Closed prasantpoudel closed 3 months ago
Please see #696 for non-English content, and #657 for non-OpenAI models. There may be commentary from the community that helps resolve your issue.
This is caused by not doing ensure_ascii=False in json.dumps. I solved this by changing the class FileWorkflowCallbacks in file_workflow_callbacks.py as follows
class FileWorkflowCallbacks(NoopWorkflowCallbacks):
"""A reporter that writes to a file."""
_out_stream: TextIOWrapper
def __init__(self, directory: str):
"""Create a new file-based workflow reporter."""
Path(directory).mkdir(parents=True, exist_ok=True)
self._out_stream = open( # noqa: PTH123, SIM115
Path(directory) / "logs.json", "a", encoding="utf-8", errors="strict"
)
def on_error(
self,
message: str,
cause: BaseException | None = None,
stack: str | None = None,
details: dict | None = None,
):
"""Handle when an error occurs."""
self._out_stream.write(
json.dumps({
"type": "error",
"data": message,
"stack": stack,
"source": str(cause),
"details": details,
}, ensure_ascii=False)
+ "\n"
)
message = f"{message} details={details}"
log.info(message)
def on_warning(self, message: str, details: dict | None = None):
"""Handle when a warning occurs."""
self._out_stream.write(
json.dumps({"type": "warning", "data": message, "details": details}, ensure_ascii=False) + "\n"
)
_print_warning(message)
def on_log(self, message: str, details: dict | None = None):
"""Handle when a log message is produced."""
self._out_stream.write(
json.dumps({"type": "log", "data": message, "details": details}, ensure_ascii=False) + "\n"
)
message = f"{message} details={details}"
log.info(message)
def _print_warning(skk):
log.warning(skk)
Is there an existing issue for this?
Describe the issue
I am testing the graphrag for japanese text to create a graph and query the text file using local and global method.
issue1- when the community report are generated , many part of the report the result are \"title\": \"\u5343\u6210\u5de5\u696d\u682a\u5f0f\u4f1a\u793e like unicode format
issue2: when i query using the local method, it don't give the output from the text that i feeded at the time of indexing process. the llm is giving it own knowledge instead of searching from the lancedb or from community report.
model use llama3 using ollama service