| { | |
| "embedder": "BAAI/bge-small-en-v1.5", | |
| "metric": "cosine(ip on normalized vectors)", | |
| "chunk_tokens": 220, | |
| "chunk_overlap": 40, | |
| "datasets": [ | |
| [ | |
| "ise-uiuc/Magicoder-OSS-Instruct-75K", | |
| null, | |
| "train" | |
| ], | |
| [ | |
| "databricks/databricks-dolly-15k", | |
| null, | |
| "train" | |
| ] | |
| ], | |
| "note": "Reasoning RAG KB built ONLY from the training-type datasets (code + report)." | |
| } |