diff --git a/README.md b/README.md
index 05065a7..a16d0a1 100644
--- a/README.md
+++ b/README.md
@@ -12,13 +12,13 @@
Your trained model, schema and verified structure. One file.
- 
- 
- 
+
+
+
- 
+
@@ -142,7 +142,7 @@ or training history.
Custom types fail at export. Sklearn and CatBoost own their class labels;
XGBoost can bind declared labels to a native copy. ONNX has no visualization
structure, and all-sklearn Pipeline structure is null. The
-[adapter contracts](./docs/format-adapters.md) define the exact support limits.
+[adapter contracts](https://docs.mlbundle.eshkra.dev/format/adapters/) define the exact support limits.
## Some notes
@@ -153,7 +153,7 @@ independent review; do not accept bundles from untrusted parties.
The reader bounds ZIP/JSON data, verifies integrity and uses closed native loaders.
Hashes detect corruption, not deliberate tampering. Native parsers and inference
still need process isolation and application-owned resource limits.
-Read [SECURITY.md](./SECURITY.md) for guarantees, limits and private reporting.
+Read [SECURITY.md](https://docs.mlbundle.eshkra.dev/security/) for guarantees, limits and private reporting.
Training rows stay local by default, but feature names, categories, labels and
learned structure can reveal information. Schemas and environment records remain
@@ -162,17 +162,18 @@ the library's scope.
## Documentation
-The [documentation site](./docs/) includes installation, adapter guides, record
-prediction and API reference. The normative contracts remain in this repository:
+The [documentation site](https://docs.mlbundle.eshkra.dev) includes installation, adapter guides, record
+prediction and API reference. It renders the normative contracts from this
+repository unchanged:
-- [Bundle format and verification](./docs/format.md)
-- [Schema, mappings and explicit samples](./docs/format-schema.md)
-- [Supported adapters and structure](./docs/format-adapters.md)
-- [Sklearn state and trusted types](./docs/format-state.md)
-- [Composite pipelines](./docs/format-pipeline.md)
-- [ONNX layouts and bounds](./docs/format-onnx.md)
+- [Bundle format and verification](https://docs.mlbundle.eshkra.dev/format/)
+- [Schema, mappings and explicit samples](https://docs.mlbundle.eshkra.dev/format/schema/)
+- [Supported adapters and structure](https://docs.mlbundle.eshkra.dev/format/adapters/)
+- [Sklearn state and trusted types](https://docs.mlbundle.eshkra.dev/format/state/)
+- [Composite pipelines](https://docs.mlbundle.eshkra.dev/format/pipeline/)
+- [ONNX layouts and bounds](https://docs.mlbundle.eshkra.dev/format/onnx/)
-The [playground](./playground/) provides editable examples and inspection of the
+The [playground](https://playground.mlbundle.eshkra.dev) provides editable examples and inspection of the
generated `.mlbundle` file. It runs scikit-learn examples in the browser and
inspects uploaded files without loading their models.
@@ -192,8 +193,8 @@ uv build
```
Tests use the installed package and enforce 80% branch-inclusive coverage.
-CI validates the built wheel on Windows and Linux. Read [AGENTS.md](./AGENTS.md)
-for contribution rules and [publishing](./docs/publishing.md) for the owner-approved
+CI validates the built wheel on Windows and Linux. Read [AGENTS.md](https://github.com/UlloaSP/mlbundle/blob/main/AGENTS.md)
+for contribution rules and [publishing](https://github.com/UlloaSP/mlbundle/blob/main/docs/publishing.md) for the automatic
release workflow.
The landing, docs and playground use [Vite+](https://viteplus.dev/guide/) and a
@@ -218,7 +219,7 @@ After all checks pass on `main`, it deploys the three tested site builds to
Cloudflare. Manual **CI Pipeline** runs on `main` use the same gates and can
redeploy without publishing Python. A change to `project.version` on `main`
automatically starts the PyPI workflow after its version and branch preflight.
-See [website deployment](./docs/deployment.md) and [publishing](./docs/publishing.md).
+See [website deployment](https://github.com/UlloaSP/mlbundle/blob/main/docs/deployment.md) and [publishing](https://github.com/UlloaSP/mlbundle/blob/main/docs/publishing.md).
## License
diff --git a/tests/test_readme.py b/tests/test_readme.py
index 1f38482..011f512 100644
--- a/tests/test_readme.py
+++ b/tests/test_readme.py
@@ -1,4 +1,4 @@
-"""Execute the README quick start and its mlbundle CLI commands."""
+"""Execute the README quick start and check the links PyPI renders."""
import re
import shlex
@@ -13,6 +13,17 @@
r"```(\w+)\n(.*?)\n```", README.read_text(encoding="utf-8"), re.DOTALL
)
PYTHON = [code for language, code in BLOCKS if language == "python"]
+LINKS = re.findall(
+ r"\]\(([^)\s]+)\)|(?:href|src|srcset)=\"([^\"]+)\"",
+ README.read_text(encoding="utf-8"),
+)
+
+
+def test_links_are_absolute():
+ # PyPI renders the README without the repository, so relative links break.
+ targets = [markdown or html for markdown, html in LINKS]
+ assert targets
+ assert [url for url in targets if not url.startswith("https://")] == []
@pytest.mark.parametrize("code", PYTHON)