Skip to content

Unable to Provide insights on Overall Data - Only Taking top 5 or 7 chunks #100

@Curiosity007

Description

@Curiosity007

.env

Generic

TEXT_EMBEDDINGS_MODEL=sentence-transformers/all-MiniLM-L6-v2
TEXT_EMBEDDINGS_MODEL_TYPE=HF # LlamaCpp or HF
USE_MLOCK=false

Ingestion

PERSIST_DIRECTORY=db
DOCUMENTS_DIRECTORY=source_documents
INGEST_CHUNK_SIZE=500
INGEST_CHUNK_OVERLAP=50
INGEST_N_THREADS=1

Generation

MODEL_TYPE=LlamaCpp # GPT4All or LlamaCpp
MODEL_PATH=eachadea/ggml-vicuna-7b-1.1/ggml-vic7b-q5_1.bin
MODEL_TEMP=0.8
MODEL_N_CTX=2048 # Max total size of prompt+answer
MODEL_MAX_TOKENS=1024 # Max size of answer
MODEL_STOP=[STOP]
CHAIN_TYPE=betterstuff
N_RETRIEVE_DOCUMENTS=100 # How many documents to retrieve from the db
N_FORWARD_DOCUMENTS=100 # How many documents to forward to the LLM, chosen among those retrieved
N_GPU_LAYERS=32

Python version

Python 3.10.10

System

Description: Ubuntu 22.04.2 LTS Release: 22.04 Codename: jammy

CASALIOY version

Latest Commit - ee9a4e5

Information

  • The official example scripts
  • My own modified scripts

Related Components

  • Document ingestion
  • GUI
  • Prompt answering

Reproduction

I have fed the system a 5000 line csv file, with 30 columns.

Now I asked about overall insight from the data.

I can see in the terminal, it is only seeing top 5 or 7 documents, which is nothing but single row. So, this is giving me answer based on 5 or 7 rows, and thus no actual insight is coming

Point to be noted - I have kept only 1 document in the source documents folder to avoid information overlapping

Expected behavior

Should be able to understand the pattern in the data, and suggest some insights based on it.

Metadata

Metadata

Assignees

No one assigned

    Labels

    help wantedExtra attention is needed

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions