Merge branch 'deepset-ai:main' into ci-deepset-ai#5931-isort

.github/workflows/benchmarks.yml

@@ -21,10 +21,10 @@ jobs:
     steps:
       - uses: actions/checkout@v4
 
-      - uses: iterative/setup-cml@v1
+      - uses: iterative/setup-cml@v2
 
       - name: AWS authentication
-        uses: aws-actions/configure-aws-credentials@8c3f20df09ac63af7b3ae3d7c91f105f857d8497
+        uses: aws-actions/configure-aws-credentials@010d0da01d0b5a38af31e9c3470dbfdabdecca3a
         with:
           aws-region: ${{ env.AWS_REGION }}
           role-to-assume: ${{ secrets.AWS_CI_ROLE_ARN }}
@@ -237,7 +237,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: AWS authentication
-        uses: aws-actions/configure-aws-credentials@8c3f20df09ac63af7b3ae3d7c91f105f857d8497
+        uses: aws-actions/configure-aws-credentials@010d0da01d0b5a38af31e9c3470dbfdabdecca3a
         with:
           aws-region: ${{ env.AWS_REGION }}
           role-to-assume: ${{ secrets.AWS_CI_ROLE_ARN }}

.github/workflows/release_notes.yml

@@ -22,9 +22,9 @@ jobs:
       - name: Checkout
         uses: actions/checkout@v4
         with:
-            # With the default value of 1, there are corner cases where tj-actions/changed-files
-            # fails with a `no merge base` error
-            fetch-depth: 0
+          # With the default value of 1, there are corner cases where tj-actions/changed-files
+          # fails with a `no merge base` error
+          fetch-depth: 0
 
       - name: Get release note files
         id: changed-files
@@ -35,5 +35,10 @@ jobs:
       - name: Check release notes
         if: steps.changed-files.outputs.any_changed == 'false' && !contains( github.event.pull_request.labels.*.name, 'ignore-for-release-notes')
         run: |
+          # Check if any of the commit messages contain tags ci/docs/test
+          if git log --pretty=%s origin/main..HEAD | grep -E '^(ci:|docs:|test:)' > /dev/null; then
+            echo "Skipping release note check for commits with 'ci:', 'docs:', or 'test:' tags."
+          else
             echo "::error::The release notes file is missing, please add one or attach the label 'ignore-for-release-notes' to this PR."
             exit 1
+          fi

.github/workflows/tests_preview.yml

@@ -29,7 +29,7 @@ jobs:
   black:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
 
       - uses: actions/setup-python@v4
         with:
@@ -109,7 +109,7 @@ jobs:
           - macos-latest
     runs-on: ${{ matrix.os }}
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
 
       - uses: actions/setup-python@v4
         with:
@@ -156,19 +156,14 @@ jobs:
   integration-tests-linux:
     name: Integration / ubuntu-latest
     needs: unit-tests
-    strategy:
-      fail-fast: false
-      matrix:
-        os:
-          - ubuntu-latest
-    runs-on: ${{ matrix.os }}
+    runs-on: ubuntu-latest
     services:
       tika:
         image: apache/tika:2.9.0.0
         ports:
           - 9998:9998
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
 
       - uses: actions/setup-python@v4
         with:
@@ -222,7 +217,7 @@ jobs:
     needs: unit-tests
     runs-on: macos-latest
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
 
       - uses: actions/setup-python@v4
         with:
@@ -280,7 +275,7 @@ jobs:
     needs: unit-tests
     runs-on: windows-latest
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
 
       - uses: actions/setup-python@v4
         with:

README.md

@@ -106,7 +106,7 @@ Then move into the cloned folder and install the project with `pip`, including t
 cd haystack && pip install -e '.[dev]'
 ```
 
-If you want to contribute to the Haystack repo, check our [Contributor Guidelines](#💙-contributing) first.
+If you want to contribute to the Haystack repo, check our [Contributor Guidelines](https://github.com/deepset-ai/haystack/blob/main/CONTRIBUTING.md) first.
 
 See the list of [dependencies](https://github.com/deepset-ai/haystack/blob/main/pyproject.toml) to check which ones you want to install (for example, `[all]`, `[dev]`, or other).
 

haystack/nodes/answer_generator/openai.py

@@ -1,6 +1,7 @@
 import logging
 import os
 from typing import List, Optional, Tuple, Union
+import warnings
 
 from haystack import Document
 from haystack.environment import HAYSTACK_REMOTE_API_TIMEOUT_SEC
@@ -21,6 +22,10 @@
 
 class OpenAIAnswerGenerator(BaseGenerator):
     """
+    This component is now deprecated and will be removed in future versions.
+    Use `PromptNode` instead of `OpenAIAnswerGenerator`,
+    as explained in https://haystack.deepset.ai/tutorials/22_pipeline_with_promptnode.
+
     Uses the GPT-3 models from the OpenAI API to generate Answers based on the Documents it receives.
     The Documents can come from a Retriever or you can supply them manually.
 
@@ -109,6 +114,12 @@ def __init__(
         :param openai_organization: The OpenAI-Organization ID, defaults to `None`. For more details, see see OpenAI
         [documentation](https://platform.openai.com/docs/api-reference/requesting-organization).
         """
+        warnings.warn(
+            "`OpenAIAnswerGenerator component is deprecated and will be removed in future versions. Use `PromptNode` "
+            "instead of `OpenAIAnswerGenerator`.",
+            category=DeprecationWarning,
+        )
+
         super().__init__(progress_bar=progress_bar)
         if (examples is None and examples_context is not None) or (examples is not None and examples_context is None):
             logger.warning(

haystack/nodes/prompt/invocation_layer/chatgpt.py

@@ -1,10 +1,17 @@
 import logging
-from typing import Optional, List, Dict, Union, Any
+from typing import Any, Dict, List, Optional, Union
 
 from haystack.nodes.prompt.invocation_layer.handlers import DefaultTokenStreamingHandler, TokenStreamingHandler
 from haystack.nodes.prompt.invocation_layer.open_ai import OpenAIInvocationLayer
 from haystack.nodes.prompt.invocation_layer.utils import has_azure_parameters
-from haystack.utils.openai_utils import openai_request, _check_openai_finish_reason, count_openai_tokens_messages
+from haystack.utils.openai_utils import (
+    _check_openai_finish_reason,
+    check_openai_async_policy_violation,
+    check_openai_policy_violation,
+    count_openai_tokens_messages,
+    openai_async_request,
+    openai_request,
+)
 
 logger = logging.getLogger(__name__)
 
@@ -43,45 +50,6 @@ def __init__(
         """
         super().__init__(api_key, model_name_or_path, max_length, api_base=api_base, **kwargs)
 
-    def _execute_openai_request(
-        self, prompt: Union[str, List[Dict]], base_payload: Dict, kwargs_with_defaults: Dict, stream: bool
-    ):
-        """
-        For more details, see [OpenAI ChatGPT API reference](https://platform.openai.com/docs/api-reference/chat).
-        """
-        if isinstance(prompt, str):
-            messages = [{"role": "user", "content": prompt}]
-        elif isinstance(prompt, list) and len(prompt) > 0 and isinstance(prompt[0], dict):
-            messages = prompt
-        else:
-            raise ValueError(
-                f"The prompt format is different than what the model expects. "
-                f"The model {self.model_name_or_path} requires either a string or messages in the ChatML format. "
-                f"For more details, see this [GitHub discussion](https://github.com/openai/openai-python/blob/main/chatml.md)."
-            )
-        extra_payload = {"messages": messages}
-        payload = {**base_payload, **extra_payload}
-        if not stream:
-            response = openai_request(url=self.url, headers=self.headers, payload=payload)
-            _check_openai_finish_reason(result=response, payload=payload)
-            assistant_response = [choice["message"]["content"].strip() for choice in response["choices"]]
-        else:
-            response = openai_request(
-                url=self.url, headers=self.headers, payload=payload, read_response=False, stream=True
-            )
-            handler: TokenStreamingHandler = kwargs_with_defaults.pop("stream_handler", DefaultTokenStreamingHandler())
-            assistant_response = self._process_streaming_response(response=response, stream_handler=handler)
-
-        # Although ChatGPT generates text until stop words are encountered, unfortunately it includes the stop word
-        # We want to exclude it to be consistent with other invocation layers
-        if "stop" in kwargs_with_defaults and kwargs_with_defaults["stop"] is not None:
-            stop_words = kwargs_with_defaults["stop"]
-            for idx, _ in enumerate(assistant_response):
-                for stop_word in stop_words:
-                    assistant_response[idx] = assistant_response[idx].replace(stop_word, "").strip()
-
-        return assistant_response
-
     def _extract_token(self, event_data: Dict[str, Any]):
         delta = event_data["choices"][0]["delta"]
         if "content" in delta:
@@ -141,3 +109,109 @@ def supports(cls, model_name_or_path: str, **kwargs) -> bool:
             and not "gpt-3.5-turbo-instruct" in model_name_or_path
         )
         return valid_model and not has_azure_parameters(**kwargs)
+
+    async def ainvoke(self, *args, **kwargs):
+        """
+        Invokes a prompt on the model. Based on the model, it takes in a prompt (or either a prompt or a list of messages)
+        and returns a list of responses using a REST invocation.
+
+        :return: The responses are being returned.
+
+        Note: Only kwargs relevant to OpenAI are passed to OpenAI rest API. Others kwargs are ignored.
+        For more details, see OpenAI [documentation](https://platform.openai.com/docs/api-reference/completions/create).
+        """
+        prompt, base_payload, kwargs_with_defaults, stream, moderation = self._prepare_invoke(*args, **kwargs)
+
+        if moderation and await check_openai_async_policy_violation(input=prompt, headers=self.headers):
+            logger.info("Prompt '%s' will not be sent to OpenAI due to potential policy violation.", prompt)
+            return []
+
+        if isinstance(prompt, str):
+            messages = [{"role": "user", "content": prompt}]
+        elif isinstance(prompt, list) and len(prompt) > 0 and isinstance(prompt[0], dict):
+            messages = prompt
+        else:
+            raise ValueError(
+                f"The prompt format is different than what the model expects. "
+                f"The model {self.model_name_or_path} requires either a string or messages in the ChatML format. "
+                f"For more details, see this [GitHub discussion](https://github.com/openai/openai-python/blob/main/chatml.md)."
+            )
+        extra_payload = {"messages": messages}
+        payload = {**base_payload, **extra_payload}
+        if not stream:
+            response = await openai_async_request(url=self.url, headers=self.headers, payload=payload)
+            _check_openai_finish_reason(result=response, payload=payload)
+            assistant_response = [choice["message"]["content"].strip() for choice in response["choices"]]
+        else:
+            response = await openai_async_request(
+                url=self.url, headers=self.headers, payload=payload, read_response=False, stream=True
+            )
+            handler: TokenStreamingHandler = kwargs_with_defaults.pop("stream_handler", DefaultTokenStreamingHandler())
+            assistant_response = self._process_streaming_response(response=response, stream_handler=handler)
+
+        # Although ChatGPT generates text until stop words are encountered, unfortunately it includes the stop word
+        # We want to exclude it to be consistent with other invocation layers
+        if "stop" in kwargs_with_defaults and kwargs_with_defaults["stop"] is not None:
+            stop_words = kwargs_with_defaults["stop"]
+            for idx, _ in enumerate(assistant_response):
+                for stop_word in stop_words:
+                    assistant_response[idx] = assistant_response[idx].replace(stop_word, "").strip()
+
+        if moderation and await check_openai_async_policy_violation(input=assistant_response, headers=self.headers):
+            logger.info("Response '%s' will not be returned due to potential policy violation.", assistant_response)
+            return []
+
+        return assistant_response
+
+    def invoke(self, *args, **kwargs):
+        """
+        Invokes a prompt on the model. Based on the model, it takes in a prompt (or either a prompt or a list of messages)
+        and returns a list of responses using a REST invocation.
+
+        :return: The responses are being returned.
+
+        Note: Only kwargs relevant to OpenAI are passed to OpenAI rest API. Others kwargs are ignored.
+        For more details, see OpenAI [documentation](https://platform.openai.com/docs/api-reference/completions/create).
+        """
+        prompt, base_payload, kwargs_with_defaults, stream, moderation = self._prepare_invoke(*args, **kwargs)
+
+        if moderation and check_openai_policy_violation(input=prompt, headers=self.headers):
+            logger.info("Prompt '%s' will not be sent to OpenAI due to potential policy violation.", prompt)
+            return []
+
+        if isinstance(prompt, str):
+            messages = [{"role": "user", "content": prompt}]
+        elif isinstance(prompt, list) and len(prompt) > 0 and isinstance(prompt[0], dict):
+            messages = prompt
+        else:
+            raise ValueError(
+                f"The prompt format is different than what the model expects. "
+                f"The model {self.model_name_or_path} requires either a string or messages in the ChatML format. "
+                f"For more details, see this [GitHub discussion](https://github.com/openai/openai-python/blob/main/chatml.md)."
+            )
+        extra_payload = {"messages": messages}
+        payload = {**base_payload, **extra_payload}
+        if not stream:
+            response = openai_request(url=self.url, headers=self.headers, payload=payload)
+            _check_openai_finish_reason(result=response, payload=payload)
+            assistant_response = [choice["message"]["content"].strip() for choice in response["choices"]]
+        else:
+            response = openai_request(
+                url=self.url, headers=self.headers, payload=payload, read_response=False, stream=True
+            )
+            handler: TokenStreamingHandler = kwargs_with_defaults.pop("stream_handler", DefaultTokenStreamingHandler())
+            assistant_response = self._process_streaming_response(response=response, stream_handler=handler)
+
+        # Although ChatGPT generates text until stop words are encountered, unfortunately it includes the stop word
+        # We want to exclude it to be consistent with other invocation layers
+        if "stop" in kwargs_with_defaults and kwargs_with_defaults["stop"] is not None:
+            stop_words = kwargs_with_defaults["stop"]
+            for idx, _ in enumerate(assistant_response):
+                for stop_word in stop_words:
+                    assistant_response[idx] = assistant_response[idx].replace(stop_word, "").strip()
+
+        if moderation and check_openai_policy_violation(input=assistant_response, headers=self.headers):
+            logger.info("Response '%s' will not be returned due to potential policy violation.", assistant_response)
+            return []
+
+        return assistant_response