dvc.lock

schema: '2.0'
stages:
  fetch-metadata:
    cmd: python scripts/fetch_eidc_metadata.py data/eidc_metadata.json -s 0
    deps:
    - path: scripts/fetch_eidc_metadata.py
      hash: md5
      md5: 82907434d9521996e30014df01bbba8e
      size: 964
    outs:
    - path: data/eidc_metadata.json
      hash: md5
      md5: fb338ea98ce71bf7f002be952b6db0e1
      size: 12275265
  prepare:
    cmd: python scripts/extract_metadata.py data/eidc_metadata.json data/extracted_metadata.json
    deps:
    - path: data/eidc_metadata.json
      hash: md5
      md5: 423dc3a61ede72e1d5c818d74277c0b4
      size: 12140491
    - path: scripts/extract_metadata.py
      hash: md5
      md5: c2fa7d2c4b8f28a6e24536ce0df244fd
      size: 1296
    outs:
    - path: data/extracted_metadata.json
      hash: md5
      md5: 7d2ae8d6a41a960592f30496eb498af7
      size: 4578493
  extract-metadata:
    cmd: python scripts/extract_metadata.py data/eidc_metadata.json data/extracted_metadata.json
    deps:
    - path: data/eidc_metadata.json
      hash: md5
      md5: fb338ea98ce71bf7f002be952b6db0e1
      size: 12275265
    - path: scripts/extract_metadata.py
      hash: md5
      md5: e66f21369c5106eaaad4476612c6fb5e
      size: 1313
    outs:
    - path: data/extracted_metadata.json
      hash: md5
      md5: 9f4fc9cb1e8af8e0f2d1c95b311989fc
      size: 4616342
  chunk-data:
    cmd: python scripts/chunk_data.py  -o data/chunked_data.json  -c 250  -ol 75  data/extracted_metadata.json  data/supporting-docs.json  -m
      0
    deps:
    - path: data/extracted_metadata.json
      hash: md5
      md5: 9f4fc9cb1e8af8e0f2d1c95b311989fc
      size: 4616342
    - path: data/supporting-docs.json
      hash: md5
      md5: 0b14da8f2e73dc8e15747f693c0f70ce
      size: 72383140
    - path: scripts/chunk_data.py
      hash: md5
      md5: 3ad449140b03e1c2904b22a5b401a12e
      size: 2705
    outs:
    - path: data/chunked_data.json
      hash: md5
      md5: b107dfb052c12ea47b04a5176e8bab4a
      size: 176342129
  create-embeddings:
    cmd: python scripts/create_embeddings.py data/chunked_data.json data/embeddings.json
      -m all-MiniLM-L6-v2
    deps:
    - path: data/chunked_data.json
      hash: md5
      md5: b107dfb052c12ea47b04a5176e8bab4a
      size: 176342129
    - path: scripts/create_embeddings.py
      hash: md5
      md5: 87bd2ed6373552bea229c9f3465fd3db
      size: 1594
    outs:
    - path: data/embeddings.json
      hash: md5
      md5: 68a9de7fcf765be8ae2f4d3ff6537228
      size: 3739724900
  upload-to-docstore:
    cmd: python scripts/upload_to_docstore.py  data/embeddings.json  -o data/chroma-data  -em
      all-MiniLM-L6-v2  -c eidc-data
    deps:
    - path: data/embeddings.json
      hash: md5
      md5: 68a9de7fcf765be8ae2f4d3ff6537228
      size: 3739724900
    - path: scripts/upload_to_docstore.py
      hash: md5
      md5: 930456cedd43723c1d643ad90c146952
      size: 2793
    outs:
    - path: data/chroma-data
      hash: md5
      md5: 486d560a81dc951bdd85772996e62f00.dir
      size: 1815042692
      nfiles: 6
  run-rag-pipeline:
    cmd: python scripts/run_rag_pipeline.py  -i data/eidc_rag_testset.csv  -o data/evaluation_data.csv  -ds
      data/chroma-data  -c eidc-data  -m llama3.1  -p data/pipeline.yml
    deps:
    - path: data/chroma-data
      hash: md5
      md5: 486d560a81dc951bdd85772996e62f00.dir
      size: 1815042692
      nfiles: 6
    - path: data/eidc_rag_testset.csv
      hash: md5
      md5: a371d83c5822d256286e80d64d58c3fe
      size: 7524
    - path: scripts/run_rag_pipeline.py
      hash: md5
      md5: 2d6dc886728d4bd46676ecd1882f1fd1
      size: 5838
    outs:
    - path: data/evaluation_data.csv
      hash: md5
      md5: a473732be874c8256f7178ef3f4dc7a9
      size: 9576
    - path: data/pipeline.yml
      hash: md5
      md5: 8e3c4e49d4d97f613e83468d010a96e9
      size: 3440
  generate-testset:
    cmd: head -n 101 data/synthetic-datasets/eidc_rag_test_sample.csv > data/eidc_rag_testset.csv
    outs:
    - path: data/eidc_rag_testset.csv
      hash: md5
      md5: a371d83c5822d256286e80d64d58c3fe
      size: 7524
  fetch-supporting-docs:
    cmd: python scripts/fetch_supporting_docs.py data/eidc_metadata.json data/supporting-docs.json
    deps:
    - path: data/eidc_metadata.json
      hash: md5
      md5: fb338ea98ce71bf7f002be952b6db0e1
      size: 12275265
    - path: scripts/fetch_supporting_docs.py
      hash: md5
      md5: 02b94a2cc7bff711784cbdec3650b618
      size: 1718
    outs:
    - path: data/supporting-docs.json
      hash: md5
      md5: 0b14da8f2e73dc8e15747f693c0f70ce
      size: 72383140
  evaluate:
    cmd: python scripts/evaluate.py data/evaluation_data.csv -m data/metrics.json
      -img data/eval.png
    deps:
    - path: data/evaluation_data.csv
      hash: md5
      md5: a473732be874c8256f7178ef3f4dc7a9
      size: 9576
    - path: scripts/evaluate.py
      hash: md5
      md5: 4154acf8e74c1d8bcd0b0da72af038e0
      size: 2728
    outs:
    - path: data/eval.png
      hash: md5
      md5: 7bfd424fa4c9a3550d6e9605bb2f6af2
      size: 89143
    - path: data/metrics.json
      hash: md5
      md5: f768092fe2696328ff4da565e763e743
      size: 270