llama-cpp-haystack 2.0.0__tar.gz → 2.1.0__tar.gz

{llama_cpp_haystack-2.0.0 → llama_cpp_haystack-2.1.0}/CHANGELOG.md

@@ -1,5 +1,21 @@
 # Changelog
 
+## [integrations/llama_cpp-v2.0.0] - 2026-01-12
+
+### ⚙️ CI
+
+- Change pytest command (#2475)
+
+### 🧹 Chores
+
+- Remove Readme API CI workflow and configs (#2573)
+- Make fmt command more forgiving (#2671)
+- [**breaking**] Llama_cpp - drop Python 3.9 and use X|Y typing (#2710)
+
+### 🌀 Miscellaneous
+
+- Enhancement: Adopt PEP 585 type hinting (part 4) (#2527)
+
 ## [integrations/llama_cpp-v1.4.0] - 2025-10-23
 
 ### 📚 Documentation

{llama_cpp_haystack-2.0.0 → llama_cpp_haystack-2.1.0}/LICENSE.txt

@@ -58,7 +58,7 @@ APPENDIX: How to apply the Apache License to your work.
 
 To apply the Apache License to your work, attach the following boilerplate notice, with the fields enclosed by brackets "[]" replaced with your own identifying information. (Don't include the brackets!)  The text should be enclosed in the appropriate comment syntax for the file format. We also recommend that a file or class name and description of purpose be included on the same "printed page" as the copyright notice for easier identification within third-party archives.
 
-Copyright [yyyy] [name of copyright owner]
+Copyright 2024 deepset GmbH
 
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.

{llama_cpp_haystack-2.0.0 → llama_cpp_haystack-2.1.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: llama-cpp-haystack
-Version: 2.0.0
+Version: 2.1.0
 Summary: An integration between the llama.cpp LLM framework and Haystack
 Project-URL: Documentation, https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/llama_cpp#readme
 Project-URL: Issues, https://github.com/deepset-ai/haystack-core-integrations/issues

{llama_cpp_haystack-2.0.0 → llama_cpp_haystack-2.1.0}/src/haystack_integrations/components/generators/llama_cpp/chat/chat_generator.py

@@ -1,3 +1,7 @@
+# SPDX-FileCopyrightText: 2024-present deepset GmbH <info@deepset.ai>
+#
+# SPDX-License-Identifier: Apache-2.0
+
 import json
 from collections.abc import Iterator
 from datetime import datetime, timezone
@@ -201,7 +205,7 @@ class LlamaCppChatGenerator:
         streaming_callback: StreamingCallbackT | None = None,
         chat_handler_name: str | None = None,
         model_clip_path: str | None = None,
-    ):
+    ) -> None:
         """
         :param model: The path of a quantized model for text generation, for example, "zephyr-7b-beta.Q4_0.gguf".
             If the model path is also specified in the `model_kwargs`, this parameter will be ignored.
@@ -263,7 +267,7 @@ class LlamaCppChatGenerator:
         self.model_clip_path = model_clip_path
         self._handler = handler
 
-    def warm_up(self):
+    def warm_up(self) -> None:
         if self._model is not None:
             return
 
@@ -347,8 +351,7 @@ class LlamaCppChatGenerator:
             - `replies`: The responses from the model
         """
         if self._model is None:
-            error_msg = "The model has not been loaded. Please call warm_up() before running."
-            raise RuntimeError(error_msg)
+            self.warm_up()
 
         if not messages:
             return {"replies": []}
@@ -381,7 +384,7 @@ class LlamaCppChatGenerator:
         )
 
         if streaming_callback:
-            response_stream = self._model.create_chat_completion(
+            response_stream = self._model.create_chat_completion(  # type: ignore[union-attr]
                 messages=formatted_messages, tools=llamacpp_tools, **updated_generation_kwargs, stream=True
             )
             return self._handle_streaming_response(
@@ -391,7 +394,7 @@ class LlamaCppChatGenerator:
             )  # we know that response_stream is Iterator[CreateChatCompletionStreamResponse]
             # because create_chat_completion was called with stream=True, but mypy doesn't know that
 
-        response = self._model.create_chat_completion(
+        response = self._model.create_chat_completion(  # type: ignore[union-attr]
             messages=formatted_messages, tools=llamacpp_tools, **updated_generation_kwargs
         )
         replies = []
@@ -399,8 +402,11 @@ class LlamaCppChatGenerator:
             msg = f"Expected a dictionary response, got a different object: {response}"
             raise ValueError(msg)
 
-        for choice in response["choices"]:
-            chat_message = self._convert_chat_completion_choice_to_chat_message(choice, response)
+        for choice in response["choices"]:  # type: ignore[index]
+            chat_message = self._convert_chat_completion_choice_to_chat_message(
+                choice,
+                response,  # type: ignore[arg-type]
+            )
             replies.append(chat_message)
         return {"replies": replies}
 

{llama_cpp_haystack-2.0.0 → llama_cpp_haystack-2.1.0}/src/haystack_integrations/components/generators/llama_cpp/generator.py

@@ -1,3 +1,7 @@
+# SPDX-FileCopyrightText: 2024-present deepset GmbH <info@deepset.ai>
+#
+# SPDX-License-Identifier: Apache-2.0
+
 from typing import Any
 
 from haystack import component, logging
@@ -32,7 +36,7 @@ class LlamaCppGenerator:
         n_batch: int | None = 512,
         model_kwargs: dict[str, Any] | None = None,
         generation_kwargs: dict[str, Any] | None = None,
-    ):
+    ) -> None:
         """
         :param model: The path of a quantized model for text generation, for example, "zephyr-7b-beta.Q4_0.gguf".
             If the model path is also specified in the `model_kwargs`, this parameter will be ignored.
@@ -64,7 +68,7 @@ class LlamaCppGenerator:
         self.generation_kwargs = generation_kwargs
         self.model: Llama | None = None
 
-    def warm_up(self):
+    def warm_up(self) -> None:
         if self.model is None:
             self.model = Llama(**self.model_kwargs)
 
@@ -84,8 +88,7 @@ class LlamaCppGenerator:
             - `meta`: metadata about the request.
         """
         if self.model is None:
-            error_msg = "The model has not been loaded. Please call warm_up() before running."
-            raise RuntimeError(error_msg)
+            self.warm_up()
 
         if not prompt:
             return {"replies": []}
@@ -93,7 +96,7 @@ class LlamaCppGenerator:
         # merge generation kwargs from init method with those from run method
         updated_generation_kwargs = {**self.generation_kwargs, **(generation_kwargs or {})}
 
-        output = self.model.create_completion(prompt=prompt, **updated_generation_kwargs)
+        output = self.model.create_completion(prompt=prompt, **updated_generation_kwargs)  # type: ignore[union-attr]
         if not isinstance(output, dict):
             msg = f"Expected a dictionary response, got a different object: {output}"
             raise ValueError(msg)

{llama_cpp_haystack-2.0.0 → llama_cpp_haystack-2.1.0}/tests/test_chat_generator.py

@@ -1,3 +1,7 @@
+# SPDX-FileCopyrightText: 2024-present deepset GmbH <info@deepset.ai>
+#
+# SPDX-License-Identifier: Apache-2.0
+
 import json
 import os
 import urllib.request
@@ -678,7 +682,6 @@ class TestLlamaCppChatGenerator:
 
         model_path = str(model_path / filename)
         generator = LlamaCppChatGenerator(model=model_path, n_ctx=8192, n_batch=512)
-        generator.warm_up()
         return generator
 
     @pytest.fixture
@@ -898,14 +901,6 @@ class TestLlamaCppChatGenerator:
         )
         assert generator.model_kwargs["n_batch"] == 1024
 
-    def test_raises_error_without_warm_up(self):
-        """
-        Test that the generator raises an error if warm_up() is not called before running.
-        """
-        generator = LlamaCppChatGenerator(model="test_model.gguf", n_ctx=512, n_batch=512)
-        with pytest.raises(RuntimeError):
-            generator.run("What is the capital of China?")
-
     def test_run_with_empty_message(self, generator_mock):
         """
         Test that an empty message returns an empty list of replies.
@@ -1179,7 +1174,6 @@ class TestLlamaCppChatGeneratorFunctionary:
                 "hf_tokenizer_path": hf_tokenizer_path,
             },
         )
-        generator.warm_up()
         return generator
 
     @pytest.mark.integration
@@ -1260,7 +1254,6 @@ class TestLlamaCppChatGeneratorChatML:
                 "chat_format": "chatml-function-calling",
             },
         )
-        generator.warm_up()
         return generator
 
     @pytest.mark.integration
@@ -1326,8 +1319,6 @@ class TestLlamaCppChatGeneratorChatML:
             generation_kwargs={"max_tokens": 50, "temperature": 0.1},
         )
 
-        generator.warm_up()
-
         result = generator.run(messages)
 
         assert "replies" in result

{llama_cpp_haystack-2.0.0 → llama_cpp_haystack-2.1.0}/tests/test_generator.py

@@ -1,3 +1,7 @@
+# SPDX-FileCopyrightText: 2024-present deepset GmbH <info@deepset.ai>
+#
+# SPDX-License-Identifier: Apache-2.0
+
 import os
 import urllib.request
 from pathlib import Path
@@ -42,7 +46,6 @@ class TestLlamaCppGenerator:
 
         model_path = str(model_path / filename)
         generator = LlamaCppGenerator(model=model_path, n_ctx=128, n_batch=128)
-        generator.warm_up()
         return generator
 
     @pytest.fixture
@@ -106,14 +109,6 @@ class TestLlamaCppGenerator:
         generator = LlamaCppGenerator(model="test_model.gguf", n_ctx=512, n_batch=512, model_kwargs={"n_batch": 1024})
         assert generator.model_kwargs["n_batch"] == 1024
 
-    def test_raises_error_without_warm_up(self):
-        """
-        Test that the generator raises an error if warm_up() is not called before running.
-        """
-        generator = LlamaCppGenerator(model="test_model.gguf", n_ctx=512, n_batch=512)
-        with pytest.raises(RuntimeError):
-            generator.run("What is the capital of China?")
-
     def test_run_with_empty_prompt(self, generator_mock):
         """
         Test that an empty prompt returns an empty list of replies.