Embedding not getting created using ollama

[guptagaurav89]

Setup notebook on local machine with docker inside WSL. Models' setup using ollama on azure VM. Connection test to language model is working fine, but embedding is failing with error -
Model Test Failed
nomic-embed-text:latest

Connection error - check network/endpoint

And with source file getting below error
INFO | commands.embedding_commands:embed_source_command:342 - Starting embedding for source: source:hlfs1gmdtn3diou7w6qb
2026-02-20 10:35:41.804 | INFO | commands.embedding_commands:embed_source_command:370 - Created 82 chunks for source source:hlfs1gmdtn3diou7w6qb (sizes: min=34, max=1198, avg=919 chars)
2026-02-20 10:35:45.992 | ERROR | surreal_commands.core.service:execute_command:283 - Command open_notebook.embed_source (command:0v7npdf84xaa8giypgft) failed after 5 attempt(s): Failed to generate embeddings using model 'nomic-embed-text:latest' (batch 1/2, 50 texts): Failed to get embeddings: All connection attempts failed

Please suggest the solution, if any configuration is missed

[lfnovo]

This is weird.. can you show how you configured the ollama provider?

[guptagaurav89]

Ollama URL - http://hostname.region.cloudapp.azure.com:11434

Also, curl test command for both chat and embedding to Ollama is working inside notebook container.

curl http://...:11434/api/embeddings -d '{

"model": "nomic-embed-text",
"> prompt": "test"
}'
{"embedding":[0.6649775505065918,0.2699969708919525,-4.426666736602783,-0.20665805041790009,1.4542583227157593,0.1424468606710434,1.1023224592208862,-0.09218410402536392,0.8590682148933411,-0.6400354504585266,0.11621516942977905,1.575396180152893,0.7940111756324768,0.13871563971042633,-1.048120141029358,-1.2554664611816406,1.2376112937927246,-2.093083620071411,-0.13985379040241241,-0.13259878754615784,0.06555783003568649,-0.2713017165660858,-1.3450208902359009,0.16999131441116333,3.123197317123413,-0.376568466424942,-1.452771782875061,1.372378945350647,-0.6342620849609375,-0.4403989017009735,0.8748071789741516,-0.6545827388763428,1.1640567779541016,-1.7903121709823608,-0.7306339144706726,-0.658701479434967,0.5709836483001709,1.1556618213653564,0.2697652280330658,-0.

[NbtKmy]

I have same issue - I have Ollama directly on the host (MacOS). And Notebook through Docker. The API endpoint is thus "http://host.docker.internal:11434". The LLM works very well on Notebook, but the embedding model not.

[crystalskim]

I have the same exact issue. Ollama on Win11, Open Notebook on Docker.

[BallyCode]

Same for me. It won't assign an embedding model. I tried downloading one of the embedding models into the ollama folder, still nothing.

[lfnovo]

This was a known bug in the Ollama embedding provider, where the configured base_url was being ignored and requests fell back to localhost — which is exactly why the language model worked but embeddings failed against a remote Ollama endpoint.

It has since been fixed upstream in Esperanto (2.19.6+), and the latest version of Open Notebook should work out of the box. Please update to the latest release and re-test — your remote Ollama embedding model should connect correctly now.

If you still run into issues on the latest version, let us know and we can dig in further.