Add contribution guide and contrib/ staging directory.

CONTRIBUTING.md

@@ -3,6 +3,118 @@
 We'd love to accept your patches and contributions to this project. There are
 just a few small guidelines you need to follow.
 
+## A Note on the Library's Current State
+
+DP Synth is **rapidly evolving** and currently in a malleable state. We have not
+yet made a commitment to backwards compatibility, which means **now is the best
+time** to propose API changes, suggest new abstractions, or challenge existing
+design decisions. If you have opinions about how the library should be
+structured, we want to hear them — once we stabilize the API, the bar for
+breaking changes will be much higher.
+
+## Types of Contributions
+
+We welcome contributions in many forms:
+
+### Bug Fixes, Issues, and Documentation
+
+- **Bug reports**: Open an [issue](https://github.com/google/dpsynth/issues)
+  describing what went wrong and how to reproduce it.
+- **Bug fixes**: Small, focused PRs that fix a specific issue are always
+  appreciated.
+- **Documentation**: Improving docstrings, adding examples, or clarifying
+  existing docs.
+
+### Privacy Hardening
+
+Finding and reporting privacy violations is especially valuable. Contributions
+that harden existing implementations — for example, integrating with an
+[OpenDP](https://opendp.org/) verification backend as an opt-in check — are
+welcome.
+
+### New Features and Mechanisms
+
+We accept new mechanisms at every level of the stack:
+
+- **Tabular data mechanisms**: New algorithms for discrete synthetic data
+  generation (e.g., alternatives to AIM, SWIFT, MST).
+- **Text generation mechanisms**: New approaches for differentially private text
+  synthesis or fine-tuning.
+- **Primitives**: New building blocks such as partition selection algorithms or
+  numerical discretization strategies.
+- **Higher-level components**: Mechanisms that build on top of synthetic data
+  generation — for example, adaptive insights from unstructured text, or
+  modality-specific pipelines that decompose complex problems into
+  subproblems the library already handles well.
+
+### API Design Proposals
+
+The library is at an early stage where API design feedback is highly impactful.
+If you think a core abstraction could be improved, open an issue to discuss it
+before writing code. We'd rather get the design right now than maintain
+backwards-incompatible wrappers later.
+
+### Contributing to Upstream Libraries
+
+DP Synth is co-developed with several companion libraries, and improvements to
+any of them directly benefit DP Synth:
+
+- [**MBI**](https://github.com/ryan112358/mbi): Graphical model estimation
+  and inference engine used by the tabular mechanisms.
+- [**dp_accounting**](https://github.com/google/differential-privacy/tree/main/python/dp_accounting):
+  Privacy accounting and budget composition.
+- [**jax_privacy**](https://github.com/google-deepmind/jax_privacy):
+  Differentially private training primitives used by the text generation
+  module.
+
+Contributions to these libraries — bug fixes, performance improvements, new
+features — can be surfaced into DP Synth once they land upstream.
+
+## The `contrib/` Directory
+
+If you're proposing a **new mechanism** or a substantial new feature that
+doesn't fit neatly into the existing module structure, place it in the
+[`contrib/`](contrib/) directory. This is a staging area for experimental or
+community-contributed code that hasn't yet been integrated into the core
+library.
+
+Once a contribution in `contrib/` has landed, matured, and proven useful, we'll
+work with you to migrate it into the appropriate core module.
+
+If your contribution is a bug fix, documentation improvement, or a change to
+an existing module, submit it directly to the relevant file — `contrib/` is
+only for new standalone additions.
+
+## Acceptance Criteria
+
+When reviewing contributions, we primarily look for two things:
+
+### 1. Simplicity
+
+Simpler contributions with smaller diffs are more likely to be accepted. For
+new mechanisms in particular:
+
+- **Aim for a single file** with at most ~500 lines of code.
+- You don't need to use the same internal helper functions that existing
+  mechanisms use. If your approach is novel, a self-contained implementation
+  is perfectly fine.
+- However, your mechanism **should conform to the same API contract** as
+  existing mechanisms (e.g., implementing `DPMechanism`, accepting the same
+  calibration/configuration interface). The exception is mechanisms designed
+  for a new data modality (e.g., relational data), where a new API surface
+  may be necessary.
+
+### 2. No New Heavy Dependencies
+
+DP Synth deliberately keeps its dependency footprint small:
+
+- **Do not introduce PyTorch, TensorFlow, or other heavyweight
+  dependencies.**
+- If you need numerical computing, use **JAX**. If you need neural network
+  layers, use **Flax**.
+- If your contribution requires a new dependency, discuss it in the issue
+  tracker first.
+
 ## Contributor License Agreement
 
 Contributions to this project must be accompanied by a Contributor License
@@ -16,20 +128,18 @@ You generally only need to submit a CLA once, so if you've already submitted one
 (even if it was for a different project), you probably don't need to do it
 again.
 
-## Bug Reports & Feature Requests
-
-Before opening a pull request, consider
-[filing an issue](https://github.com/google/dpsynth/issues) to discuss the
-change. This helps us coordinate efforts and provide guidance on the best
-approach.
-
 ## Code Reviews
 
 All submissions, including submissions by project members, require review. We
 use GitHub pull requests for this purpose. Consult
 [GitHub Help](https://help.github.com/articles/about-pull-requests/) for more
 information on using pull requests.
 
+Before opening a pull request, consider
+[filing an issue](https://github.com/google/dpsynth/issues) to discuss the
+change. This helps us coordinate efforts and provide guidance on the best
+approach.
+
 ## Community Guidelines
 
 This project follows

dpsynth/contrib/__init__.py

@@ -0,0 +1,20 @@
+# Copyright 2026 Google LLC
+#
+# 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.
+
+"""Community-contributed mechanisms and utilities.
+
+This package is a staging area for experimental or community-contributed code.
+New mechanisms that don't yet fit into the core library structure should be
+placed here. See CONTRIBUTING.md for details.
+"""