sanjeevafk commited on
Commit
ce4d537
·
verified ·
1 Parent(s): 1215b69

chore: clear old shards before full restore

Browse files
Files changed (4) hide show
  1. LICENSE_SUMMARY.md +0 -13
  2. README.md +0 -59
  3. SOURCES_MANIFEST.yaml +0 -7
  4. train-00000.parquet +0 -3
LICENSE_SUMMARY.md DELETED
@@ -1,13 +0,0 @@
1
- # License Summary
2
-
3
- This dataset is mixed-license. Downstream consumers must respect the upstream license attached to each chunk.
4
-
5
- | Upstream License | Documents |
6
- | --- | ---: |
7
- | unknown | 628 |
8
-
9
- ## Policy
10
-
11
- - No dataset-wide MIT claim is made for upstream source content.
12
- - Per-chunk `upstream_license`, `source`, and `source_url` fields are authoritative.
13
- - Unknown licenses remain in the export and should be filtered for conservative enterprise use.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
README.md DELETED
@@ -1,59 +0,0 @@
1
- ---
2
- language:
3
- - en
4
- tags:
5
- - rag
6
- - retrieval
7
- - technical-docs
8
- - programming
9
- size_categories:
10
- - 100K<n<1M
11
- ---
12
-
13
- # depthapi_technical_corpus
14
-
15
- ## Summary
16
-
17
- `depthapi_technical_corpus` is a mixed-license retrieval corpus built from technical documentation, books, engineering writeups, and repository-derived reference material. It is intended for open-source RAG systems, embedding benchmarks, reranker evaluation, and enterprise retrieval experiments.
18
-
19
- ## Ingestion Process
20
-
21
- The corpus was assembled from multiple upstream source types and normalized through a reproducible local pipeline backed by Supabase.
22
-
23
- - `Scrapling` was used for live technical documentation crawling and structured HTML extraction from documentation websites.
24
- - [`D4Vinci/Scrapling`](https://github.com/D4Vinci/Scrapling) was used for live technical documentation crawling and structured HTML extraction from documentation websites.
25
- - [`opendataloader-project/opendataloader-pdf`](https://github.com/opendataloader-project/opendataloader-pdf) was used for PDF extraction when ingesting book-like and document-style technical sources into normalized markdown/text blocks.
26
- - Source material was then normalized, rechunked deterministically, deduplicated, validated, and exported to parquet for Hugging Face datasets compatibility.
27
-
28
- ## Schema
29
-
30
- Each chunk includes:
31
-
32
- - `chunk_id`
33
- - `source`
34
- - `source_url`
35
- - `upstream_license`
36
- - `document_id`
37
- - `chunk_index`
38
- - `retrieved_at`
39
- - `chunker_version`
40
- - `content_hash`
41
- - `content`
42
-
43
- ## Usage
44
-
45
- ```python
46
- from datasets import load_dataset
47
-
48
- dataset = load_dataset("sanjeevafk/depthapi_technical_corpus", split="train", streaming=True)
49
- for row in dataset.take(3):
50
- print(row["chunk_id"], row["source"], row["upstream_license"])
51
- ```
52
-
53
- ## Retrieval Benchmark Example
54
-
55
- Use `data/research_corpus/benchmarks/queries.jsonl` with your retriever and score against `qrels.jsonl`.
56
-
57
- ## Licensing
58
-
59
- This is a mixed-license dataset. Downstream users must inspect `upstream_license` and `SOURCES_MANIFEST.yaml` before redistribution or commercial use.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
SOURCES_MANIFEST.yaml DELETED
@@ -1,7 +0,0 @@
1
- version: 1
2
- sources:
3
- - source: Full-Stack FastAPI Template
4
- source_url: unknown
5
- upstream_license: unknown
6
- retrieved_at: unknown
7
- documents: 628
 
 
 
 
 
 
 
 
train-00000.parquet DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:4555e1268e27f8acfe0c6384024dcb6498a634f89356c810981c8c3fe2ae8bfd
3
- size 120949