llama-cpp-haystack 0.4.1__tar.gz → 0.4.3__tar.gz

llama_cpp_haystack-0.4.3/CHANGELOG.md

@@ -0,0 +1,93 @@
+# Changelog
+
+## [integrations/llama_cpp-v0.4.2] - 2024-12-10
+
+### 🧪 Testing
+
+- Do not retry tests in `hatch run test` command (#954)
+
+### ⚙️ CI
+
+- Adopt uv as installer (#1142)
+
+### 🧹 Chores
+
+- Update ruff linting scripts and settings (#1105)
+- Unpin `llama-cpp-python` (#1115)
+- Fix linting/isort (#1215)
+- Use text instead of content for ChatMessage in Llama.cpp, Langfuse and Mistral (#1238)
+
+### 🌀 Miscellaneous
+
+- Chore: lamma_cpp - ruff update, don't ruff tests (#998)
+- Fix: pin `llama-cpp-python<0.3.0` (#1111)
+
+## [integrations/llama_cpp-v0.4.1] - 2024-08-08
+
+### 🐛 Bug Fixes
+
+- Replace DynamicChatPromptBuilder with ChatPromptBuilder (#940)
+
+### ⚙️ CI
+
+- Retry tests to reduce flakyness (#836)
+
+### 🧹 Chores
+
+- Update ruff invocation to include check parameter (#853)
+- Pin `llama-cpp-python>=0.2.87` (#955)
+
+### 🌀 Miscellaneous
+
+- Ci: install `pytest-rerunfailures` where needed; add retry config to `test-cov` script (#845)
+- Fix: pin llama-cpp-python to an older version (#943)
+- Refactor: introduce `_convert_message_to_llamacpp_format` utility function (#939)
+
+## [integrations/llama_cpp-v0.4.0] - 2024-05-13
+
+### 🐛 Bug Fixes
+
+- Llama.cpp: change wrong links and imports (#436)
+- Fix order of API docs (#447)
+
+### 📚 Documentation
+
+- Update category slug (#442)
+- Small consistency improvements (#536)
+- Disable-class-def (#556)
+
+### 🧹 Chores
+
+- [**breaking**] Rename model_path to model in the Llama.cpp integration (#243)
+
+### 🌀 Miscellaneous
+
+- Generate api docs (#353)
+- Model_name_or_path > model (#418)
+- Llama.cpp - review docstrings (#510)
+- Llama.cpp - update examples (#511)
+- Make tests show coverage (#566)
+- Remove references to Python 3.7 (#601)
+- Chore: add license classifiers (#680)
+- Chore: change the pydoc renderer class (#718)
+- Basic implementation of llama.cpp chat generation (#723)
+
+## [integrations/llama_cpp-v0.2.1] - 2024-01-18
+
+### 🌀 Miscellaneous
+
+- Update import paths for beta5 (#233)
+
+## [integrations/llama_cpp-v0.2.0] - 2024-01-17
+
+### 🌀 Miscellaneous
+
+- Mount llama_cpp in haystack_integrations (#217)
+
+## [integrations/llama_cpp-v0.1.0] - 2024-01-09
+
+### 🚀 Features
+
+- Add Llama.cpp Generator (#179)
+
+<!-- generated by git-cliff -->

{llama_cpp_haystack-0.4.1 → llama_cpp_haystack-0.4.3}/PKG-INFO

@@ -1,6 +1,6 @@
-Metadata-Version: 2.3
+Metadata-Version: 2.4
 Name: llama-cpp-haystack
-Version: 0.4.1
+Version: 0.4.3
 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-0.4.1 → llama_cpp_haystack-0.4.3}/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-0.4.1 → llama_cpp_haystack-0.4.3}/pyproject.toml

@@ -45,6 +45,7 @@ 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",
@@ -53,30 +54,33 @@ dependencies = [
     "transformers[sentencepiece]",
 ]
 [tool.hatch.envs.default.scripts]
-test = "pytest --reruns 3 --reruns-delay 30 -x {args:tests}"
-test-cov = "coverage run -m pytest --reruns 3 --reruns-delay 30 -x {args:tests}"
+test = "pytest {args:tests}"
+test-cov = "coverage run -m pytest {args:tests}"
+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 check {args:.}", "black --check --diff {args:.}"]
-fmt = ["black {args:.}", "ruff --fix {args:.}", "style"]
+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"]
@@ -86,6 +90,8 @@ skip-string-normalization = true
 [tool.ruff]
 target-version = "py38"
 line-length = 120
+
+[tool.ruff.lint]
 select = [
     "A",
     "ARG",
@@ -131,10 +137,10 @@ unfixable = [
     "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

{llama_cpp_haystack-0.4.1 → llama_cpp_haystack-0.4.3}/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-0.4.1 → llama_cpp_haystack-0.4.3}/src/haystack_integrations/components/generators/llama_cpp/chat/chat_generator.py

@@ -2,7 +2,7 @@ import logging
 from typing import Any, Dict, List, Optional
 
 from haystack import component
-from haystack.dataclasses import ChatMessage, ChatRole
+from haystack.dataclasses import ChatMessage
 from llama_cpp import Llama
 from llama_cpp.llama_tokenizer import LlamaHFTokenizer
 
@@ -17,10 +17,14 @@ def _convert_message_to_llamacpp_format(message: ChatMessage) -> Dict[str, str]:
         - `content`
         - `name` (optional)
     """
-    formatted_msg = {"role": message.role.value, "content": message.content}
+    formatted_msg = {"role": message.role.value, "content": message.text}
     if message.name:
         formatted_msg["name"] = message.name
 
+    if formatted_msg["role"] == "tool":
+        formatted_msg["name"] = message.tool_call_result.origin.tool_name
+        formatted_msg["content"] = message.tool_call_result.result
+
     return formatted_msg
 
 
@@ -114,26 +118,31 @@ class LlamaCppChatGenerator:
         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 = [
-            ChatMessage(
-                content=choice["message"]["content"],
-                role=ChatRole[choice["message"]["role"].upper()],
-                name=None,
-                meta={
-                    "response_id": response["id"],
-                    "model": response["model"],
-                    "created": response["created"],
-                    "index": choice["index"],
-                    "finish_reason": choice["finish_reason"],
-                    "usage": response["usage"],
-                },
-            )
-            for choice in response["choices"]
-        ]
-
-        for reply, choice in zip(replies, response["choices"]):
+
+        replies = []
+
+        for choice in response["choices"]:
+            meta = {
+                "response_id": response["id"],
+                "model": response["model"],
+                "created": response["created"],
+                "index": choice["index"],
+                "finish_reason": choice["finish_reason"],
+                "usage": response["usage"],
+            }
+
+            name = None
             tool_calls = choice.get("message", {}).get("tool_calls", [])
             if tool_calls:
-                reply.meta["tool_calls"] = tool_calls
-        reply.name = tool_calls[0]["function"]["name"] if tool_calls else None
+                meta["tool_calls"] = tool_calls
+                name = tool_calls[0]["function"]["name"]
+
+            reply = ChatMessage.from_assistant(choice["message"]["content"], meta=meta)
+            if name:
+                if hasattr(reply, "_name"):
+                    reply._name = name  # new ChatMessage
+                elif hasattr(reply, "name"):
+                    reply.name = name  # legacy ChatMessage
+            replies.append(reply)
+
         return {"replies": replies}

{llama_cpp_haystack-0.4.1 → llama_cpp_haystack-0.4.3}/tests/test_chat_generator.py

@@ -10,6 +10,7 @@ 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.chat.chat_generator import (
     LlamaCppChatGenerator,
     _convert_message_to_llamacpp_format,
@@ -40,11 +41,11 @@ def test_convert_message_to_llamacpp_format():
     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",
-    }
+    converted_message = _convert_message_to_llamacpp_format(message)
+
+    assert converted_message["role"] in ("function", "tool")
+    assert converted_message["name"] == "function_name"
+    assert converted_message["content"] == "Function call"
 
 
 class TestLlamaCppChatGenerator:
@@ -162,7 +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):
@@ -182,7 +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
@@ -205,7 +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):
@@ -269,7 +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
@@ -307,15 +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:
@@ -341,7 +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",
@@ -398,7 +399,6 @@ class TestLlamaCppChatGeneratorFunctionary:
                                 "type": "string",
                                 "description": "The city and state, e.g. San Francisco, CA",
                             },
-                            "unit": {"type": "string", "enum": ["celsius", "fahrenheit"]},
                         },
                         "required": ["location"],
                     },
@@ -406,7 +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,
@@ -430,8 +431,8 @@ class TestLlamaCppChatGeneratorFunctionary:
         second_response = generator.run(messages=messages)
         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-0.4.1 → llama_cpp_haystack-0.4.3}/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