Fetching metadata from the HF Docker repository...
Sep 15 22:24 chore: update docs/repo url Aug 24 18:53 feat: set KLEA_RAG_ENV_FILE 18:43 chore: generalise file copy 11:17 fix: correct command 11:14 feat: use local guard model Not ideal, but better than disabling 10:02 fix(docker): use non-root user for all ollama ops Otherwise during runtime, it still cant write to model specific folders
which were created as root. 09:54 chore: correct ollama model dir It was being run as the root user, but then used as the non-root user. 09:39 chore: pull model again on runtime Should be cached, so shouldnt require too much time. 09:32 chore: regenerate all stores Aug 22 11:48 chore: improve file listing 11:46 chore: list all files recursively 11:01 feat: use local ollama for embedding model 10:02 chore: standardize to `store/` for all stores 09:58 feat: separate source and app folders Aug 19 16:05 chore: update readme 16:04 chore: use a better name 16:03 chore: import scripts to process nml-docs 15:11 Merge branch 'main' of hf.co:spaces/NeuroML/NeuroKLEA 15:10 chore: regenerate nml-docs Now includes bibliography correctly Aug 17 13:59 chore: temporarily disable guard node Keeps erring on HF. Not yet diagnosed. Aug 12 18:40 feat: correctly build nml-docs 17:23 fix(stores): regenerate with correct mapping file 16:47 fix(data): correct metadata map format 15:46 fix: also update non-mcp config 15:25 feat: update vector stores We regenerated them using the new docling flow, also generated the bm25
sources. Aug 3 21:11 chore: add k values for nml docs vector store Jul 30 20:01 fix: check in new config 19:52 feat: temporarily disable mcp server 17:01 feat: update to use new webgui Jul 2 22:20 fix: update package extras Jun 26 14:12 fix: correct server url 13:55 feat: copy in vector stores 13:06 feat: update configs for klea updates Jun 19 15:20 rename to neuroklea 15:11 feat: update for latest rag May 14 08:30 fix: update vector store config 08:16 feat: update docker and env files Models on HF have changed again. Waiting to get access from
Llama 4 for the guard model. Feb 18 17:42 feat: limit api to internal hostname Feb 15 22:42 chore: update envvars Jan 30 12:17 feat: switch model Another inference provider gone missing. Jan 28 15:17 feat: use smaller Qwen3 non-thinking model 15:17 feat: correct vector store paths 14:35 feat(stores): make a couple relative To see if the rag code works correctly here too. 12:56 feat: switch to Qwen3-32B gpt-oss doesn't seem to follow instructions---it doesn't even classify
queries correctly..
https://huggingface.co/inference/models?model=openai%2Fgpt-oss-120b&search=Qwen%2FQwen3
https://huggingface.co/Qwen/Qwen3-32B 12:43 feat: switch to gpt-oss-120b for chat https://huggingface.co/inference/models?model=openai%2Fgpt-oss-120b&search=openai%2Fgpt-oss-120b
The qwen model we were using doesn't seem to have any inference
providers now. Jan 27 14:35 chore: use fastapi `run` 13:40 fix: correct scripts path 13:37 chore(hf): debug 13:22 feat(deployment): update huggingface Dec 3, 2025 21:23 fix: correct readme yaml 
