openclaw - 💡(How to fix) Fix NC criminal court pilot: connect to Grey Matter and run verifiable prediction benchmark [1 comments, 2 participants]

Fix Plan

To connect the NC criminal court pilot to Grey Matter and run the prediction benchmark, we need to perform the following steps:

• Translate the knowledge graph schema to Grey Matter engram headers and edges
• Ingest the data into Grey Matter brain
• Define and run a verifiable benchmark

Step-by-Step Solution

1. Translate knowledge graph schema:
   • Use a Python script to read knowledge_graph.json and create Grey Matter engram headers and edges.
   • Example code:

import json

with open('knowledge_graph.json') as f: knowledge_graph = json.load(f)

engram_headers = [] engram_edges = []

for node in knowledge_graph['nodes']: engram_headers.append({ 'id': node['id'], 'type': node['type'] })

for edge in knowledge_graph['edges']: engram_edges.append({ 'source': edge['source'], 'target': edge['target'], 'type': edge['type'] })

2. **Ingest data into Grey Matter brain**:
   * Use the Grey Matter API to create a new brain and add the translated engram headers and edges.
   * Example code:
     ```python
import requests

brain_id = 'nc-criminal-court-pilot'
response = requests.post(f'https://grey-matter.com/brains/{brain_id}/engrams', json={'headers': engram_headers, 'edges': engram_edges})

1. Define and run verifiable benchmark:
   • Create a test harness using prediction_test.py to predict rulings based on case facts and statutes.
   • Compare the results to a keyword/string matching baseline.

Verification

• Verify that the data has been successfully ingested into Grey Matter brain by checking the brain's engram headers and edges.
• Run the prediction benchmark and compare the results to the ground truth (published decisions).

Extra Tips

• Ensure that the knowledge graph schema is correctly translated to Grey Matter engram headers and edges.
• Use a robust testing framework to verify the accuracy of the prediction benchmark.