Create README.md

README.md

@@ -0,0 +1,68 @@
+---
+task_categories:
+- text-generation
+tags:
+- code
+- elixir
+- the-stack
+- pretraining
+size_categories:
+- 100K<n<1M
+
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: data/*.parquet
+---
+
+# Curated Elixir Stack v3
+
+A curated, syntax-verified Elixir source code dataset derived from `HuggingFaceCode/stack-v3-full`. This was also an exploration for me to learn how datasets are curated.
+
+## Dataset Summary
+* **Source:** Derived directly from the public Elixir partition of [The Stack v3](https://huggingface.co/datasets/HuggingFaceCode/stack-v3-full) by BigCode.
+* **File Count:** ~564,000 files.
+* **Target Size Bounds:** Filtered to files between 1.5 KB and 25 KB post-cleaning.
+* **Syntax Parsing:** Validated using `tree-sitter-elixir` (AST parse validation to drop truncated code and merge conflicts).
+* **Architectural Tags:** Tagged via pattern matching across common Elixir domains (`test`, `phoenix`, `liveview`, `genserver`, `supervisor`, `ecto`, `documented`, `typespec`, `macro`).
+* **Metadata Preserved:** Original `content_id` and originating `repo_ids` are preserved in the Parquet schema to maintain source attribution and traceability back to original repositories.
+
+---
+
+## Important Notice: Automated Filtering & Sanitization Limitations
+
+**Please read before using this dataset:**
+
+1. **No Guarantee of PII Removal:** Personal Identifiable Information (such as email addresses, IP addresses, or potential secret keys) was processed strictly via automated regular expressions and replaced with sentinel tokens (`<EMAIL>`, `<IP_ADDRESS>`, `<SECRET_TOKEN>`, `<PRIVATE_KEY>`). **These heuristics are non-exhaustive and are not guaranteed to catch all sensitive or personal information.**
+2. **No Guarantee of Complete License/Header Stripping:** Automated comment parsing was applied to remove common top-of-file copyright preambles to reduce repetitive legal boilerplate. Some non-standard, malformed, or mid-file license blocks may remain.
+3. **No Code Endorsement or Security Auditing:** Files in this dataset originate from public third-party repositories. The maintainers of this curated derivative have not audited the code for security vulnerabilities, malware, or runtime fitness.
+
+---
+
+## Disclaimer of Warranties & Limitation of Liability
+
+THIS DATASET IS PROVIDED "AS IS" AND "AS AVAILABLE," WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, TITLE, AND NON-INFRINGEMENT. 
+
+IN NO EVENT SHALL THE CURATORS OR CONTRIBUTORS BE LIABLE FOR ANY CLAIM, DAMAGES, OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT, OR OTHERWISE, ARISING FROM, OUT OF, OR IN CONNECTION WITH THE DATASET, THE USE OR REDISTRIBUTION OF THE DATASET, OR ANY UPSTREAM DATA DERIVED FROM THE STACK V3. 
+
+Downstream users, developers, and model trainers are solely responsible for ensuring that their use of this data complies with applicable data privacy laws, intellectual property rights, and third-party repository licenses.
+
+---
+
+## Provenance, Attribution & Upstream Licensing
+
+* This dataset is an independent derivative of [The Stack v3](https://huggingface.co/datasets/HuggingFaceCode/stack-v3-full) created by BigCode.
+* Individual code records retain the respective open-source licenses of their source repositories as indexed by BigCode. Refer to the `repo_ids` column for source repository links.
+* Use of this dataset must adhere to ethical AI norms and acceptable use guidelines as defined by BigCode.
+
+---
+
+## Opt-Out and Takedown Requests
+
+This dataset fully respects the developer opt-out framework established by BigCode. 
+
+If you are a copyright holder or developer and wish to have your code or repository removed from this curated derivative:
+1. Open an issue in the **Community** tab of this repository with our repository URL and/or the relevant `content_id`.
+
+Upon notification, the requested records will be purged from the source files in the subsequent release.