sembr 0.2.0__tar.gz → 0.2.2__tar.gz

{sembr-0.2.0/sembr.egg-info → sembr-0.2.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sembr
-Version: 0.2.0
+Version: 0.2.2
 Summary: A semantic linebreaker powered by transformers
 Author: admk
 License-Expression: MIT
@@ -76,6 +76,8 @@ to accelerate inference.
 
 ### Usage
 
+#### Command Line Interface
+
 To use SemBr,
 run the following command in your terminal:
 ```shell
@@ -110,7 +112,7 @@ Additionally,
 you can specify the following options
 to customize the behavior of SemBr:
 
-* `-m <model_name>`:
+* `-m <model_name>`, `--model-name <model_name>`:
   The name of the Hugging Face model to use.
   - The default is
     [`admko/sembr2023-bert-small`][sembr-bert-small].
@@ -119,7 +121,7 @@ to customize the behavior of SemBr:
     and then specify the path to the model directory,
     or prepend `TRANSFORMERS_OFFLINE=1` to the command
     to use the cached model.
-* `-l`:
+* `-l`, `--listen`:
   Serves the SemBr API on a local server.
   - Each instance of `sembr` run
     will detect if the API is accessible,
@@ -127,13 +129,58 @@ to customize the behavior of SemBr:
   - This option is useful
     to avoid the time taken to initialize the model
     by keeping it in memory in a separate process.
-* `-p <port>`:
+* `-p <port>`, `--port <port>`:
   The port to serve the SemBr API on.
   - The default is `8384`.
-* `-s <ip>`:
+* `-s <ip>`, `--server <ip>`:
   The IP address to serve the SemBr API on.
   - The default is `127.0.0.1`.
+* `-b <int>`, `--batch_size <int>`:
+  The number of lines to process in a batch.
+  Default is `8`.
+* `-d <int>`, `--overlap-divisor <int>`:
+  The overlap divisor for tiled inference.
+  Default is `8`.
+* `-f <func>`, `--predict-func <func>`:
+  The prediction function to use.
+  Options are `argmax`, `logit_adjustment`, `greedy_line_breaks`.
+  Default is `argmax`.
+* `-t <int>`, `--tokens-per-line <int>`:
+  Maximum tokens per line for greedy line breaking.
+  This is only effective
+  when using the `greedy_line_breaks` prediction function.
+* `--bits <4|8>`:
+  Quantization bits for model weights (4 or 8).
+  Requires CUDA. Not supported on MPS.
+* `--dtype <dtype>`:
+  Data type for model weights (e.g. `float16`, `bfloat16`).
+  Default is `float32`.
+* `--mcp`:
+  Start MCP server mode instead of processing text.
+
+
+#### MCP Server
+
+Alternatively,
+you can run `sembr` as an [MCP server][mcp].
+Simply add the following configuration
+to your MCP server configuration:
+```json
+"mcpServers": {
+  "sembr": {
+    "type": "stdio",
+    "command": "uvx",
+    "args": [
+      "sembr",
+      "--mcp"
+    ],
+  }
+}
+```
 
+The server also supports the formatting options described above.
+It will expose a `wrap_text` tool
+for the MCP client to use.
 
 ## What are Semantic Line Breaks?
 
@@ -316,13 +363,10 @@ to save best models.
     - [ ] Inference queue.
     - [ ] Daemon with model unloading.
   - Editor integration:
-    - [ ] NeoVim plugin.
-    - [ ] VSCode extension.
-  - [ ] Use the [Hugging Face API][hfapi] for inference.
-    It is free to use but has a rate limit,
-    and also does not return logit values,
-    so no additional algorithms
-    can be used to improve the predictions.
+    - [x] ~~NeoVim plugin.~~
+    - [x] ~~VSCode extension.~~
+    - [x] MCP server.
+  - [x] ~~Use the [Hugging Face API][hfapi] for inference.~~
 - Accuracy:
   - Some lines are too short or too long:
     - [x] Long lines can be penalized greedily
@@ -335,8 +379,8 @@ to save best models.
   - [ ] Performance and accuracy benchmarking,
         and comparisons with related works.
 - Performance:
-  - [ ] Improve inference speed.
-  - [ ] Reduce memory usage.
+  - [x] Improve inference speed.
+  - [x] Reduce memory usage.
 
 
 ## Related Projects and References
@@ -360,6 +404,7 @@ Semantic line breaking:
 
 [pypi]: https://pypi.org/project/sembr
 [uv]: https://github.com/astral-sh/uv
+[mcp]: https://modelcontextprotocol.io/overview
 
 [sembr]: https://sembr.org
 [semlf]: https://rhodesmill.org/brandon/2012/one-sentence-per-line

{sembr-0.2.0 → sembr-0.2.2}/README.md

@@ -50,6 +50,8 @@ to accelerate inference.
 
 ### Usage
 
+#### Command Line Interface
+
 To use SemBr,
 run the following command in your terminal:
 ```shell
@@ -84,7 +86,7 @@ Additionally,
 you can specify the following options
 to customize the behavior of SemBr:
 
-* `-m <model_name>`:
+* `-m <model_name>`, `--model-name <model_name>`:
   The name of the Hugging Face model to use.
   - The default is
     [`admko/sembr2023-bert-small`][sembr-bert-small].
@@ -93,7 +95,7 @@ to customize the behavior of SemBr:
     and then specify the path to the model directory,
     or prepend `TRANSFORMERS_OFFLINE=1` to the command
     to use the cached model.
-* `-l`:
+* `-l`, `--listen`:
   Serves the SemBr API on a local server.
   - Each instance of `sembr` run
     will detect if the API is accessible,
@@ -101,13 +103,58 @@ to customize the behavior of SemBr:
   - This option is useful
     to avoid the time taken to initialize the model
     by keeping it in memory in a separate process.
-* `-p <port>`:
+* `-p <port>`, `--port <port>`:
   The port to serve the SemBr API on.
   - The default is `8384`.
-* `-s <ip>`:
+* `-s <ip>`, `--server <ip>`:
   The IP address to serve the SemBr API on.
   - The default is `127.0.0.1`.
+* `-b <int>`, `--batch_size <int>`:
+  The number of lines to process in a batch.
+  Default is `8`.
+* `-d <int>`, `--overlap-divisor <int>`:
+  The overlap divisor for tiled inference.
+  Default is `8`.
+* `-f <func>`, `--predict-func <func>`:
+  The prediction function to use.
+  Options are `argmax`, `logit_adjustment`, `greedy_line_breaks`.
+  Default is `argmax`.
+* `-t <int>`, `--tokens-per-line <int>`:
+  Maximum tokens per line for greedy line breaking.
+  This is only effective
+  when using the `greedy_line_breaks` prediction function.
+* `--bits <4|8>`:
+  Quantization bits for model weights (4 or 8).
+  Requires CUDA. Not supported on MPS.
+* `--dtype <dtype>`:
+  Data type for model weights (e.g. `float16`, `bfloat16`).
+  Default is `float32`.
+* `--mcp`:
+  Start MCP server mode instead of processing text.
+
+
+#### MCP Server
+
+Alternatively,
+you can run `sembr` as an [MCP server][mcp].
+Simply add the following configuration
+to your MCP server configuration:
+```json
+"mcpServers": {
+  "sembr": {
+    "type": "stdio",
+    "command": "uvx",
+    "args": [
+      "sembr",
+      "--mcp"
+    ],
+  }
+}
+```
 
+The server also supports the formatting options described above.
+It will expose a `wrap_text` tool
+for the MCP client to use.
 
 ## What are Semantic Line Breaks?
 
@@ -290,13 +337,10 @@ to save best models.
     - [ ] Inference queue.
     - [ ] Daemon with model unloading.
   - Editor integration:
-    - [ ] NeoVim plugin.
-    - [ ] VSCode extension.
-  - [ ] Use the [Hugging Face API][hfapi] for inference.
-    It is free to use but has a rate limit,
-    and also does not return logit values,
-    so no additional algorithms
-    can be used to improve the predictions.
+    - [x] ~~NeoVim plugin.~~
+    - [x] ~~VSCode extension.~~
+    - [x] MCP server.
+  - [x] ~~Use the [Hugging Face API][hfapi] for inference.~~
 - Accuracy:
   - Some lines are too short or too long:
     - [x] Long lines can be penalized greedily
@@ -309,8 +353,8 @@ to save best models.
   - [ ] Performance and accuracy benchmarking,
         and comparisons with related works.
 - Performance:
-  - [ ] Improve inference speed.
-  - [ ] Reduce memory usage.
+  - [x] Improve inference speed.
+  - [x] Reduce memory usage.
 
 
 ## Related Projects and References
@@ -334,6 +378,7 @@ Semantic line breaking:
 
 [pypi]: https://pypi.org/project/sembr
 [uv]: https://github.com/astral-sh/uv
+[mcp]: https://modelcontextprotocol.io/overview
 
 [sembr]: https://sembr.org
 [semlf]: https://rhodesmill.org/brandon/2012/one-sentence-per-line

{sembr-0.2.0 → sembr-0.2.2}/sembr/__init__.py

@@ -1,5 +1,5 @@
 __toolname__ = __name__
-__version__ = "0.2.0"
+__version__ = "0.2.2"
 __author__ = "admk"
 __license__ = "MIT"
 __url__ = f"https://github.com/admk/{__name__}"

{sembr-0.2.0 → sembr-0.2.2}/sembr/cli.py

@@ -89,6 +89,8 @@ def start_server(port, tokenizer, model, processor, wrap_kwargs=None):
         text = form['text']
         kwargs = dict(wrap_kwargs or {})
         for k, v in form.items():
+            if k == 'text':
+                continue
             if k in ['batch_size', 'tokens_per_line', 'overlap_divisor']:
                 v = int(v)
             kwargs[k] = v
@@ -157,7 +159,7 @@ def wrap_kwargs(args):
     }
 
 
-def main():
+def main() -> int:
     parser = cli_parser()
     args = parser.parse_args()
     if args.debug:
@@ -167,13 +169,21 @@ def main():
         debugpy.wait_for_client()
     if args.mcp:
         from .mcp import mcp
+        unsupported = ['input_file', 'output_file', 'listen']
+        for arg_name in unsupported:
+            if getattr(args, arg_name) in [None, False]:
+                continue
+            message = f'--{arg_name} is not supported in MCP mode.'
+            print(message, file=sys.stderr)
+            return 1
         mcp.run()
-        return
+        return 0
     kwargs = wrap_kwargs(args)
     if args.listen:
         tokenizer, model, processor = init(
             args.model_name, args.bits, args.dtype)
-        return start_server(args.port, tokenizer, model, processor, kwargs)
+        start_server(args.port, tokenizer, model, processor, kwargs)
+        return 0
     if args.input_file is not None:
         with open(args.input_file, 'r', encoding='utf-8') as f:
             text = f.read()
@@ -181,8 +191,8 @@ def main():
         text = sys.stdin.read()
     else:
         parser.print_help()
-        print('\nNo input file or stdin text provided.')
-        return
+        print('\nNo input file or stdin text provided.', file=sys.stderr)
+        return 1
     if check_server(args.server, args.port):
         result = rewrap_on_server(text, args.server, args.port, kwargs)
     else:
@@ -192,10 +202,11 @@ def main():
         result = sembr(text, tokenizer, model, processor, **kwargs)
     if args.output_file is None:
         print(result)
-        return
+        return 0
     with open(args.output_file, 'w', encoding='utf-8') as f:
         f.write(result)
+    return 0
 
 
 if __name__ == '__main__':
-    main()
+    sys.exit(main())

{sembr-0.2.0 → sembr-0.2.2}/sembr/mcp.py

@@ -60,22 +60,5 @@ def wrap_text(
         structured_content={"success": True, "output": wrapped_text})
 
 
-@mcp.tool(
-    description="Apply semantic line breaks to file",
-    tags=["sembr", "semantic linebreak", "format", "file"],
-)
-def process_file(
-    file_path: Annotated[str, Field(description="File path to process")],
-) -> ToolResult:
-    try:
-        with open(file_path, 'r', encoding='utf-8') as f:
-            text = f.read()
-    except Exception as e:
-        return ToolResult(
-            content=[TextContent(type="text", text=f"Error reading file: {file_path}")],
-            structured_content={"success": False, "error": str(e)})
-    return wrap_text(text)
-
-
 if __name__ == "__main__":
     mcp.run()

{sembr-0.2.0 → sembr-0.2.2/sembr.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sembr
-Version: 0.2.0
+Version: 0.2.2
 Summary: A semantic linebreaker powered by transformers
 Author: admk
 License-Expression: MIT
@@ -76,6 +76,8 @@ to accelerate inference.
 
 ### Usage
 
+#### Command Line Interface
+
 To use SemBr,
 run the following command in your terminal:
 ```shell
@@ -110,7 +112,7 @@ Additionally,
 you can specify the following options
 to customize the behavior of SemBr:
 
-* `-m <model_name>`:
+* `-m <model_name>`, `--model-name <model_name>`:
   The name of the Hugging Face model to use.
   - The default is
     [`admko/sembr2023-bert-small`][sembr-bert-small].
@@ -119,7 +121,7 @@ to customize the behavior of SemBr:
     and then specify the path to the model directory,
     or prepend `TRANSFORMERS_OFFLINE=1` to the command
     to use the cached model.
-* `-l`:
+* `-l`, `--listen`:
   Serves the SemBr API on a local server.
   - Each instance of `sembr` run
     will detect if the API is accessible,
@@ -127,13 +129,58 @@ to customize the behavior of SemBr:
   - This option is useful
     to avoid the time taken to initialize the model
     by keeping it in memory in a separate process.
-* `-p <port>`:
+* `-p <port>`, `--port <port>`:
   The port to serve the SemBr API on.
   - The default is `8384`.
-* `-s <ip>`:
+* `-s <ip>`, `--server <ip>`:
   The IP address to serve the SemBr API on.
   - The default is `127.0.0.1`.
+* `-b <int>`, `--batch_size <int>`:
+  The number of lines to process in a batch.
+  Default is `8`.
+* `-d <int>`, `--overlap-divisor <int>`:
+  The overlap divisor for tiled inference.
+  Default is `8`.
+* `-f <func>`, `--predict-func <func>`:
+  The prediction function to use.
+  Options are `argmax`, `logit_adjustment`, `greedy_line_breaks`.
+  Default is `argmax`.
+* `-t <int>`, `--tokens-per-line <int>`:
+  Maximum tokens per line for greedy line breaking.
+  This is only effective
+  when using the `greedy_line_breaks` prediction function.
+* `--bits <4|8>`:
+  Quantization bits for model weights (4 or 8).
+  Requires CUDA. Not supported on MPS.
+* `--dtype <dtype>`:
+  Data type for model weights (e.g. `float16`, `bfloat16`).
+  Default is `float32`.
+* `--mcp`:
+  Start MCP server mode instead of processing text.
+
+
+#### MCP Server
+
+Alternatively,
+you can run `sembr` as an [MCP server][mcp].
+Simply add the following configuration
+to your MCP server configuration:
+```json
+"mcpServers": {
+  "sembr": {
+    "type": "stdio",
+    "command": "uvx",
+    "args": [
+      "sembr",
+      "--mcp"
+    ],
+  }
+}
+```
 
+The server also supports the formatting options described above.
+It will expose a `wrap_text` tool
+for the MCP client to use.
 
 ## What are Semantic Line Breaks?
 
@@ -316,13 +363,10 @@ to save best models.
     - [ ] Inference queue.
     - [ ] Daemon with model unloading.
   - Editor integration:
-    - [ ] NeoVim plugin.
-    - [ ] VSCode extension.
-  - [ ] Use the [Hugging Face API][hfapi] for inference.
-    It is free to use but has a rate limit,
-    and also does not return logit values,
-    so no additional algorithms
-    can be used to improve the predictions.
+    - [x] ~~NeoVim plugin.~~
+    - [x] ~~VSCode extension.~~
+    - [x] MCP server.
+  - [x] ~~Use the [Hugging Face API][hfapi] for inference.~~
 - Accuracy:
   - Some lines are too short or too long:
     - [x] Long lines can be penalized greedily
@@ -335,8 +379,8 @@ to save best models.
   - [ ] Performance and accuracy benchmarking,
         and comparisons with related works.
 - Performance:
-  - [ ] Improve inference speed.
-  - [ ] Reduce memory usage.
+  - [x] Improve inference speed.
+  - [x] Reduce memory usage.
 
 
 ## Related Projects and References
@@ -360,6 +404,7 @@ Semantic line breaking:
 
 [pypi]: https://pypi.org/project/sembr
 [uv]: https://github.com/astral-sh/uv
+[mcp]: https://modelcontextprotocol.io/overview
 
 [sembr]: https://sembr.org
 [semlf]: https://rhodesmill.org/brandon/2012/one-sentence-per-line