beginner needs some help pls

[Andrew1457]

I want to try Leettools, but I am struggling a bit. Here are the steps I have already done and the error message. Thanks in advance for your help.

First, I created a new Anaconda environment. Then, I installed Leettools with pip. I also downloaded the .env.ollama file and modified it to:

EDS_DEFAULT_LLM_BASE_URL=http://localhost:11434/api/generate
EDS_DEFAULT_INFERENCE_MODEL=llama3.2:3b-instruct-fp16
EDS_DEFAULT_EMBEDDING_MODEL=nomic-embed-text
EDS_EMBEDDING_MODEL_DIMENSION=768

Then, I ran the following command, like in the example:
leet flow -t answer -q "How does GraphRAG work?" -k graphrag -l info
Here is the full error message:

[24/01/25 17:15:28.543] INFO [31980-26820] Loaded flow_options: {} (C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\cli\flow\flow_cli.py:287 flow)
[24/01/25 17:15:29.193] INFO [31980-26820] Initializing the DuckDB StrategyStore. (C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\core\strategy\_impl\duckdb\strategy_store_duckdb.py:62 __init__)

Error: 'charmap' codec can't decode byte 0x9d in position 637: character maps to <undefined>

Use -l DEBUG to see the details.

====================================================================================================

Traceback (most recent call last):
  File "C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\cli\flow\flow_cli.py", line 289, in flow
    result_article = _run_flow_cli(**locals())
                     ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\cli\flow\flow_cli.py", line 37, in _run_flow_cli
    history_manager = get_history_manager(context)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\chat\history_manager.py", line 347, in get_history_manager
    return _HMInstances().get_instance(context)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\chat\history_manager.py", line 335, in get_instance
    hm_instance = factory_util.create_manager_with_repo_type(
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\common\utils\factory_util.py", line 94, in create_manager_with_repo_type
    return create_object(
           ^^^^^^^^^^^^^^
  File "C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\common\utils\factory_util.py", line 51, in create_object
    return cls(*args, **kwwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\chat\_impl\duckdb\history_manager_duckdb.py", line 57, in __init__
    self.strategy_store = context.get_strategy_store()
                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\context_manager.py", line 103, in get_strategy_store
    self._strategy_store = create_strategy_store(
                           ^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\core\strategy\strategy_store.py", line 90, in create_strategy_store
    return factory_util.create_manager_with_repo_type(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\common\utils\factory_util.py", line 94, in create_manager_with_repo_type
    return create_object(
           ^^^^^^^^^^^^^^
  File "C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\common\utils\factory_util.py", line 51, in create_object
    return cls(*args, **kwwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\core\strategy\_impl\duckdb\strategy_store_duckdb.py", line 80, in __init__
    self._run_init()
  File "C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\core\strategy\_impl\duckdb\strategy_store_duckdb.py", line 183, in _run_init
    self.default_strategy = self.create_strategy_from_path(self.default_path)
                            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\core\strategy\_impl\duckdb\strategy_store_duckdb.py", line 388, in create_strategy_from_path
    rewrite_up = self._create_prompt_from_file(
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\andre\anaconda3\envs\test2\Lib\site-packages\leettools\core\strategy\_impl\duckdb\strategy_store_duckdb.py", line 109, in _create_prompt_from_file
    prompt_text = f.read()
                  ^^^^^^^^
  File "C:\Users\andre\anaconda3\envs\test2\Lib\encodings\cp1252.py", line 23, in decode
    return codecs.charmap_decode(input,self.errors,decoding_table)[0]
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
UnicodeDecodeError: 'charmap' codec can't decode byte 0x9d in position 637: character maps to <undefined>

[pengfeng]

Thanks for trying LeetTools out! I think it is a locale mismatch, can you try "locale" command and paste the output?

I will add a patch to enforce the encoding as well.

[Andrew1457]

allright that makes sense. I changed the url to http://localhost:11434/api but nothing changed
query.log
here is the full log. I am out of ideas......

[leettools-dev]

Can you change the -k parameter to "-k graphrag2" since it seems that the existing KB "graphrag" was created using the default openai embedder.

Also, you can turn on more logging by using changing "-l info" to "-l debug" so that we can see which embedder is actually used during the embedding process.

[Andrew1457]

I tried it again. I added this to the command: -e .env.ollama

leet flow -e .env.ollama -t answer -q "How does GraphRAG work?" -k graphrag2 -l debug

Now it uses the right API in the log, but it still fails.
query.log

[pengfeng]

Thanks for reporting back! I didn't notice your first command that no -e option was specified. You can put the config in .env if you are fixed using Ollama so that you don't need to use -e every time.

The new problem is kind of strange since if the embedding for the document segment works, the embedding of the query should work too since they are using the same API. I will try to find a Windows machine to reproduce the problem.

[pengfeng]

We added an embedder check for queries so that we print out a warning when the KB's embedder and the default embedder in the env file are not compatible. Also cleaned up the debug information to remove some unnecessary messages. Please check the new version to see if you can reproduce the error. Thanks!

[pengfeng]

Usually you can use "LANG=en_US.UTF-8" to set the locale so that some of the nice emojis in the logs can be processed:-)

LANG=en_US.UTF-8 leet flow -t answer -q "How does GraphRAG work?" -k graphrag -l info

[pengfeng]

Another user encountered the same problem and it turned out the problem was

EDS_DEFAULT_LLM_BASE_URL=http://localhost:11434/api

should be

EDS_DEFAULT_LLM_BASE_URL=http://localhost:11434/v1

Since Ollama responds to both endpoints and the error got propagated into the OpenAI client lib.