aluha501 commited on
Commit
eeb7471
·
verified ·
1 Parent(s): d6d02a6

Publish one-argument RAG access documentation

Browse files
docs/releases/v0.7.0-multisource-merged.md CHANGED
@@ -5,7 +5,17 @@ Release status: published experimental derivative.
5
  Published to the existing Hugging Face dataset:
6
  [`aluha501/community-qa-lab`](https://huggingface.co/datasets/aluha501/community-qa-lab)
7
 
8
- Hub commit: [`2eb94ed72d6dc0985a85437940ec40423e7f6703`](https://huggingface.co/datasets/aluha501/community-qa-lab/commit/2eb94ed72d6dc0985a85437940ec40423e7f6703)
 
 
 
 
 
 
 
 
 
 
9
 
10
  ## Published configurations
11
 
@@ -18,27 +28,28 @@ Hub commit: [`2eb94ed72d6dc0985a85437940ec40423e7f6703`](https://huggingface.co/
18
 
19
  ### Primary use: community content and RAG
20
 
21
- `corpus_rag_multisource_merged_v0.1` is the primary v0.7 content corpus. Each
22
- row is a licensed question/post document with its source and provenance
23
- metadata. It is intended for ordinary NLP, semantic search and RAG indexing:
 
24
 
25
  ```python
26
  from datasets import load_dataset
27
 
28
- rag = load_dataset(
29
- "aluha501/community-qa-lab",
30
- "corpus_rag_multisource_merged_v0.1",
31
- )
32
  ```
33
 
34
  Use `document_text` as the text input. `accepted`, marked-answer status and
35
  other interaction fields are not factual ground-truth annotations; they are
36
  retained only as optional metadata for filtering and analysis.
37
 
38
- The new files are namespaced under `v0.7.0-multisource-merged/` in the Hub
39
- repository. Existing `default`, `technical_qa`, and audit configurations were
40
- preserved unchanged. The uploaded Parquet and release artifacts were read back
41
- and verified byte-for-byte at the Hub commit.
 
 
 
42
 
43
  The DPO and listwise views use accepted/marked-answer interaction signals as
44
  weak labels; they do not represent factual truth. The RAG view contains only
 
5
  Published to the existing Hugging Face dataset:
6
  [`aluha501/community-qa-lab`](https://huggingface.co/datasets/aluha501/community-qa-lab)
7
 
8
+ The root `default` files are the merged RAG corpus, so ordinary users can load
9
+ the release with one dataset identifier and no second config argument:
10
+
11
+ ```python
12
+ from datasets import load_dataset
13
+
14
+ rag = load_dataset("aluha501/community-qa-lab")
15
+ ```
16
+
17
+ The publication commit is updated whenever the Hub card or root projection is
18
+ refreshed; the final commit is recorded in the release manifest.
19
 
20
  ## Published configurations
21
 
 
28
 
29
  ### Primary use: community content and RAG
30
 
31
+ `corpus_rag_multisource_merged_v0.1` is the source view used to populate the
32
+ root `default` files. Each row is a licensed question/post document with its
33
+ source and provenance metadata. It is intended for ordinary NLP, semantic
34
+ search and RAG indexing:
35
 
36
  ```python
37
  from datasets import load_dataset
38
 
39
+ rag = load_dataset("aluha501/community-qa-lab")
 
 
 
40
  ```
41
 
42
  Use `document_text` as the text input. `accepted`, marked-answer status and
43
  other interaction fields are not factual ground-truth annotations; they are
44
  retained only as optional metadata for filtering and analysis.
45
 
46
+ The derived view files remain namespaced under
47
+ `v0.7.0-multisource-merged/` in the Hub repository for reproducibility. The
48
+ root `default` projection now points to the merged RAG corpus. The historical
49
+ v0.4.0 files remain recoverable under `releases/v0.4.0/`; the named
50
+ `legacy_default_v0.4.0` configuration is provided for explicit compatibility.
51
+ The uploaded Parquet and release artifacts are read back and verified
52
+ byte-for-byte at the final Hub commit.
53
 
54
  The DPO and listwise views use accepted/marked-answer interaction signals as
55
  weak labels; they do not represent factual truth. The RAG view contains only