Indexing stuck at 'Workflow extract_graph started.' stage. Fails with 18.18 percent complete

[manuchadha1979]

The indexing stage of the Quickstart notebook is stuck at the following stage:

{
'status_code': 200,
'index_name': 'graphrag-index16',
'storage_name': 'filesforgraphrag16',
'status': 'running',
'percent_complete': 18.18,
'progress': 'Workflow extract_graph started.',
}

after few mins

{
'status_code': 200,
'index_name': 'graphrag-index16',
'storage_name': 'filesforgraphrag16',
'status': 'failed',
'percent_complete': 18.18,
'progress': 'Workflow extract_graph started.',
}

[RichardHallgren]

I get the same error as you, same files etc worked with the old solution accelerator.

Did you get any errors with the CosmosDB during creation etc? My guess it's something with assigning access to CosmosDB that isn't working correctly during deployment.

[timothymeyers]

Check the logs for your graphrag-index-manager kubernetes job and your indexing-job- job.

You may be able to find out what the error was.

[manuchadha1979]

sorry, I don't know how to. The deployment was successful so I don't think there was an issue. Is there a way to try old accelerator. I just want to see how to deploy /run it for learning perspective

[RichardHallgren]

For me it was the autogenerated prompts for extracting entities that had an error in it. The generated prompt had } where it shouldn't be, after removing the extra } it worked. I would start double checking your prompts to see if you can find anything, or try with default prompts.

[manuchadha1979]

are you referring to quickstart or advanced notebook?

[RichardHallgren]

are you referring to quickstart or advanced notebook?

The GraphRAG Accelerator API gives you the possibility to generate prompts, there is an example in the advanced notebook. Those generated prompts caused my problem, but best is of course to try and dig into the logs and see what you can find.

[smallangi]

I am also experiencing this. we see the following error in the Storage acct logs/report folder. Upon checking we see that extract_entities/graph_extractor.py does not exist in the graphrag.

"stack": "Traceback (most recent call last):\n File "/usr/local/lib/python3.10/site-packages/graphrag/index/operations/extract_entities/graph_extractor.py", line 127, in call\n result = await self._process_document(text, prompt_variables)\n File "/usr/local/lib/python3.10/site-packages/graphrag/index/operations/extract_entities/graph_extractor.py", line 156, in _process_document\n self._extraction_prompt.format(**{\nValueError: unexpected '{' in field name\n",

[RichardHallgren]

That's the same issue I had, but with }, if you change the manual prompt (remove that unexpected '{' from the prompt or close it if needed) it should work.