llguidance 1.4.0__cp39-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl → 1.5.0__cp39-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl

llguidance/_lib.pyi

@@ -570,6 +570,26 @@ class LLExecutor:
         State rollback is performed to maintain matcher consistency.
         """
 
+    def consume_token_par(
+        self,
+        matchers: List[Tuple[LLMatcher, TokenId]],
+    ) -> List[bool]:
+        """
+        Consume a single token for each matcher in parallel.
+
+        Args:
+            matchers: List of tuples containing:
+                - LLMatcher: The matcher object
+                - TokenId: The token ID to consume
+
+        Returns:
+            List[bool]: Success/failure for each matcher (in order).
+
+        Note:
+            Matchers that fail (return False) are left in an error state,
+            consistent with the behavior of consume_token on a single matcher.
+        """
+
 class JsonCompileOptions(TypedDict, total=False):
     # defaults to ","
     item_separator: Optional[str]

llguidance/mlx.py

@@ -8,7 +8,7 @@ from .numpy import get_bitmask_shape, allocate_token_bitmask, fill_next_token_bi
 from numpy.typing import NDArray
 
 
-@mx.custom_function # type: ignore[misc]
+@mx.custom_function # type: ignore[untyped-decorator]
 def apply_token_bitmask_kernel(data: mx.array, mask: mx.array) -> mx.array:
     source = """
         uint batch = thread_position_in_grid.y;  // Batch index

llguidance/numpy.py

@@ -80,3 +80,22 @@ def fill_next_token_bitmask_par_with_draft_tokens(executor: LLExecutor,
     batch, vocab = bitmask.shape
     assert bitmask.flags["C_CONTIGUOUS"], "Mask must be contiguous"
     executor.unsafe_compute_mask_ptr_with_draft_token(matchers, bitmask.ctypes.data, vocab * 4, batch)
+
+
+def consume_token_par(executor: LLExecutor,
+                      matchers: List[Tuple[LLMatcher, int]]) -> List[bool]:
+    """
+    Consume a single token for each matcher in parallel.
+
+    Args:
+        executor: The LLExecutor to use for parallel execution.
+        matchers: List of tuples containing (LLMatcher, token_id).
+
+    Returns:
+        List[bool]: Success/failure for each matcher (in order).
+
+    Note:
+        Matchers that fail (return False) are left in an error state,
+        consistent with the behavior of consume_token on a single matcher.
+    """
+    return executor.consume_token_par(matchers)

llguidance/torch.py

@@ -77,3 +77,22 @@ def fill_next_token_bitmask_par_with_draft_tokens(executor: LLExecutor,
     batch, vocab = bitmask.shape
     assert bitmask.is_contiguous(), "Mask must be contiguous"
     executor.unsafe_compute_mask_ptr_with_draft_token(matchers, bitmask.data_ptr(), vocab * 4, batch)
+
+
+def consume_token_par(executor: LLExecutor,
+                      matchers: List[Tuple[LLMatcher, int]]) -> List[bool]:
+    """
+    Consume a single token for each matcher in parallel.
+
+    Args:
+        executor: The LLExecutor to use for parallel execution.
+        matchers: List of tuples containing (LLMatcher, token_id).
+
+    Returns:
+        List[bool]: Success/failure for each matcher (in order).
+
+    Note:
+        Matchers that fail (return False) are left in an error state,
+        consistent with the behavior of consume_token on a single matcher.
+    """
+    return executor.consume_token_par(matchers)

{llguidance-1.4.0.dist-info → llguidance-1.5.0.dist-info}/METADATA

@@ -1,14 +1,20 @@
 Metadata-Version: 2.4
 Name: llguidance
-Version: 1.4.0
+Version: 1.5.0
 License-File: LICENSE
 Summary: Bindings for the Low-level Guidance (llguidance) Rust library for use within Guidance
 Author: Michal Moskal
 License-Expression: MIT
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown; charset=UTF-8; variant=GFM
-Project-URL: repository, https://github.com/microsoft/llguidance
 Project-URL: issue_tracker, https://github.com/microsoft/llguidance/issues
+Project-URL: repository, https://github.com/microsoft/llguidance
+
+<div align="right">
+  <a href="https://discord.gg/cjPfAK43dz"><img src="https://img.shields.io/badge/Discord-Join%20Us-5865F2?logo=discord&logoColor=white" alt="Discord"></a>
+  <a href="mailto:guidanceai@microsoft.com"><img src="https://img.shields.io/badge/Email-guidanceai%40microsoft.com-0078D4?logo=microsoft-outlook&logoColor=white" alt="Email"></a>
+  <img src="https://img.shields.io/badge/Hours-10am--2pm%20Pacific-gray" alt="Hours">
+</div>
 
 # Low-level Guidance (llguidance)
 
@@ -35,7 +41,7 @@ Project-URL: issue_tracker, https://github.com/microsoft/llguidance/issues
 ## About
 
 This library implements constrained decoding (also called constrained sampling or
-structured outputs) for Large Langauge Models (LLMs).
+structured outputs) for Large Language Models (LLMs).
 It can enforce arbitrary context-free grammar on the output of LLM
 and is fast - on the order of 50μs of CPU time per token
 (for 128k tokenizer) with negligible startup costs.

{llguidance-1.4.0.dist-info → llguidance-1.5.0.dist-info}/RECORD

@@ -1,10 +1,7 @@
-llguidance-1.4.0.dist-info/METADATA,sha256=NpVepLhgx0KTvLTXRN1-k-RYfpKXNziLxSMQh2QZpdk,10302
-llguidance-1.4.0.dist-info/WHEEL,sha256=dY-16_uqpeJeKa9TH_TLDl5sXNjJIyZ_Q6aJg0vnMsg,145
-llguidance-1.4.0.dist-info/licenses/LICENSE,sha256=ws_MuBL-SCEBqPBFl9_FqZkaaydIJmxHrJG2parhU4M,1141
 llguidance/__init__.py,sha256=F9svXvm6oafbuUf_eq34PHJV4c7-yN133vmbWN6nIkc,590
 llguidance/_grammar_from.py,sha256=-vHqkPqJe6t0JKKuQhlUu08kYpPIVknMh8tZlh8FYeQ,2384
-llguidance/_lib.abi3.so,sha256=2vjFu2aGBBXoinSsRMSlVvn557RHIboGHcEOAdpsNzk,7191256
-llguidance/_lib.pyi,sha256=1m69VQ-7XaQLJKf6Vw5qf-UEh1cEigVMZM0Hj-XgSiw,24213
+llguidance/_lib.abi3.so,sha256=uU52BBgsVfyqn1voEbl6eBL9iXt1_5gSbtE_ZlRNjWI,7005360
+llguidance/_lib.pyi,sha256=UD2X7uyV7D-nwKJ9GpXrUFA3ZoIjTmKVPjFcjkhunpQ,24830
 llguidance/_struct_tag.py,sha256=83okmGWShxZud7S2vHjPRiInhFw0QVTHkeN8wtR8hR8,4430
 llguidance/_tokenizer.py,sha256=yC-RcgyMZN-olV-PnN4XkjlH-fOU8E9jrwO8VkXLv4M,1084
 llguidance/_util.py,sha256=6JV5SxjoH7hZPaSHhPRD_G6JzIhbKFFTqWTpp88VIiU,260
@@ -12,9 +9,12 @@ llguidance/cli.py,sha256=jhXdWbJC5rs6J8aknHtiuJeIWHcajZ7jYUj0ydeDJ68,2384
 llguidance/gbnf_to_lark.py,sha256=32XJ5Dzq-iSySnkV_rLaNZ888JjHBIr_QkSYdhtMAME,16635
 llguidance/hf.py,sha256=sLJKZxGpftuAY5eSCYpogzim3WZA7-vs9SOVgnzf3xI,1933
 llguidance/llamacpp.py,sha256=d_LjNbomBhj7uTo90h0muvPkOMso8NLe7H_YfPCqA8U,2284
-llguidance/mlx.py,sha256=ydMNSjnI2db8F685waOSWQLIoQTB9oehGBXnp9T3A6Q,2427
-llguidance/numpy.py,sha256=ZxHoQ_OmAgUHmxa_6KA8ruuKzXZLjmanPM9BYlKI0bo,3327
+llguidance/mlx.py,sha256=zvE3hxZNNXgX0JhNb19eK3bbHDaZWochEOH12C5fCgk,2440
+llguidance/numpy.py,sha256=mXvm4gp7WX_avVYKz2v4AlE4_v0dAgLuqmWy3PKOu6g,3947
 llguidance/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 llguidance/tiktoken.py,sha256=-mYDPxq4LM2FI9K8kLL0Us0qN-fZpNK0qIwboDPjfSk,1101
-llguidance/torch.py,sha256=qLwQjaxZbvaVss7Qy_XP-SEPGYRqJRkvd6KE4LmzZJ8,3372
-llguidance-1.4.0.dist-info/RECORD,,
+llguidance/torch.py,sha256=BhR7hKaZK0sltMXqr3vIcMx0FXz9k8XW5FbJIL0QvGU,3992
+llguidance-1.5.0.dist-info/METADATA,sha256=TLsiUVDRHA9S5IbbaDeKmUj6-C9qvZNSoj6-FfM5yc0,10751
+llguidance-1.5.0.dist-info/WHEEL,sha256=btCZOUXJMzNcRkylrJSGktvUfAhpsQmBEeAGzSFPuxs,145
+llguidance-1.5.0.dist-info/licenses/LICENSE,sha256=ws_MuBL-SCEBqPBFl9_FqZkaaydIJmxHrJG2parhU4M,1141
+llguidance-1.5.0.dist-info/RECORD,,

{llguidance-1.4.0.dist-info → llguidance-1.5.0.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: maturin (1.10.1)
+Generator: maturin (1.11.5)
 Root-Is-Purelib: false
 Tag: cp39-abi3-manylinux_2_17_aarch64
 Tag: cp39-abi3-manylinux2014_aarch64