Custom knowledge source embedding failing for local Ollama model

General
1

Hi,
I have a custom knowledge source that I have created in my application. On creating the crew, like this:

crew = Crew(
      agents=[json_analyst, summarizer_agent],
      tasks=[ analysis_task, summary_task],
      verbose=True,
      knowledge_sources=[knowledge_source],
      embedder={
        “provider”: “ollama”,
        “config”: {
        “model”: “nomic-embed-text”
        }
      },
      process=Process.sequential
)

the script fails while creating the embedding with the following error:

"Failed to upsert documents: Expected Embedings to be non-empty list or numpy array, got  in upsert."

My llm for agent is configured like this:

llm =  ChatOpenAI(
    model="ollama/phi3:3.8b",
    base_url="http://localhost:11434/v1"
)

Any pointers around this? How to get embedding working with Ollama models?

2

Can you verify that the Ollama server is running on your system and accessible? If you run this http://localhost:11434/ on your browser, it should look like this.

Screenshot 2025-01-07 at 2.12.20 AM

Also check if you’re using the current version of CrewAI (v0.95.0)
Another side note tid bit, try using the LLM class instead of ChatOpenAI like so:

llm = LLM(
    model="ollama/llama3:70b",
    base_url="http://localhost:11434"
)
3

Yes, my Ollama server is running and I confirmed that. I am also running 0.95.0. Am I required to also use embedder_config in my knowledge source in addition to passing that information to the crew?

4

Issue seems to be resolved when using the previous version instead (0.86.0)