Batch entity summarization

[joerattazzi-microsoft]

Description

In cases where you're updating your graph with a large number of inputs, the asyncio.gather command during Updating Final Entities step can completely block and time out your calls. Link to code:

graphrag/graphrag/index/update/entities.py

Line 138 in f14cda2

results = await asyncio.gather(*tasks)

Here, I'm proposing a change to batch the calls through a configurable value

Related Issues

#1707

Proposed Changes

• Configurable batch_size added to summarize_descriptions_config
• Do batching of calls to the LLM for run_entity_summarization
• Add logging to help clarify what's going on

Checklist

• I have tested these changes locally.
• I have reviewed the code changes.
• I have updated the documentation (if necessary).
• I have added appropriate unit tests (if applicable).

Additional Notes

[Add any additional notes or context that may be helpful for the reviewer(s).]

Tested with a fairly large data set via our Synapse pipeline