents.

Screenshots or Logs (if applicable)

Environment

• OS: Ubuntu Linux
• Agno Version: 1.0.4
• Additional Environment Details: Python 3.12.8

Possible Solutions (optional)

At this code in the file agno/vectordb/chroma Line no. 205 - 214

ids = result.get("ids", [[]])[
metadata = result.get("metadatas", [[]])[0]  # type: 
documents = result.get("documents", [[]])[0]  # type: 
embeddings = result.get("
distances = result.get("distances", [[]])[0]  # type: 
uris = result.get("
data = result.get("
metadata["distances"] = distances  # type: 
metadata["uris"] = uris  # type: 
metadata["data"] = data  # type: ignore

The metadata will receive the value [[]] (list of list), But indices like "distances", "uris", "data" is directly accessed from metadata which will have value as (list of dict).
This will throw the error saying list indices must be int, not str.