Ollama endpoint seems not to work

[Vasilije1990]

Hello Vasilije,

I just returned from my holiday, and decided to give cognee a go. I am interested to test the local model integration instead of with any third-party. As such, as per docs (https://docs.cognee.ai/core_concepts/local_models), I added the option to use the Ollama service hosted on the same dev machine:
cognee.config.llm_provider = 'ollama'

Using the same code on the same site, I get the following exception when I tried to run:

---

Running cognify to create knowledge graph...

InvalidValueError: LLM API key is not set. (Status code: 422)Coroutine task errored: extract_graph_from_data
('LLM API key is not set.', 'InvalidValueError')
Traceback (most recent call last):
File "/Users/myacc/codes/python/cognee/lib/python3.11/site-packages/cognee/modules/pipelines/operations/run_tasks.py", line 116, in run_tasks_base
task_result = await running_task.run(*args)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/myacc/codes/python/cognee/lib/python3.11/site-packages/cognee/tasks/graph/extract_graph_from_data.py", line 19, in extract_graph_from_data
chunk_graphs = await asyncio.gather(
^^^^^^^^^^^^^^^^^^^^^
File "/Users/myacc/codes/python/cognee/lib/python3.11/site-packages/cognee/modules/data/extraction/knowledge_graph/extract_content_graph.py", line 7, in extract_content_graph
llm_client = get_llm_client()
^^^^^^^^^^^^^^^^
File "/Users/myacc/codes/python/cognee/lib/python3.11/site-packages/cognee/infrastructure/llm/get_llm_client.py", line 37, in get_llm_client
raise InvalidValueError(message="LLM API key is not set.")
cognee.exceptions.exceptions.InvalidValueError: ('LLM API key is not set.', 'InvalidValueError')Async generator task errored: extract_chunks_from_documents
('LLM API key is not set.', 'InvalidValueError')

Even though I am using a local LLM, it is still asking for LLM api key. When I simply added some random key, the client ends up reaching out to OpenAI, and fails.

I searched your Discord community and there was a post from some months ago with a similar issue as mine. But I don't see any specific suggestions on how to resolve the issue to use just Ollama without any OpenAI reliance.

If you have any suggestion, I will be keen to explore.

Thanks,
Kennedy

[Vasilije1990]

The issue should be resolved now with the latest release