@Jedi did you manage to resolve this

sshinji3046

@Jedi did you manage to resolve this issue ? I too face similar issues

16 comments

LLogan M

just do pip install -U sqlite-utils

sshinji3046

@Logan M - The issue still persists. It is in the latest version of sqlite-utils-3.36

LLogan M

hmm yea, found the bug, pushing a fix soon

LLogan M

should be fixed if you just do pip install -U llama-index-core

sshinji3046

@Logan M i had a quick check , and also updated the both sqlite-utils and llama-index-core but the issue still remains. I am using mistral-7b instruct model but i think it should not matter , on the choice of llm
SW versions
llama-index-core 0.10.14
sqlite-fts4 1.0.3
sqlite-utils 3.36

LLogan M

Works fine for me on a fresh venv
https://colab.research.google.com/drive/1QUymvROoOT_-8HXyYw9n9s1vdNj_PyvG?usp=sharing

sshinji3046

i will give it a try with mistral instruct . Now i am curious why it didnt work for me . I will just add the hugging face llm function to the colab notebook and check

sshinji3046

I think for soem reason it doesnt work for mistral instruct . I also tried a simple print statement just to see perhaps if there was something wrong . Added loading of model from hugging face
notebook link : https://colab.research.google.com/drive/18o5tvNZWufcseIHvXJEQv1P4i1K74s2p?usp=sharing

LLogan M

Very spooky, I checked the source code and the fix I made didn't get comitted. No idea how my notebook worked though

LLogan M

ok, patched it, NOW it works (although the model does not write a correct query lol)

sshinji3046

ok you mean something like this right ?

Plain Text

OperationalError: near "sql": syntax error

The above exception was the direct cause of the following exception:

ValueError                                Traceback (most recent call last)
/usr/local/lib/python3.10/dist-packages/llama_index/core/query_engine/jsonalyze_query_engine.py in default_jsonalyzer(list_of_dict, query_bundle, llm, table_name, prompt, sql_parser)
    108     except sqlite_utils.utils.sqlite3.OperationalError as exc:
    109         print_text(f"Error executing query: {sql_query}")
--> 110         raise ValueError("Invalid query") from exc
    111 
    112     return sql_query, table_schema, results

LLogan M

yup exactly 👀

LLogan M

So now everything is working, just the LLM is not cooperating 😅 Pretty common when trying to parse output from open-source models

sshinji3046

oh shoot. The json parser would have been great to query different part of the jsons .... too bad this tool is not working with other open source models. but in general a good concept.

LLogan M

Maybe with a different model or some prompt engineering it might perform better 🤔

sshinji3046

actuallly not sure , it is said the mistral 7b instruct model is the best is what i have understood

Add a reply

Find answers from the community

@Jedi did you manage to resolve this