Error when load doc

[iriderprokhorov]

Hi! Thank you for project!
I got error when load docs
Docker installation, MODEL=deepseek-r1:7b,
Docker version 28.1.1, build 4eba37
Python 3.11.11

LOGS

deepgraph-rag-service | IndexError: list index out of range
deepgraph-rag-service | 2025-07-31 11:19:11.512 Uncaught app execution
deepgraph-rag-service | Traceback (most recent call last):
deepgraph-rag-service | File "/usr/local/lib/python3.11/site-packages/streamlit/runtime/scriptrunner/exec_code.py", lin e 128, in exec_func_with_error_handling
deepgraph-rag-service | result = func()
deepgraph-rag-service | ^^^^^^
deepgraph-rag-service | File "/usr/local/lib/python3.11/site-packages/streamlit/runtime/scriptrunner/script_runner.py", line 669, in code_to_exec
deepgraph-rag-service | exec(code, module.dict) # noqa: S102
deepgraph-rag-service | ^^^^^^^^^^^^^^^^^^^^^^^^^^^
deepgraph-rag-service | File "/usr/src/app/app.py", line 64, in
deepgraph-rag-service | process_documents(uploaded_files,reranker,EMBEDDINGS_MODEL, OLLAMA_BASE_URL)
deepgraph-rag-service | File "/usr/src/app/utils/doc_handler.py", line 60, in process_documents
deepgraph-rag-service | vector_store = FAISS.from_documents(texts, embeddings)
deepgraph-rag-service | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
deepgraph-rag-service | File "/usr/local/lib/python3.11/site-packages/langchain_core/vectorstores/base.py", line 848, i n from_documents
deepgraph-rag-service | return cls.from_texts(texts, embedding, metadatas=metadatas, **kwargs)
deepgraph-rag-service | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
deepgraph-rag-service | File "/usr/local/lib/python3.11/site-packages/langchain_community/vectorstores/faiss.py", line 1044, in from_texts
deepgraph-rag-service | return cls.__from(
deepgraph-rag-service | ^^^^^^^^^^^
deepgraph-rag-service | File "/usr/local/lib/python3.11/site-packages/langchain_community/vectorstores/faiss.py", line 1001, in __from
deepgraph-rag-service | index = faiss.IndexFlatL2(len(embeddings[0]))
deepgraph-rag-service | ~~~~~~~~~~^^^
deepgraph-rag-service | IndexError: list index out of range

[onestardao]

you just stepped on one of the most invisible traps in RAG pipelines

when ingesting docs, if your text chunk turns out empty or too short to produce a valid embedding, most frameworks silently pass in a zero vector without warning

this leads to runtime errors like “list index out of range” or, worse, hallucinated matches later on

quick fix: before pushing to your vector DB, always add a sanity check to skip chunks where len(embedding) == 0 or where the chunk is just whitespace

bonus tip: also log how many chunks get dropped during ingestion. you’d be surprised how often the pipeline runs without storing anything real