raftaar 0.1.0__tar.gz

raftaar-0.1.0/CHANGELOG.md

@@ -0,0 +1,78 @@
+# Changelog
+
+## 0.1.0
+
+Named **Raftaar** — رفتار, *behaviour*, in Persian, Turkish and Urdu.
+
+Behaviour cloning is the paradigm this tool serves, so the name says what the
+library is about in one word to anyone who knows the field, and pairs with
+[daftar](https://pypi.org/project/daftar/) (دفتر, *ledger*) as a second package
+by the same author.
+
+Renamed twice before release, from DemoScope via KineLens. "Demo" reads as a
+demonstration *of a product* rather than demonstration *data*.
+
+Deliberately not named `trajlint`, `trajscope` or `trajaudit`, all of which were
+free: `trajlens` already exists on PyPI as a structural linter for LeRobot
+datasets, and those names would read as derivative of a tool operating one layer
+below this one.
+
+`rename.py` is kept in the repository until first publication, so the name stays
+cheap to change. It detects the current package name from `src/` rather than
+hardcoding it, so it works however many times it is run.
+
+## Packaging work (pre-release)
+
+Raftaar existed as a working prototype: five detectors, a synthetic
+environment with injected faults, a validation harness, a CLI and a browser
+demo. This release turns it into something installable with `pip install
+raftaar` without changing what the detectors do.
+
+### Packaging
+
+- `src/` layout, `pyproject.toml`, Apache 2.0 licence, `raftaar` console
+  script replacing `python -m raftaar.cli`.
+- **The dependency surface is now three packages**: numpy, scipy,
+  scikit-learn. The prototype's `requirements.txt` also pulled matplotlib,
+  Flask and gunicorn — a web server and a plotting stack, for a CPU dataset
+  audit. Those are now the `[plot]` and `[web]` extras.
+- `matplotlib` is imported lazily inside `report.plot()`. `raftaar scan`
+  writes its JSON and markdown on a machine without it and prints what it
+  skipped, instead of failing after the scan you just waited for.
+- The Flask demo, its templates and the Dockerfile moved to `web/`, outside the
+  installable package. They are the Cloud Run demo, not the library.
+
+### Tests
+
+29 tests, where there were none. The detector tests are the point: every fault
+is injected deliberately into the synthetic environment, so each detector is
+scored against ground truth, in both directions — sensitivity (an injected
+fault must be caught) and specificity (clean data must stay quiet).
+
+Also pinned: the CLI round-trip, the public API, that importing `raftaar`
+pulls neither matplotlib nor Flask nor torch, and that the synthetic
+environment is deterministic under a fixed seed — which it has to be, since it
+is the ground truth everything else is scored against.
+
+### A documented limitation
+
+Writing the specificity tests surfaced something the README had overstated.
+"Clean data → no warnings" holds at 120 episodes, but below roughly 90
+`ACTION_INCONSISTENCY` fires at *warning* severity on clean data. Near-identical
+states genuinely do disagree when there are few episodes per region of the state
+space, so the detector is under-powered rather than wrong — but a user scanning
+40 episodes sees a warning on data that is fine.
+
+Both behaviours are now pinned by tests, and the README says so. The test for
+the spurious warning carries a note that if it ever starts failing because the
+warning no longer appears, that is an improvement and the test should be
+deleted.
+
+### Optional provenance
+
+`raftaar.provenance` records what produced a scan — dataset content hash,
+Raftaar version, every detector's headline number — using
+[daftar](https://pypi.org/project/daftar/) when installed. Entirely optional:
+`available()` is checked before anything is imported, `track()` yields `None`
+when daftar is absent, and `tracked_scan()` returns exactly what
+`scan(load_dataset(path))` returns either way.

raftaar-0.1.0/LICENSE

@@ -0,0 +1,202 @@
+
+                                 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
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

raftaar-0.1.0/MANIFEST.in

@@ -0,0 +1,7 @@
+include LICENSE
+include README.md
+include CHANGELOG.md
+recursive-include tests *.py
+recursive-include examples *.md *.py
+recursive-include web *.py *.html *.md
+include web/Dockerfile

raftaar-0.1.0/PKG-INFO

@@ -0,0 +1,242 @@
+Metadata-Version: 2.4
+Name: raftaar
+Version: 0.1.0
+Summary: Know what your robot demonstrations will teach, before you train. Predicts which policy class will fail on a dataset, and where.
+Author: Aslan Satary Dizaji
+License-Expression: Apache-2.0
+Project-URL: Homepage, https://github.com/aslansd/raftaar
+Project-URL: Repository, https://github.com/aslansd/raftaar
+Project-URL: Issues, https://github.com/aslansd/raftaar/issues
+Keywords: robotics,behaviour-cloning,imitation-learning,dataset-quality,demonstrations,lerobot,data-curation
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Science/Research
+Classifier: Programming Language :: Python :: 3
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: numpy>=1.26
+Requires-Dist: scipy>=1.11
+Requires-Dist: scikit-learn>=1.4
+Provides-Extra: plot
+Requires-Dist: matplotlib>=3.8; extra == "plot"
+Provides-Extra: validate
+Requires-Dist: matplotlib>=3.8; extra == "validate"
+Provides-Extra: web
+Requires-Dist: flask>=3.0; extra == "web"
+Requires-Dist: gunicorn>=22.0; extra == "web"
+Provides-Extra: provenance
+Requires-Dist: daftar>=0.8; extra == "provenance"
+Provides-Extra: dev
+Requires-Dist: pytest>=8; extra == "dev"
+Requires-Dist: matplotlib>=3.8; extra == "dev"
+Provides-Extra: all
+Requires-Dist: matplotlib>=3.8; extra == "all"
+Requires-Dist: flask>=3.0; extra == "all"
+Requires-Dist: gunicorn>=22.0; extra == "all"
+Requires-Dist: daftar>=0.8; extra == "all"
+Dynamic: license-file
+
+# Raftaar
+
+**رفتار** — *behaviour.* The same word in Persian, Turkish and Urdu.
+
+**Know what your demonstrations will teach, before you train.**
+
+Raftaar reads a robot demonstration dataset and predicts *which policy class
+will fail on it, and where* — without training anything. It runs on a CPU in
+minutes.
+
+Structural validators ("is this file corrupt, are the camera intrinsics valid")
+are commoditising fast. Raftaar works one level up, on the **distribution**:
+the geometry of the demonstrations themselves, and what a learning algorithm
+will converge to when trained on them.
+
+```bash
+pip install raftaar
+```
+
+Python 3.10+. Apache 2.0. Three dependencies — numpy, scipy, scikit-learn. No
+GPU, no torch, no downloads.
+
+---
+
+## The failure that matters
+
+A dataset containing two valid but incompatible strategies — half the operators
+reach around an obstacle on the left, half on the right — will train a
+regression policy to do neither. The policy converges to the average, the
+average is a path nobody demonstrated, and **the training loss goes down the
+entire time**.
+
+Validation loss does not catch it either, because the validation split has the
+same structure as the training split. The same is true of coverage holes, of
+silent recalibrations mid-collection, and of actuator channels that stopped
+moving. Every one of these is detectable before training. None of them is
+currently detected.
+
+### Where this sits
+
+The robot-data tooling layer is filling in from the bottom. Structural linters
+that check schemas, timestamps and decodability already exist and are improving
+fast — `trajlens` is one, and it audits the public LeRobot Hub today. That layer
+is commoditising and Raftaar does not compete with it.
+
+Raftaar works on the layer above: not "is this file valid" but **"what will a
+policy learn from this distribution, and which policy class will fail"**. The
+two are complementary — run a structural linter first, then this.
+
+---
+
+## Quickstart
+
+```bash
+# 1. A dataset with known, deliberately injected faults
+raftaar synth field_data --episodes 120 \
+    --faults bimodal_detour coverage_hole demonstrator_jitter camera_shard dead_actuator
+
+# 2. Diagnose it — no training
+raftaar scan datasets/field_data --out reports/field_data
+
+# 3. Check the diagnosis was right — train policies and roll them out
+raftaar validate datasets/field_data --trials 40
+```
+
+From Python:
+
+```python
+from raftaar import load_dataset, scan
+
+report = scan(load_dataset("datasets/field_data"))
+for finding in report["findings"]:
+    print(finding["severity"], finding["id"], finding["where"])
+    print("   ", finding["what"])
+```
+
+---
+
+## The five detectors
+
+| Detector | Question it answers |
+|---|---|
+| `AVERAGING_HAZARD` | Do demonstrations contain competing strategies whose *average* is a trajectory nobody performed? |
+| `COVERAGE_GAP` | Which task conditions were never demonstrated? (sample-size aware — sparsity is not a hole) |
+| `ACTION_INCONSISTENCY` | Do different episodes, in near-identical states, disagree on what to do? |
+| `REPRESENTATION_SHARD` | Was part of this dataset recorded under a different calibration or lighting? |
+| `IDLE_CHANNEL` | Which action channels carry no signal? (sharpened with `--reference`) |
+
+`AVERAGING_HAZARD` is the differentiated one. It clusters demonstrations into
+strategies, measures how far their average sits from the nearest real strategy
+in units of within-strategy spread, and **suppresses splits that the task
+conditions themselves explain** — paths to objects on the left and on the right
+*should* differ, and that is the policy's job, not a defect.
+
+---
+
+## Why the naive version doesn't work
+
+Three things had to be right before any of this detected anything:
+
+1. **Strategy signatures.** Comparing raw trajectories fails: variation in
+   *where the object was* swamps variation in *how the operator got there*.
+   Subtracting the straight line between a phase's own endpoints isolates
+   strategy from condition.
+2. **A confound guard.** A mode split that the initial conditions predict is
+   suppressed, or every well-collected dataset would be flagged.
+3. **Elbow selection, not BIC minimum.** BIC falls monotonically as components
+   are added, because real demonstration clusters always have sub-structure.
+
+---
+
+## Does it actually predict anything?
+
+`raftaar validate` trains a unimodal regression policy (stands in for
+ACT-style behaviour cloning) and a mode-conditioned policy (stands in for
+diffusion/flow policies) on the same data and rolls both out. 120 episodes,
+40 trials.
+
+| Dataset | Raftaar verdict | Unimodal BC | Mode-conditioned BC |
+|---|---|---|---|
+| `clean` | no warnings | **100% success, 0% collision** | 78% success, 15% collision |
+| `bimodal_only` | AVERAGING_HAZARD | 75% success, **25% collision** | **100% success, 0% collision** |
+| `field_data` | 5 findings | 20% success, **80% collision** | 88% success, 5% collision |
+
+The prediction runs both ways. On clean data the tool stays quiet and the simple
+policy wins outright — reaching for a multimodal policy there makes things
+*worse*. On bimodal data the tool warns, and the simple policy drives into the
+obstacle. That symmetry is what makes the output actionable rather than
+decorative.
+
+### A caveat on small datasets
+
+"No warnings" on clean data holds at 120 episodes. Below roughly 90,
+`ACTION_INCONSISTENCY` fires at *warning* severity on clean data, because
+near-identical states genuinely do disagree when there are few episodes per
+region of the state space. The detector is measuring something real; it is
+under-powered, not wrong. Both behaviours are pinned by tests in
+`tests/test_detectors.py` so the boundary is documented rather than folklore.
+
+---
+
+## Optional extras
+
+The core install is deliberately small. Everything else is opt-in:
+
+```bash
+pip install "raftaar[plot]"        # matplotlib, for the diagnostic figure
+pip install "raftaar[validate]"    # the rollout study
+pip install "raftaar[provenance]"  # record what produced each scan
+pip install "raftaar[web]"         # the browser demo in web/
+pip install "raftaar[all]"
+```
+
+`raftaar scan` writes its markdown report and JSON with no plotting stack
+installed, and tells you what it skipped. A dataset audit should not require
+matplotlib to run.
+
+### Provenance
+
+A scan is a measurement and `validate` is an experiment, so both depend on
+seeds, library versions and the exact dataset that went in. With
+[daftar](https://pypi.org/project/daftar/) installed:
+
+```python
+from raftaar.provenance import tracked_scan
+
+report = tracked_scan("datasets/field_data", label="field-data-audit")
+```
+
+This records the dataset's content hash, the Raftaar version, every detector's
+headline number and all findings, so two scans of the same dataset can be diffed
+rather than eyeballed. Without daftar it behaves exactly like
+`scan(load_dataset(path))` — the integration is a no-op, not an error.
+
+---
+
+## Tests
+
+```bash
+pip install -e ".[dev]"
+pytest -q          # 29 tests
+```
+
+The detector tests are the ones that matter. Every fault is *injected
+deliberately* into the synthetic environment, so each detector is scored against
+ground truth rather than opinion — and both directions are tested:
+
+- **sensitivity** — a dataset built with a fault must raise that fault's finding
+- **specificity** — clean data must stay quiet, because a detector that fires on
+  everything is worse than no detector
+
+---
+
+## Status
+
+Research prototype, honestly labelled. The synthetic environment exists so the
+detectors can be scored against ground truth. The next milestone is a
+`LeRobotDataset` adapter and a run across the public dataset hub — at which
+point the claims above get tested against data nobody generated to order.
+
+## Licence
+
+Apache 2.0.