dataforge-07 0.1.0__tar.gz

dataforge_07-0.1.0/LICENSE

@@ -0,0 +1,176 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS

dataforge_07-0.1.0/MANIFEST.in

@@ -0,0 +1,33 @@
+prune data_quality_env
+prune tests
+prune docs
+prune scripts
+prune training
+prune playground
+prune playground-model
+prune dataforge-mcp
+prune benchmark_results
+prune datasets
+prune eval
+prune logs
+prune build
+prune dist
+prune .github
+prune .hf-space-repo
+prune .hf-space-stage
+prune .hf-space-stage-plan
+exclude analyze_trajectory.py
+exclude benchmark.py
+exclude client.py
+exclude compat.py
+exclude generate_datasets.py
+exclude heuristic_baseline.py
+exclude inference.py
+exclude models.py
+exclude random_baseline.py
+exclude run_baseline.py
+exclude test_env.py
+exclude verify_all_fields.py
+exclude verify_nuclear.py
+exclude verify_score_range.py
+global-exclude __pycache__ *.py[cod] *.ipynb .pytest_cache .mypy_cache .ruff_cache .hypothesis

dataforge_07-0.1.0/PKG-INFO

@@ -0,0 +1,436 @@
+Metadata-Version: 2.4
+Name: dataforge_07
+Version: 0.1.0
+Summary: DataForge: CLI-first data-quality detection and reversible repair for tabular data.
+License-Expression: Apache-2.0
+Project-URL: Homepage, https://github.com/Aegis15/dataforge
+Project-URL: Repository, https://github.com/Aegis15/dataforge
+Project-URL: Documentation, https://dataforge.praneshrajan15.workers.dev/playground
+Keywords: data-quality,ai-agent,llm,rl,smt,dbt
+Classifier: Development Status :: 3 - Alpha
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Requires-Python: <3.13,>=3.11
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: pydantic>=2.7
+Requires-Dist: typer<0.25,>=0.24
+Requires-Dist: rich>=13.7
+Requires-Dist: textual<9,>=8.2
+Requires-Dist: z3-solver>=4.13
+Requires-Dist: pyyaml>=6.0
+Requires-Dist: pandas>=2.2
+Requires-Dist: httpx>=0.27
+Requires-Dist: python-dotenv>=1.0
+Provides-Extra: bench
+Requires-Dist: pandas>=2.2; extra == "bench"
+Requires-Dist: httpx>=0.27; extra == "bench"
+Requires-Dist: tenacity>=8.3; extra == "bench"
+Requires-Dist: python-dotenv>=1.0; extra == "bench"
+Requires-Dist: pyarrow>=16.0; extra == "bench"
+Provides-Extra: causal
+Requires-Dist: pandas>=2.2; extra == "causal"
+Requires-Dist: numpy>=1.26; extra == "causal"
+Requires-Dist: networkx>=3.3; extra == "causal"
+Requires-Dist: causal-learn>=0.1.4; extra == "causal"
+Requires-Dist: hyppo>=0.5.2; extra == "causal"
+Requires-Dist: scipy>=1.13; extra == "causal"
+Provides-Extra: dev
+Requires-Dist: pytest>=9.0.3; extra == "dev"
+Requires-Dist: pytest-cov>=5.0; extra == "dev"
+Requires-Dist: pytest-benchmark>=4.0; extra == "dev"
+Requires-Dist: pytest-xdist>=3.6; extra == "dev"
+Requires-Dist: hypothesis>=6.100; extra == "dev"
+Requires-Dist: mutmut>=3.5; extra == "dev"
+Requires-Dist: build>=1.2; extra == "dev"
+Requires-Dist: pip-audit<3,>=2.10; extra == "dev"
+Requires-Dist: cyclonedx-bom<8,>=7.3; extra == "dev"
+Requires-Dist: cryptography>=46.0.7; extra == "dev"
+Requires-Dist: idna>=3.15; extra == "dev"
+Requires-Dist: pip>=26.1.1; extra == "dev"
+Requires-Dist: urllib3>=2.7; extra == "dev"
+Requires-Dist: ruff>=0.11; extra == "dev"
+Requires-Dist: mypy>=1.10; extra == "dev"
+Requires-Dist: pandas-stubs>=2.2; extra == "dev"
+Requires-Dist: types-PyYAML; extra == "dev"
+Requires-Dist: huggingface_hub==1.13.0; extra == "dev"
+Requires-Dist: httpx>=0.27; extra == "dev"
+Requires-Dist: tenacity>=8.3; extra == "dev"
+Requires-Dist: python-dotenv>=1.0; extra == "dev"
+Requires-Dist: pyarrow>=16.0; extra == "dev"
+Requires-Dist: networkx>=3.3; extra == "dev"
+Requires-Dist: causal-learn>=0.1.4; extra == "dev"
+Requires-Dist: hyppo>=0.5.2; extra == "dev"
+Requires-Dist: scipy>=1.13; extra == "dev"
+Requires-Dist: sqlglot>=25.0; extra == "dev"
+Requires-Dist: duckdb>=1.0; extra == "dev"
+Provides-Extra: train
+Requires-Dist: trl==1.4.0; extra == "train"
+Requires-Dist: transformers==5.7.0; extra == "train"
+Requires-Dist: accelerate==1.13.0; extra == "train"
+Requires-Dist: peft==0.19.1; extra == "train"
+Requires-Dist: bitsandbytes==0.49.2; extra == "train"
+Requires-Dist: datasets==4.8.5; extra == "train"
+Requires-Dist: huggingface_hub==1.13.0; extra == "train"
+Requires-Dist: pyyaml==6.0.3; extra == "train"
+Requires-Dist: pandas==2.3.3; extra == "train"
+Requires-Dist: tensorboard==2.20.0; extra == "train"
+Provides-Extra: eval
+Requires-Dist: matplotlib>=3.9; extra == "eval"
+Requires-Dist: seaborn>=0.13; extra == "eval"
+Provides-Extra: providers
+Requires-Dist: httpx>=0.27; extra == "providers"
+Requires-Dist: tenacity>=8.3; extra == "providers"
+Requires-Dist: python-dotenv>=1.0; extra == "providers"
+Provides-Extra: pandas
+Requires-Dist: pandas>=2.2; extra == "pandas"
+Provides-Extra: playground
+Requires-Dist: pandas>=2.2; extra == "playground"
+Requires-Dist: fastapi>=0.136.1; extra == "playground"
+Requires-Dist: starlette<2,>=1.0.1; extra == "playground"
+Requires-Dist: uvicorn[standard]>=0.35; extra == "playground"
+Requires-Dist: python-multipart>=0.0.27; extra == "playground"
+Requires-Dist: slowapi>=0.1.9; extra == "playground"
+Provides-Extra: openenv
+Requires-Dist: pandas>=2.2; extra == "openenv"
+Requires-Dist: openenv-core[core]>=0.2.2; extra == "openenv"
+Requires-Dist: authlib!=1.7.0,>=1.7.1; extra == "openenv"
+Requires-Dist: cryptography>=46.0.7; extra == "openenv"
+Requires-Dist: duckdb>=1.0; extra == "openenv"
+Requires-Dist: sqlglot>=25.0; extra == "openenv"
+Requires-Dist: scipy>=1.13; extra == "openenv"
+Requires-Dist: networkx>=3.3; extra == "openenv"
+Requires-Dist: causal-learn>=0.1.4; extra == "openenv"
+Requires-Dist: hyppo>=0.5.2; extra == "openenv"
+Provides-Extra: all
+Requires-Dist: dataforge_07[bench,causal,dev,eval,openenv,pandas,playground,providers,train]; extra == "all"
+Dynamic: license-file
+
+# DataForge
+
+DataForge is a CLI-first data-quality repair toolkit for tabular data. It
+detects common CSV issues, proposes deterministic repairs, checks proposed
+changes through safety and verification gates, and records applied changes in a
+reversible transaction log.
+
+The final public product name is DataForge. The PyPI/TestPyPI distribution
+family is `dataforge_07*` because the unqualified `dataforge` project name is
+occupied by unrelated packages. Installing `dataforge_07` still provides the
+`dataforge` import namespace and `dataforge` CLI. `dataforge15` is only a
+temporary staging alias retained for local compatibility.
+
+The current repository is an alpha implementation. It also contains the
+OpenEnv-compatible training environment, the SFT warmup workflow, a local MCP
+server package, and playground/demo sources. Warehouse integrations and
+production model-quality claims remain future work.
+
+Before any public release, review `THREAT_MODEL.md` and `docs/docs/release.md`.
+They define the security, supply-chain, and evidence gates that separate the
+current alpha from the full original DataForge vision.
+
+## Current Status
+
+Shipped in the current worktree:
+
+- `dataforge profile`, `dataforge repair`, `dataforge revert`,
+  `dataforge watch`, `dataforge audit`, and `dataforge bench`
+- Three detector families: `type_mismatch`, `decimal_shift`, `fd_violation`
+- Reviewable schema inference in `profile --json`, including inferred column
+  types, domains, regex candidates, uniqueness, and FD candidates
+- Pending constraint review artifacts via `profile --constraints-out`, which
+  can feed repair only after individual candidates are marked accepted
+- Matching deterministic repairers wired through SafetyFilter -> SMTVerifier
+- Backend-neutral `PatchPlan` and `TableStore` contracts for CSV, DuckDB, and
+  dry-run-only cloud warehouse boundaries
+- Reversible hash-chained transaction journals with immutable source snapshots
+- Public backend repair engine at `dataforge.engine.repair`
+- Real-world benchmark harness for Hospital, Flights, and Beers
+- OpenEnv-compatible HTTP environment with eight typed actions, including
+  read-only `ROOT_CAUSE`
+- Causal root-cause analyzer for cascading data-quality errors
+- Standalone `dataforge-mcp` package exposing DataForge tools over MCP
+- Week 9 SFT oracle trajectory workflow, readiness gate, Kaggle notebook, and
+  release verifier
+- Separate Gradio model-demo Space source for the published 0.5B SFT smoke
+  checkpoint
+
+Not shipped yet:
+
+- published `dataforge_07`, `dataforge_07_mcp`, `dataforge_07_evals`,
+  `dataforge_07_dbt`, and `dataforge_07_agent_patterns` packages
+- committed production verification for the Cloudflare Workers playground
+- warehouse-native or external adapter packages
+- credentialed Snowflake, BigQuery, or Databricks apply/revert conformance
+- design-partner, pilot-user, or customer validation evidence is not yet claimed
+- A production-quality trained model family
+- Autonomous repair in the playground or model demo
+
+## Quickstart
+
+```bash
+python -m pip install -e ".[dev]"
+dataforge profile fixtures/hospital_10rows.csv --schema fixtures/hospital_schema.yaml
+dataforge profile fixtures/hospital_10rows.csv --constraints-out constraints.json
+dataforge constraints review constraints.json
+dataforge repair fixtures/hospital_10rows.csv --schema fixtures/hospital_schema.yaml --dry-run
+dataforge repair fixtures/hospital_10rows.csv --constraints constraints.json --dry-run
+dataforge watch fixtures/hospital_10rows.csv --schema fixtures/hospital_schema.yaml --once --json
+dataforge bench --methods random,heuristic --datasets hospital,flights,beers --seeds 3 --seed-list 0,1,2
+```
+
+`dataforge15` remains a temporary staging compatibility alias, but public docs
+and release evidence must use `dataforge_07` for PyPI distribution identity and
+`dataforge` for the installed CLI/import identity.
+
+To apply repairs, use `--apply`. Applied repairs write a transaction journal and
+source snapshot before mutating the CSV, so they can be reverted:
+
+```bash
+dataforge repair path/to/file.csv --schema path/to/schema.yaml --apply
+dataforge audit <txn-id>
+dataforge revert <txn-id>
+dataforge revert <txn-id> --search-root path/to --json
+```
+
+Warehouse targets use `warehouse://` URIs and always emit a `patch_plan_v1`
+contract before any mutation. DuckDB is the local conformance backend; cloud
+warehouse adapters are dry-run-only boundaries until credentialed apply,
+audit, and rollback suites are enabled:
+
+```bash
+dataforge repair "warehouse://duckdb?database=dev.duckdb&relation=main.model&row_id=id" --dry-run --json
+dataforge repair "warehouse://snowflake?relation=PUBLIC.MODEL&row_id=ID" --dry-run --json
+```
+
+DuckDB `--apply` requires a stable row identity, records the patch plan in the
+transaction journal, and can be reverted through the same `audit` and `revert`
+commands. Snowflake, BigQuery, and Databricks apply are intentionally refused
+until their conformance gates prove reversible transactions.
+
+New transaction logs are local tamper-evident hash chains. `dataforge audit`
+verifies the chain head, event order, replayability, and revert prerequisites;
+legacy v1 logs remain replayable but are reported as unverified because they do
+not contain event hashes.
+
+## Week 9 SFT Warmup
+
+The current SFT workflow builds split-safe `expert_v1` trajectory records from
+dirty/clean CSV diffs. Exact repairs in the primary dataset are labeled
+`oracle_from_clean_diff`, not inferred from Groq, Cerebras, or Gemini teacher
+guesses. Clean train chunks are retained as `finish` examples so the model
+learns when no repair is justified.
+
+```powershell
+$env:HF_TOKEN="..."
+.\.venv\Scripts\python.exe scripts\data\build_oracle_sft_trajectories.py
+.\.venv\Scripts\python.exe scripts\data\validate_sft_readiness.py
+```
+
+This writes local ignored JSONL at `data/sft_traj/expert_v1.jsonl` and an
+auditable row split at `data/sft_traj/split_manifest.json`. Push the dataset
+bundle only after the readiness gate passes:
+
+```powershell
+$env:HF_TOKEN="..."
+.\.venv\Scripts\python.exe scripts\data\build_oracle_sft_trajectories.py --push-to-hub --hf-dataset-repo Praneshrajan15/dataforge-sft-trajectories
+```
+
+The current public smoke checkpoint is
+`Praneshrajan15/DataForge-0.5B-SFT`, with trajectories at
+`Praneshrajan15/dataforge-sft-trajectories`. It proves the dataset, Kaggle
+training, merge, evaluation, and Hub upload path; it is not a production
+model-quality claim. Verify release artifacts before citing them:
+
+```powershell
+.\.venv\Scripts\python.exe scripts\model\verify_sft_release.py --output eval\results\sft_release_v0_smoke.json
+.\.venv\Scripts\python.exe scripts\model\verify_sft_release.py --min-dataset-records 272 --require-sha-metrics --output eval\results\sft_release_contract_v2_20260515.json
+```
+
+## Week 12 GRPO Path
+
+The repository now contains a gated GRPO post-training path for free-tier
+experiments:
+
+- `training/configs/grpo_05b.yaml` targets `DataForge-0.5B-SFT` -> `DataForge-0.5B-GRPO`.
+- `training/configs/grpo_15b.yaml` requires a verified `DataForge-1.5B-SFT`
+  prerequisite before attempting `DataForge-1.5B-GRPO`.
+- `training/rewards/dataforge_reward.py` scores completions locally through the
+  `repair_contract_v1` exact-repair contract.
+- `training/kaggle/grpo_kaggle.ipynb` blocks Hub upload unless GRPO beats SFT
+  by at least 3 absolute F1 points on `DataForge-Bench-light-verified`.
+
+No GRPO checkpoint is described as a quality milestone in this README until
+`scripts/model/verify_grpo_release.py` produces committed verification
+evidence. Refresh benchmark tables only from generated JSON:
+
+After GRPO eval evidence exists:
+
+```powershell
+.\.venv\Scripts\python.exe scripts\bench\refresh_benchmark_table.py --skip-agent-run --trained-model-json eval\results\grpo_model_comparison.json
+```
+
+## MCP Server
+
+The nested `dataforge-mcp/` source directory builds the standalone
+`dataforge_07_mcp` distribution. It is not published yet, so install it from
+source while release ownership is pending:
+
+```bash
+cd dataforge-mcp
+python -m pip install -e ".[dev]"
+dataforge-mcp serve
+```
+
+Tools: `dataforge_profile`, `dataforge_detect_errors`,
+`dataforge_verify_fix`, `dataforge_apply_repairs`, and `dataforge_revert`.
+The default transport is stdio. MCP reads and writes are sandboxed to configured
+allowed roots; dry-run works by default, while apply requires `--enable-apply`.
+Streamable HTTP is available for local experiments.
+
+The monorepo `packages/` directory contains the side-package release sources
+for `dataforge_07_evals`, `dataforge_07_dbt`, and
+`dataforge_07_agent_patterns`.
+
+## Playground And Model Demo
+
+- `playground/api/` is the API backend for the CSV playground. Public Space
+  deployments use `dataforge-playground`.
+- `playground/web/` is the static browser UI deployed through Cloudflare
+  Workers Static Assets. Its primary workflow is `POST /api/analyze`: upload a
+  CSV, review categorical risk and pending inferred constraints, inspect
+  verified dry-run repairs and non-repairs, then export a receipt with the
+  local CLI apply/audit/revert command shape.
+- The current verified public playground URL is
+  `https://dataforge.praneshrajan15.workers.dev/playground`, backed by
+  `https://Praneshrajan15-dataforge-playground.hf.space`.
+- That Workers URL is the production playground surface for the full original
+  vision; this is the release URL.
+- `playground-model/` is a separate Gradio Space demo for the published
+  `DataForge-0.5B-SFT` smoke checkpoint. It accepts small CSV snippets and is
+  intentionally limited to demo use.
+
+The playground does not persist uploaded files, does not use browser storage,
+does not mutate data in the hosted flow, and does not call an LLM unless a
+backend provider key is explicitly configured.
+
+## Benchmark Results
+
+<!-- BENCH:START -->
+Generated from `eval/results/agent_comparison.json` (schema `dataforge_benchmark_run_v2`, seeds `0, 1, 2`, git `dbd1bed0a03c`, dirty `true`).
+
+| Method | Precision | Recall | F1 | Avg Steps | Quota Units | GPU Hours |
+| --- | --- | --- | --- | --- | --- | --- |
+| heuristic | 0.3167 | 0.3025 | 0.2772 | 374.33 | 0.0000 | 0.0000 |
+| random | 0.0038 | 0.0003 | 0.0005 | 150.33 | 0.0000 | 0.0000 |
+
+See `BENCHMARK_REPORT.md` for per-dataset tables, error bars, and citation-only SOTA rows.
+
+Dataset bytes are pinned to BigDaMa/raha revision `7be1334b8c7bbdac3f47ef514fb3e1e8c5fc181c` for hospital, flights, beers; dirty/clean SHA-256s are recorded in the JSON metadata.
+<!-- BENCH:END -->
+
+## Local Setup
+
+```bash
+make setup
+make lint
+make type
+make test
+make backend-gate
+make release-gate
+```
+
+Verification works on Linux, macOS, and Windows with Git Bash available for GNU
+Make recipes. Python support is `>=3.11,<3.13`.
+
+`profile --constraints-out` writes a strict `constraint_review_v1` JSON artifact.
+Every inferred candidate starts as `pending`; repair ignores pending and
+rejected candidates. In v1, only accepted `column_type`, `domain_bound`, and
+`functional_dependency` candidates affect repair. Accepted regex and uniqueness
+candidates remain review evidence until verifier support is added. Use
+`dataforge constraints review constraints.json` for the Textual review UI, or
+use deterministic CI flags such as `--accept cnd-... --no-tui --json`.
+
+`make backend-gate` is the release-quality backend check: lint, format, strict
+mypy, root tests, MCP tests, README truth, benchmark truth, OpenAPI snapshot
+drift, secret scan, dependency audit availability, SBOM generation
+availability, and package build availability for both `dataforge_07` and
+`dataforge_07_mcp`. The gate covers the core `dataforge_07` distribution and
+release surfaces; the historical
+`data_quality_env` namespace remains source-tree regression coverage, not part
+of the `dataforge` wheel or source distribution.
+
+Before release, run `scripts/ci/backend_gate.py --require-optional` so
+dependency audit, SBOM generation, and package builds are hard failures rather
+than availability checks.
+
+Release doctor scopes:
+
+```bash
+dataforge release doctor --core --json
+dataforge release doctor --maintainer-deploy --json
+dataforge release gate --json
+dataforge release full-vision --json
+```
+
+`--core` is the default OSS release check. `--maintainer-deploy` additionally
+checks maintainer-specific Hugging Face, Kaggle OAuth plus clean-config Kaggle
+CLI execution, and Cloudflare state.
+`release gate` is the authoritative fresh-user proof: it builds the
+distribution, audits wheel contents, creates a dependency wheelhouse, installs
+with `pip --no-index --find-links`, then runs profile, repair dry-run, apply,
+constraint review, audit, revert, and post-revert audit from outside the source
+checkout.
+
+Configure pending trusted publishers for `dataforge_07` on TestPyPI and PyPI
+before tagging. The real PyPI workflow refuses pre-release metadata and should
+only run after trusted publishing, attestations, and fresh-install evidence are
+verified. `dataforge release full-vision --json` is expected to fail until PyPI
+publication evidence, dbt-duckdb proof, not yet met design-partner evidence,
+and model-family evidence are real.
+
+Windows setup:
+
+```powershell
+winget install -e --id Python.Python.3.12
+winget install -e --id ezwinports.make
+py -3.12 -m venv .venv
+.\.venv\Scripts\Activate.ps1
+python -m pip install -e ".[all]"
+make lint && make type && make test
+```
+
+## Environment Variables
+
+Provider keys belong in a root `.env` file, which is gitignored and loaded with
+`python-dotenv` where needed.
+
+- `GROQ_API_KEY`
+- `GEMINI_API_KEY`
+- `CEREBRAS_API_KEY`
+- `OPENROUTER_API_KEY`
+- `HF_TOKEN`
+
+## When DataForge Is The Wrong Tool
+
+Do not use DataForge for streaming data, very large warehouse tables, regulated
+workflows where every fix must be human-authored, strict low-latency SLAs, or
+teams already well served by maintained Great Expectations/dbt suites. DataForge
+is currently best suited to local CSV profiling, repair experiments, benchmark
+runs, and training/evaluation research.
+
+## Repository Docs
+
+- [.cursor/rules/dataforge.md](.cursor/rules/dataforge.md) - always-applied contribution rules
+- [ARCHITECTURE.md](ARCHITECTURE.md) - current system architecture and dependencies
+- [DECISIONS.md](DECISIONS.md) - technical decision log
+- [CONTRIBUTING.md](CONTRIBUTING.md) - workflow and code standards
+- [CLAUDE.md](CLAUDE.md) - living gotcha log for agent sessions
+- [CURSOR_MASTER.md](CURSOR_MASTER.md) - context and prompt pack
+- [META_CONTEXT.md](META_CONTEXT.md) - project meta-context
+- [FILE_STRUCTURE.md](FILE_STRUCTURE.md) - current and planned directory map
+- [SECURITY.md](SECURITY.md) - vulnerability reporting policy
+- [specs/SPEC_TEMPLATE.md](specs/SPEC_TEMPLATE.md) - template for new module specs
+
+## License
+
+Apache-2.0. See [LICENSE](LICENSE).