llama-cpp-haystack 0.4.0__tar.gz → 0.4.2__tar.gz

{llama_cpp_haystack-0.4.0 → llama_cpp_haystack-0.4.2}/.gitignore

@@ -135,3 +135,12 @@ dmypy.json
 # Docs generation artifacts
 _readme_*.md
 .idea
+
+# macOS
+.DS_Store
+
+# http cache (requests-cache)
+**/http_cache.sqlite
+
+# ruff
+.ruff_cache

llama_cpp_haystack-0.4.2/CHANGELOG.md

@@ -0,0 +1,50 @@
+# Changelog
+
+## [integrations/llama_cpp-v0.4.1] - 2024-08-08
+
+### 🐛 Bug Fixes
+
+- Replace DynamicChatPromptBuilder with ChatPromptBuilder (#940)
+
+### ⚙️ Miscellaneous Tasks
+
+- Retry tests to reduce flakyness (#836)
+- Update ruff invocation to include check parameter (#853)
+- Pin `llama-cpp-python>=0.2.87` (#955)
+
+## [integrations/llama_cpp-v0.4.0] - 2024-05-13
+
+### 🐛 Bug Fixes
+
+- Fix commit (#436)
+
+
+- Fix order of API docs (#447)
+
+This PR will also push the docs to Readme
+
+### 📚 Documentation
+
+- Update category slug (#442)
+- Small consistency improvements (#536)
+- Disable-class-def (#556)
+
+### ⚙️ Miscellaneous Tasks
+
+- [**breaking**] Rename model_path to model in the Llama.cpp integration (#243)
+
+### Llama.cpp
+
+- Generate api docs (#353)
+
+## [integrations/llama_cpp-v0.2.1] - 2024-01-18
+
+## [integrations/llama_cpp-v0.2.0] - 2024-01-17
+
+## [integrations/llama_cpp-v0.1.0] - 2024-01-09
+
+### 🚀 Features
+
+- Add Llama.cpp Generator (#179)
+
+<!-- generated by git-cliff -->

{llama_cpp_haystack-0.4.0 → llama_cpp_haystack-0.4.2}/PKG-INFO

@@ -1,14 +1,13 @@
 Metadata-Version: 2.3
 Name: llama-cpp-haystack
-Version: 0.4.0
+Version: 0.4.2
 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
 Project-URL: Source, https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/llama_cpp
 Author: Ashwin Mathur
 Author-email: deepset GmbH <info@deepset.ai>
-License-Expression: Apache-2.0
-License-File: LICENSE.txt
+License: Apache-2.0
 Classifier: Development Status :: 4 - Beta
 Classifier: License :: OSI Approved :: Apache Software License
 Classifier: Programming Language :: Python
@@ -21,7 +20,7 @@ Classifier: Programming Language :: Python :: Implementation :: CPython
 Classifier: Programming Language :: Python :: Implementation :: PyPy
 Requires-Python: >=3.8
 Requires-Dist: haystack-ai
-Requires-Dist: llama-cpp-python
+Requires-Dist: llama-cpp-python>=0.2.87
 Description-Content-Type: text/markdown
 
 # llama-cpp-haystack

{llama_cpp_haystack-0.4.0 → llama_cpp_haystack-0.4.2}/examples/rag_pipeline_example.py

@@ -6,6 +6,7 @@ from haystack.components.embedders import SentenceTransformersDocumentEmbedder,
 from haystack.components.retrievers import InMemoryEmbeddingRetriever
 from haystack.components.writers import DocumentWriter
 from haystack.document_stores import InMemoryDocumentStore
+
 from haystack_integrations.components.generators.llama_cpp import LlamaCppGenerator
 
 # Load first 100 rows of the Simple Wikipedia Dataset from HuggingFace

{llama_cpp_haystack-0.4.0 → llama_cpp_haystack-0.4.2}/pyproject.toml

@@ -15,7 +15,7 @@ authors = [
     { name = "Ashwin Mathur", email = "" },
 ]
 classifiers = [
-  "License :: OSI Approved :: Apache Software License",
+    "License :: OSI Approved :: Apache Software License",
     "Development Status :: 4 - Beta",
     "Programming Language :: Python",
     "Programming Language :: Python :: 3.8",
@@ -26,10 +26,7 @@ classifiers = [
     "Programming Language :: Python :: Implementation :: CPython",
     "Programming Language :: Python :: Implementation :: PyPy",
 ]
-dependencies = [
-  "haystack-ai",
-  "llama-cpp-python"
-]
+dependencies = ["haystack-ai", "llama-cpp-python>=0.2.87"]
 
 [project.urls]
 Documentation = "https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/llama_cpp#readme"
@@ -48,59 +45,42 @@ root = "../.."
 git_describe_command = 'git describe --tags --match="integrations/llama_cpp-v[0-9]*"'
 
 [tool.hatch.envs.default]
+installer = "uv"
 dependencies = [
     "coverage[toml]>=6.5",
     "pytest",
+    "pytest-rerunfailures",
     "haystack-pydoc-tools",
-    "transformers[sentencepiece]"
+    "transformers[sentencepiece]",
 ]
 [tool.hatch.envs.default.scripts]
 test = "pytest {args:tests}"
 test-cov = "coverage run -m pytest {args:tests}"
-cov-report = [
-    "- coverage combine",
-    "coverage report",
-]
-cov = [
-    "test-cov",
-    "cov-report",
-]
-docs = [
-  "pydoc-markdown pydoc/config.yml"
-]
+test-cov-retry = "test-cov --reruns 3 --reruns-delay 30 -x"
+cov-report = ["- coverage combine", "coverage report"]
+cov = ["test-cov", "cov-report"]
+cov-retry = ["test-cov-retry", "cov-report"]
+docs = ["pydoc-markdown pydoc/config.yml"]
 [[tool.hatch.envs.all.matrix]]
 python = ["3.8", "3.9", "3.10", "3.11", "3.12"]
 
 
 [tool.hatch.envs.lint]
+installer = "uv"
 detached = true
-dependencies = [
-    "black>=23.1.0",
-    "mypy>=1.0.0",
-    "ruff>=0.0.243",
-]
+dependencies = ["pip", "black>=23.1.0", "mypy>=1.0.0", "ruff>=0.0.243"]
 
 [tool.hatch.envs.lint.scripts]
 typing = "mypy --install-types --non-interactive --explicit-package-bases {args:src/ tests}"
-style = [
-    "ruff {args:.}",
-    "black --check --diff {args:.}",
-]
-fmt = [
-    "black {args:.}",
-    "ruff --fix {args:.}",
-    "style",
-]
-all = [
-    "style",
-    "typing",
-]
+style = ["ruff check {args:.}", "black --check --diff {args:.}"]
+fmt = ["black {args:.}", "ruff check --fix {args:.}", "style"]
+all = ["style", "typing"]
 
 [tool.hatch.metadata]
 allow-direct-references = true
 
-[tool.ruff.isort]
-known-first-party = ["src"]
+[tool.ruff.lint.isort]
+known-first-party = ["haystack_integrations"]
 
 [tool.black]
 target-version = ["py38"]
@@ -110,6 +90,8 @@ skip-string-normalization = true
 [tool.ruff]
 target-version = "py38"
 line-length = 120
+
+[tool.ruff.lint]
 select = [
     "A",
     "ARG",
@@ -140,19 +122,25 @@ ignore = [
     # Allow non-abstract empty methods in abstract base classes
     "B027",
     # Ignore checks for possible passwords
-    "S105", "S106", "S107",
+    "S105",
+    "S106",
+    "S107",
     # Ignore complexity
-    "C901", "PLR0911", "PLR0912", "PLR0913", "PLR0915",
+    "C901",
+    "PLR0911",
+    "PLR0912",
+    "PLR0913",
+    "PLR0915",
 ]
 unfixable = [
     # Don't touch unused imports
     "F401",
 ]
 
-[tool.ruff.flake8-tidy-imports]
+[tool.ruff.lint.flake8-tidy-imports]
 ban-relative-imports = "parents"
 
-[tool.ruff.per-file-ignores]
+[tool.ruff.lint.per-file-ignores]
 # Tests can use magic values, assertions, and relative imports
 "tests/**/*" = ["PLR2004", "S101", "TID252"]
 # Examples can print their output
@@ -167,27 +155,16 @@ parallel = false
 
 [tool.coverage.report]
 omit = ["*/tests/*", "*/__init__.py"]
-show_missing=true
-exclude_lines = [
-  "no cov",
-  "if __name__ == .__main__.:",
-  "if TYPE_CHECKING:",
-]
+show_missing = true
+exclude_lines = ["no cov", "if __name__ == .__main__.:", "if TYPE_CHECKING:"]
 
 
 [tool.pytest.ini_options]
 markers = [
     "integration: marks tests as slow (deselect with '-m \"not integration\"')",
 ]
-addopts = [
-    "--import-mode=importlib",
-]
+addopts = ["--import-mode=importlib"]
 
 [[tool.mypy.overrides]]
-module = [
-    "haystack.*",
-    "haystack_integrations.*",
-    "pytest.*",
-    "llama_cpp.*"
-]
+module = ["haystack.*", "haystack_integrations.*", "pytest.*", "llama_cpp.*"]
 ignore_missing_imports = true

{llama_cpp_haystack-0.4.0 → llama_cpp_haystack-0.4.2}/src/haystack_integrations/components/generators/llama_cpp/__init__.py

@@ -5,4 +5,4 @@
 from .chat.chat_generator import LlamaCppChatGenerator
 from .generator import LlamaCppGenerator
 
-__all__ = ["LlamaCppGenerator", "LlamaCppChatGenerator"]
+__all__ = ["LlamaCppChatGenerator", "LlamaCppGenerator"]

{llama_cpp_haystack-0.4.0 → llama_cpp_haystack-0.4.2}/src/haystack_integrations/components/generators/llama_cpp/chat/chat_generator.py

@@ -9,6 +9,21 @@ from llama_cpp.llama_tokenizer import LlamaHFTokenizer
 logger = logging.getLogger(__name__)
 
 
+def _convert_message_to_llamacpp_format(message: ChatMessage) -> Dict[str, str]:
+    """
+    Convert a message to the format expected by Llama.cpp.
+    :returns: A dictionary with the following keys:
+        - `role`
+        - `content`
+        - `name` (optional)
+    """
+    formatted_msg = {"role": message.role.value, "content": message.text}
+    if message.name:
+        formatted_msg["name"] = message.name
+
+    return formatted_msg
+
+
 @component
 class LlamaCppChatGenerator:
     """
@@ -96,7 +111,7 @@ class LlamaCppChatGenerator:
             return {"replies": []}
 
         updated_generation_kwargs = {**self.generation_kwargs, **(generation_kwargs or {})}
-        formatted_messages = [msg.to_openai_format() for msg in messages]
+        formatted_messages = [_convert_message_to_llamacpp_format(msg) for msg in messages]
 
         response = self.model.create_chat_completion(messages=formatted_messages, **updated_generation_kwargs)
         replies = [

{llama_cpp_haystack-0.4.0 → llama_cpp_haystack-0.4.2}/tests/test_chat_generator.py

@@ -6,11 +6,15 @@ from unittest.mock import MagicMock
 
 import pytest
 from haystack import Document, Pipeline
-from haystack.components.builders.dynamic_chat_prompt_builder import DynamicChatPromptBuilder
+from haystack.components.builders import ChatPromptBuilder
 from haystack.components.retrievers.in_memory import InMemoryBM25Retriever
 from haystack.dataclasses import ChatMessage, ChatRole
 from haystack.document_stores.in_memory import InMemoryDocumentStore
-from haystack_integrations.components.generators.llama_cpp import LlamaCppChatGenerator
+
+from haystack_integrations.components.generators.llama_cpp.chat.chat_generator import (
+    LlamaCppChatGenerator,
+    _convert_message_to_llamacpp_format,
+)
 
 
 @pytest.fixture
@@ -29,6 +33,21 @@ def download_file(file_link, filename, capsys):
             print("\nModel file already exists.")
 
 
+def test_convert_message_to_llamacpp_format():
+    message = ChatMessage.from_system("You are good assistant")
+    assert _convert_message_to_llamacpp_format(message) == {"role": "system", "content": "You are good assistant"}
+
+    message = ChatMessage.from_user("I have a question")
+    assert _convert_message_to_llamacpp_format(message) == {"role": "user", "content": "I have a question"}
+
+    message = ChatMessage.from_function("Function call", "function_name")
+    assert _convert_message_to_llamacpp_format(message) == {
+        "role": "function",
+        "content": "Function call",
+        "name": "function_name",
+    }
+
+
 class TestLlamaCppChatGenerator:
     @pytest.fixture
     def generator(self, model_path, capsys):
@@ -144,7 +163,7 @@ class TestLlamaCppChatGenerator:
         assert isinstance(result["replies"], list)
         assert len(result["replies"]) == 1
         assert isinstance(result["replies"][0], ChatMessage)
-        assert result["replies"][0].content == "Generated text"
+        assert result["replies"][0].text == "Generated text"
         assert result["replies"][0].role == ChatRole.ASSISTANT
 
     def test_run_with_generation_kwargs(self, generator_mock):
@@ -164,7 +183,7 @@ class TestLlamaCppChatGenerator:
         mock_model.create_chat_completion.return_value = mock_output
         generation_kwargs = {"max_tokens": 128}
         result = generator.run([ChatMessage.from_system("Write a 200 word paragraph.")], generation_kwargs)
-        assert result["replies"][0].content == "Generated text"
+        assert result["replies"][0].text == "Generated text"
         assert result["replies"][0].meta["finish_reason"] == "length"
 
     @pytest.mark.integration
@@ -187,7 +206,7 @@ class TestLlamaCppChatGenerator:
             assert "replies" in result
             assert isinstance(result["replies"], list)
             assert len(result["replies"]) > 0
-            assert any(answer.lower() in reply.content.lower() for reply in result["replies"])
+            assert any(answer.lower() in reply.text.lower() for reply in result["replies"])
 
     @pytest.mark.integration
     def test_run_rag_pipeline(self, generator):
@@ -213,9 +232,7 @@ class TestLlamaCppChatGenerator:
             instance=InMemoryBM25Retriever(document_store=document_store, top_k=1),
             name="retriever",
         )
-        pipeline.add_component(
-            instance=DynamicChatPromptBuilder(runtime_variables=["query", "documents"]), name="prompt_builder"
-        )
+        pipeline.add_component(instance=ChatPromptBuilder(variables=["query", "documents"]), name="prompt_builder")
         pipeline.add_component(instance=generator, name="llm")
         pipeline.connect("retriever.documents", "prompt_builder.documents")
         pipeline.connect("prompt_builder.prompt", "llm.messages")
@@ -245,7 +262,7 @@ class TestLlamaCppChatGenerator:
                 "retriever": {"query": question},
                 "prompt_builder": {
                     "template_variables": {"location": location},
-                    "prompt_source": messages,
+                    "template": messages,
                     "query": question,
                 },
             }
@@ -253,7 +270,7 @@ class TestLlamaCppChatGenerator:
 
         replies = result["llm"]["replies"]
         assert len(replies) > 0
-        assert any("bioluminescent waves" in reply.content for reply in replies)
+        assert any("bioluminescent waves" in reply.text.lower() for reply in replies)
         assert all(reply.role == ChatRole.ASSISTANT for reply in replies)
 
     @pytest.mark.integration
@@ -291,15 +308,15 @@ class TestLlamaCppChatGenerator:
         assert len(result["replies"]) > 0
         assert all(reply.role == ChatRole.ASSISTANT for reply in result["replies"])
         for reply in result["replies"]:
-            assert json.loads(reply.content)
-            assert isinstance(json.loads(reply.content), dict)
-            assert "people" in json.loads(reply.content)
-            assert isinstance(json.loads(reply.content)["people"], list)
-            assert all(isinstance(person, dict) for person in json.loads(reply.content)["people"])
-            assert all("name" in person for person in json.loads(reply.content)["people"])
-            assert all("age" in person for person in json.loads(reply.content)["people"])
-            assert all(isinstance(person["name"], str) for person in json.loads(reply.content)["people"])
-            assert all(isinstance(person["age"], int) for person in json.loads(reply.content)["people"])
+            assert json.loads(reply.text)
+            assert isinstance(json.loads(reply.text), dict)
+            assert "people" in json.loads(reply.text)
+            assert isinstance(json.loads(reply.text)["people"], list)
+            assert all(isinstance(person, dict) for person in json.loads(reply.text)["people"])
+            assert all("name" in person for person in json.loads(reply.text)["people"])
+            assert all("age" in person for person in json.loads(reply.text)["people"])
+            assert all(isinstance(person["name"], str) for person in json.loads(reply.text)["people"])
+            assert all(isinstance(person["age"], int) for person in json.loads(reply.text)["people"])
 
 
 class TestLlamaCppChatGeneratorFunctionary:
@@ -325,7 +342,7 @@ class TestLlamaCppChatGeneratorFunctionary:
         hf_tokenizer_path = "meetkai/functionary-small-v2.4-GGUF"
         generator = LlamaCppChatGenerator(
             model=model_path,
-            n_ctx=8192,
+            n_ctx=512,
             n_batch=512,
             model_kwargs={
                 "chat_format": "functionary-v2",
@@ -382,7 +399,6 @@ class TestLlamaCppChatGeneratorFunctionary:
                                 "type": "string",
                                 "description": "The city and state, e.g. San Francisco, CA",
                             },
-                            "unit": {"type": "string", "enum": ["celsius", "fahrenheit"]},
                         },
                         "required": ["location"],
                     },
@@ -390,7 +406,8 @@ class TestLlamaCppChatGeneratorFunctionary:
             }
         ]
 
-        response = generator.run(messages=messages, generation_kwargs={"tools": tools})
+        tool_choice = {"type": "function", "function": {"name": "get_current_temperature"}}
+        response = generator.run(messages=messages, generation_kwargs={"tools": tools, "tool_choice": tool_choice})
 
         available_functions = {
             "get_current_temperature": self.get_current_temperature,
@@ -412,11 +429,10 @@ class TestLlamaCppChatGeneratorFunctionary:
             messages.append(function_message)
 
         second_response = generator.run(messages=messages)
-        print(second_response)
         assert "replies" in second_response
         assert len(second_response["replies"]) > 0
-        assert any("San Francisco" in reply.content for reply in second_response["replies"])
-        assert any("72" in reply.content for reply in second_response["replies"])
+        assert any("San Francisco" in reply.text for reply in second_response["replies"])
+        assert any("72" in reply.text for reply in second_response["replies"])
 
 
 class TestLlamaCppChatGeneratorChatML:

{llama_cpp_haystack-0.4.0 → llama_cpp_haystack-0.4.2}/tests/test_generator.py

@@ -9,6 +9,7 @@ from haystack.components.builders.answer_builder import AnswerBuilder
 from haystack.components.builders.prompt_builder import PromptBuilder
 from haystack.components.retrievers.in_memory import InMemoryBM25Retriever
 from haystack.document_stores.in_memory import InMemoryDocumentStore
+
 from haystack_integrations.components.generators.llama_cpp import LlamaCppGenerator