Sentinel Documentation

.github/workflows/docs.yml

@@ -73,58 +73,3 @@ jobs:
               body: '⚠️ **Documentation is out of sync with the code!**\n\nPlease run `python docs/generate_docs.py` and commit the updated documentation files.'
             })
 
-  build:
-    runs-on: ubuntu-latest
-    if: github.event_name == 'push' || github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request'
-    needs: [check-docs]
-    # Make this job informational only for PRs - don't block PR
-    continue-on-error: ${{ github.event_name == 'pull_request' }}
-    steps:
-      - name: Checkout repository
-        uses: actions/checkout@v4
-
-      - name: Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: '3.11'
-          cache: 'pip'
-
-      - name: Install Poetry
-        run: |
-          curl -sSL https://install.python-poetry.org | python3 -
-          echo "$HOME/.local/bin" >> $GITHUB_PATH
-
-      - name: Install dependencies
-        run: |
-          poetry install --with docs
-
-      - name: Generate RST files
-        run: |
-          cd ${{ github.workspace }}
-          poetry run python docs/generate_docs.py
-
-      - name: Build documentation
-        run: |
-          cd ${{ github.workspace }}/docs
-          poetry run sphinx-build -b html source build/html
-
-      - name: Setup Pages
-        uses: actions/configure-pages@v4
-
-      - name: Upload artifact
-        uses: actions/upload-pages-artifact@v3
-        with:
-          path: ${{ github.workspace }}/docs/build/html
-
-  deploy:
-    # Only deploy on push to main or workflow_dispatch, skip on PRs
-    if: github.event_name == 'push' && github.ref == 'refs/heads/main' || github.event_name == 'workflow_dispatch'
-    needs: build
-    environment:
-      name: github-pages
-      url: ${{ steps.deployment.outputs.page_url }}
-    runs-on: ubuntu-latest
-    steps:
-      - name: Deploy to GitHub Pages
-        id: deployment
-        uses: actions/deploy-pages@v4

README.md

@@ -109,11 +109,6 @@ saved_config = index.save(
     aws_access_key_id="YOUR_ACCESS_KEY_ID",  # Optional if using environment credentials
     aws_secret_access_key="YOUR_SECRET_ACCESS_KEY"  # Optional if using environment credentials
 )
-
-# If you need to extract the model name from an existing SentenceTransformer instance (best effort):
-from sentinel.embeddings.sbert import extract_model_name_from_sentence_transformer
-model_name = extract_model_name_from_sentence_transformer(model)
-# Returns the model name if it can be determined, or None otherwise
 ```
 
 ## How It Works
@@ -186,6 +181,15 @@ poetry run sphinx-build -b html source build/html
 
 Then open `docs/build/html/index.html` in your browser.
 
+## Examples
+To run the notebook examples
+```bash
+# Install with examples dependencies
+poetry install --with examples
+poetry install --extras=sbert
+poetry run jupyter notebook
+```
+
 ## License
 
 Apache License 2.0

docs/source/sriracha.rst → docs/source/sentinel.rst

@@ -26,10 +26,10 @@ sentinel.score_types
    :undoc-members:
    :show-inheritance:
 
-sentinel.sentinel
+sentinel.sentinel_local_index
 -----------------------------
 
-.. automodule:: sentinel.sentinel
+.. automodule:: sentinel.sentinel_local_index
    :members:
    :undoc-members:
    :show-inheritance:

src/sentinel/embeddings/sbert.py

@@ -46,46 +46,6 @@ def get_sentence_transformer_and_scaling_fn(
 
     return model, None
 
-
-def extract_model_name_from_sentence_transformer(
-    model: SentenceTransformer,
-) -> Optional[str]:
-    """
-    Attempt to extract the model name from a SentenceTransformer instance.
-
-    This is a best-effort function as SentenceTransformer doesn't directly store
-    the original model name. It tries to infer it from the model's configuration.
-
-    Args:
-        model: A SentenceTransformer model instance
-
-    Returns:
-        A string representing the best guess at the model name, or
-        None if it cannot be determined
-    """
-    # Try to get the name from the model's modules
-    if hasattr(model, "modules") and model.modules:
-        # Most SentenceTransformer models use a Transformer as the first module
-        if hasattr(model.modules[0], "auto_model") and hasattr(
-            model.modules[0].auto_model, "config"
-        ):
-            # Try to get the name from the config
-            if hasattr(model.modules[0].auto_model.config, "name_or_path"):
-                return model.modules[0].auto_model.config.name_or_path
-
-    # Try to get from the model's save directory name
-    if hasattr(model, "get_sentence_embedding_dimension") and hasattr(
-        model, "get_config_dict"
-    ):
-        config = model.get_config_dict()
-        if "__path__" in config and config["__path__"]:
-            # Extract the final path component as the model name
-            return os.path.basename(os.path.normpath(config["__path__"]))
-
-    # Return None if we can't determine the name
-    return None
-
-
 def e5_scaling_function(score: float) -> float:
     """
     Scale the similarity score for E5 embeddings.

src/sentinel/sentinel_local_index.py

@@ -93,10 +93,7 @@ def __init__(
             `model, scale_fn = get_sentence_transformer_and_scaling_fn(encoder_model_name_or_path)`
 
             When saving the index, you must provide the exact encoder_model_name_or_path
-            as SentenceTransformer doesn't store the original model name. If needed, you can use
-            `sentinel.embeddings.sbert.extract_model_name_from_sentence_transformer(model)`
-            to attempt extracting the model name, but this is a best-effort function that returns
-            None if it cannot determine the model name.
+            as SentenceTransformer doesn't store the original model name.
 
         Use the class method `load` to load an index from S3 or local storage.
         """
@@ -139,10 +136,7 @@ def save(
             path: Path to save the index to (local directory or S3 URI).
             encoder_model_name_or_path: Name or path of the sentence transformer encoder model used.
                 This must be the exact name used to create the SentenceTransformer as it cannot be
-                reliably extracted from the model instance. If you need to try extracting the model name
-                from an existing instance, use
-                `sentinel.embeddings.sbert.extract_model_name_from_sentence_transformer(model)`,
-                which will return the model name if it can be determined or None otherwise.
+                reliably extracted from the model instance.
             aws_access_key_id: Optional AWS access key ID for S3 access.
             aws_secret_access_key: Optional AWS secret access key for S3 access.