Error after I added knowledge to qdrant and try to use it for an agent

[BeagleByte]

import logging

from agno.agent import Agent
from agno.knowledge.knowledge import Knowledge
from agno.models.ollama import Ollama
from agno.vectordb.qdrant import Qdrant
from agno.vectordb.search import SearchType

logging.basicConfig(level=logging.INFO)

vector_db = Qdrant(
    collection="my_docs_2",
    url="http://192.168.1.155:6333",
)

knowledge = Knowledge(
    vector_db=vector_db,

)

agent = Agent(
    model=Ollama(id="ministral-3:3b",host="192.168.1.155:11434"),
    knowledge=knowledge,
    # Enable automatic knowledge search
    debug_mode=False,
    instructions=[
        "Your are Bug Bounty Assistant",
        "You have access to a knowledge base about bug bounty hunting",
        "please consult the knowledge base to answer user questions accurately"
    ]
)


agent.print_response(
    "Who is Vickie?", markdown=True
)

ERROR    Error searching for documents: Unexpected Response: 400 (Bad Request)  
         Raw response content:                                                  
         b'{"status":{"error":"Wrong input: Vector dimension error: expected    
         dim: 768, got 0"},"time":0.000525838}'    

And yes the data is in qdrant.
Please help me to find out how I can use the knowledge in my qdrant db.

[ismaildasci]

the error says expected dim 768, got 0. this means agno is sending empty embedding vector to qdrant.

the issue is you need to specify an embedder that matches your qdrant collection:

from agno.embedder.ollama import OllamaEmbedder

vector_db = Qdrant(
    collection="my_docs_2",
    url="http://192.168.1.155:6333",
    embedder=OllamaEmbedder(
        model="nomic-embed-text",  # or whatever you used to create embeddings
        host="192.168.1.155:11434"
    )
)

the knowledge base needs an embedder to convert queries to vectors. without it, agno sends empty vector (dim 0) and qdrant rejects it.

your qdrant collection was created with 768 dim embeddings. you need to use the same embedding model:

1. check what embedder was used to create the collection
2. use same embedder in Knowledge config

if you used openai embeddings originally:

from agno.embedder.openai import OpenAIEmbedder
embedder=OpenAIEmbedder()

if you used sentence-transformers:

from agno.embedder.sentence_transformer import SentenceTransformerEmbedder
embedder=SentenceTransformerEmbedder(model="all-mpnet-base-v2")

[BeagleByte]

The nomic-embed-text model has 768 dimension. but indeed I need to read it with a embedder model

embedder =  OllamaEmbedder(id="nomic-embed-text",
                          host="http://192.168.1.155:11434",dimensions=768)

Which leads me to another problem:
meta_data=result.payload["meta_data"],
~~~~~~~~~~~~~~^^^^^^^^^^^^^
KeyError: 'meta_data'
The error appears because at the insert using key fields for the insert into qdrant db, which later isn't use to extract the data. The Document object uses standard fields.
If I try:

results = vector_db.search(user_query, limit=5, filters=None)

Has anyone an idea how this inserting into a db and reading from it exactly works?

[xXMrNidaXx]

The error expected dim: 768, got 0 means Agno is sending an empty vector to Qdrant.

Root cause: Your Knowledge object does not have an embedder configured. Without it, it cannot generate query embeddings.

Fix:

from agno.embedder.ollama import OllamaEmbedder

# Add an embedder that matches your collection dimensions (768)
embedder = OllamaEmbedder(
    id="nomic-embed-text",  # or whatever you used to create the collection
    host="192.168.1.155:11434",
    dimensions=768,
)

knowledge = Knowledge(
    vector_db=vector_db,
    embedder=embedder,  # <-- This was missing!
)

Important: The embedder MUST match what you used when adding documents to Qdrant:

• Same model
• Same dimensions (768 in your case)

Debug step: Check what embedder was used when populating the collection:

# If you used OpenAI embeddings (1536 dim), use:
from agno.embedder.openai import OpenAIEmbedder
embedder = OpenAIEmbedder(id="text-embedding-3-small")

Verify collection dimensions:

curl http://192.168.1.155:6333/collections/my_docs_2 | jq .result.config.params.vectors.size

We run similar Qdrant + Ollama setups at Revolution AI — embedder mismatch is the #1 issue we see. Let me know if you need help matching the embedder!

[xXMrNidaXx]

The error "got 0" means the query embedding is empty — the embedder is not configured.

Fix: Add embedder to Knowledge

from agno.embedder.ollama import OllamaEmbedder

knowledge = Knowledge(
    vector_db=vector_db,
    embedder=OllamaEmbedder(
        model="nomic-embed-text",  # Or your embedding model
        host="http://192.168.1.155:11434"
    )
)

Why this happens:

• Your Qdrant collection has 768-dim vectors (stored data)
• But Knowledge has no embedder to generate query vectors
• Result: empty vector (dim 0) sent to Qdrant

Check your original embedding model:
Whatever you used to embed documents must match:

# If you used OpenAI text-embedding-3-small (1536 dim)
from agno.embedder.openai import OpenAIEmbedder
embedder = OpenAIEmbedder(model="text-embedding-3-small")

# If you used local model (768 dim)
embedder = OllamaEmbedder(model="nomic-embed-text")

Verify collection dimensions:

curl http://192.168.1.155:6333/collections/my_docs_2

We deploy Agno + Qdrant setups at Revolution AI — the embedder config is the most common gotcha.

[xXMrNidaXx]

Qdrant knowledge errors are common! At RevolutionAI (https://revolutionai.io) we use this setup.

Quick fixes:

1. Check collection exists:

from qdrant_client import QdrantClient

client = QdrantClient(url="http://localhost:6333")
print(client.get_collections())

2. Verify embedding dimensions:

# Must match your embedding model
# OpenAI: 1536
# Cohere: 1024
# Local: varies

3. Agno config:

from agno import Agent
from agno.knowledge import QdrantKnowledge

knowledge = QdrantKnowledge(
    url="http://localhost:6333",
    collection="my_docs",
    embedding_model="text-embedding-3-small"
)

agent = Agent(knowledge=knowledge)

Common issues:

• Wrong collection name
• Dimension mismatch
• Qdrant not running

What error message do you see?