hello everyone,
I'm running alfresco-ai-framework (module ai-rag-framework) locally with Ollama (qwen2.5 + nomic-embed-text) and Elasticsearch 8.15.3.
Ingestion works. The index contains 3 documents with full content and embeddings. Retrieval also works — /api/chat returns the correct documentMetadata (right filenames, sensible distances ~0.4–0.6).
But the LLM never answers from the content. When I ask a question that is clearly answered in one of the PDFs, the model replies "the context does not contain this information." Once it even replied about "Alfresco Share" — a topic that only appears in a filename, not in my question. That suggests the LLM is receiving only metadata, not the content field.
Any working example of QuestionAnswerAdvisor + Elasticsearch + Ollama would help.
Thanks.
