feat: preview extra (#5869)

* copy the deps list over from haystack-ai

* fix lazyimport usage

* keep jinja and openai

* fix ci

* reno

* separate out preview unit tests

* fix import error message for tika

* tika

* add preview to all

* wrap torch

* remove comment

* unwrap openai and jinja

.github/workflows/tests.yml

@@ -114,7 +114,6 @@ jobs:
           - nodes
           - agents
           - cli
-          - preview
           - prompt
           - pipelines
           - utils
@@ -184,6 +183,70 @@ jobs:
                 - "branch:${{ github.ref_name }}"
                 - "url:https://github.com/${{ github.repository }}/actions/runs/${{ github.run_id }}"
 
+  unit-tests-preview:
+    name: Unit / preview / ${{ matrix.os }}
+    needs: black
+    strategy:
+      fail-fast: false
+      matrix:
+        os:
+          - ubuntu-latest
+          - windows-latest
+    runs-on: ${{ matrix.os }}
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-python@v4
+        with:
+          python-version: ${{ env.PYTHON_VERSION }}
+
+      - name: Install Haystack
+        run: pip install .[preview,dev] langdetect transformers[torch,sentencepiece]==4.32.1 sentence-transformers>=2.2.0 pypdf openai-whisper tika
+
+      - name: Run
+        run: pytest --cov-report xml:coverage.xml --cov="haystack" -m "unit" test/preview
+
+      - name: Coveralls Parallel
+        # We upload only coverage for ubuntu as handling both os
+        # complicates the workflow too much for little to no gain
+        if: matrix.os == 'ubuntu-latest'
+        uses: coverallsapp/github-action@v2
+        with:
+          path-to-lcov: coverage.xml
+          flag-name: preview
+          parallel: true
+
+      - name: Calculate alert data
+        id: calculator
+        shell: bash
+        if: (success() || failure()) && github.ref_name == 'main'
+        run: |
+          if [ "${{ job.status }}" = "success" ]; then
+            echo "alert_type=success" >> "$GITHUB_OUTPUT";
+          else
+            echo "alert_type=error" >> "$GITHUB_OUTPUT";
+          fi
+
+      - name: Send event to Datadog
+        if: (success() || failure()) && github.ref_name == 'main'
+        uses: masci/datadog@v1
+        with:
+          api-key: ${{ secrets.CORE_DATADOG_API_KEY }}
+          api-url: https://api.datadoghq.eu
+          events: |
+            - title: "${{ github.workflow }} workflow"
+              text: "Job ${{ github.job }} in branch ${{ github.ref_name }}"
+              alert_type: "${{ steps.calculator.outputs.alert_type }}"
+              source_type_name: "Github"
+              host: ${{ github.repository_owner }}
+              tags:
+                - "project:${{ github.repository }}"
+                - "job:${{ github.job }}"
+                - "run_id:${{ github.run_id }}"
+                - "workflow:${{ github.workflow }}"
+                - "branch:${{ github.ref_name }}"
+                - "url:https://github.com/${{ github.repository }}/actions/runs/${{ github.run_id }}"
+
   upload-coverage:
     needs: unit-tests
     runs-on: ubuntu-latest
@@ -852,8 +915,7 @@ jobs:
           sudo apt install ffmpeg  # for local Whisper tests
 
       - name: Install Haystack
-        # FIXME Use haystack-ai dependency list
-        run: pip install .[dev,inference,file-conversion] langdetect
+        run: pip install .[dev,preview] langdetect transformers[torch,sentencepiece]==4.32.1 sentence-transformers>=2.2.0 pypdf openai-whisper tika
 
       - name: Run tests
         run: |

haystack/preview/components/audio/whisper_local.py

@@ -3,10 +3,12 @@ from typing import List, Optional, Dict, Any, Union, BinaryIO, Literal, get_args
 import logging
 from pathlib import Path
 
-import torch
-import whisper
-
 from haystack.preview import component, Document, default_to_dict, default_from_dict, ComponentError
+from haystack.preview.lazy_imports import LazyImport
+
+with LazyImport("Run 'pip install openai-whisper'") as whisper_import:
+    import torch
+    import whisper
 
 
 logger = logging.getLogger(__name__)
@@ -38,6 +40,7 @@ class LocalWhisperTranscriber:
             - `large-v2`
         :param device: Name of the torch device to use for inference. If None, CPU is used.
         """
+        whisper_import.check()
         if model_name_or_path not in get_args(WhisperLocalModel):
             raise ValueError(
                 f"Model name '{model_name_or_path}' not recognized. Choose one among: "

haystack/preview/components/file_converters/tika.py

@@ -6,7 +6,7 @@ from haystack.preview.lazy_imports import LazyImport
 from haystack.preview import component, Document, default_to_dict, default_from_dict
 
 
-with LazyImport("Run 'pip install farm-haystack[file-conversion]' or 'pip install tika'") as tika_import:
+with LazyImport("Run 'pip install tika'") as tika_import:
     from tika import parser as tika_parser
 
 logger = logging.getLogger(__name__)

haystack/preview/components/file_converters/txt.py

@@ -8,7 +8,7 @@ from tqdm import tqdm
 from haystack.preview.lazy_imports import LazyImport
 from haystack.preview import Document, component, default_to_dict, default_from_dict
 
-with LazyImport("Run 'pip install farm-haystack[preprocessing]'") as langdetect_import:
+with LazyImport("Run 'pip install langdetect'") as langdetect_import:
     import langdetect
 
 

haystack/preview/components/readers/extractive.py

@@ -6,7 +6,9 @@ import warnings
 from haystack.preview import component, default_from_dict, default_to_dict, ComponentError, Document, ExtractedAnswer
 from haystack.preview.lazy_imports import LazyImport
 
-with LazyImport(message="Run 'pip install farm-haystack[inference]'") as torch_and_transformers_import:
+with LazyImport(
+    "Run 'pip install transformers[torch,sentencepiece]==4.32.1 sentence-transformers>=2.2.0'"
+) as torch_and_transformers_import:
     from transformers import AutoModelForQuestionAnswering, AutoTokenizer
     from tokenizers import Encoding
     import torch

haystack/preview/embedding_backends/sentence_transformers_backend.py

@@ -2,7 +2,7 @@ from typing import List, Optional, Union, Dict
 
 from haystack.preview.lazy_imports import LazyImport
 
-with LazyImport(message="Run 'pip install farm-haystack[inference]'") as sentence_transformers_import:
+with LazyImport(message="Run 'pip install sentence-transformers>=2.2.0'") as sentence_transformers_import:
     from sentence_transformers import SentenceTransformer
 
 

pyproject.toml

@@ -78,13 +78,6 @@ dependencies = [
   # Schema validation
   "jsonschema",
 
-  # Preview
-  "canals==0.8.0",
-  "openai",
-  "Jinja2",
-  "openai-whisper",  # FIXME https://github.com/deepset-ai/haystack/issues/5731
-  "pypdf",
-
   # Agent events
   "events",
 
@@ -92,6 +85,18 @@ dependencies = [
 ]
 
 [project.optional-dependencies]
+preview = [
+  "canals==0.8.0",
+  "requests",
+  "pandas",
+  "rank_bm25",
+  "tqdm",
+  "tenacity",
+  "lazy-imports",
+
+  "Jinja2",
+  "openai",
+]
 inference = [
   "transformers[torch,sentencepiece]==4.32.1",
   "sentence-transformers>=2.2.0",  # See haystack/nodes/retriever/_embedding_encoder.py, _SentenceTransformersEmbeddingEncoder
@@ -240,11 +245,11 @@ formatting = [
 ]
 
 all = [
-  "farm-haystack[inference,docstores,audio,crawler,preprocessing,file-conversion,pdf,ocr,ray,onnx,beir,metrics,aws]",
+  "farm-haystack[inference,docstores,audio,crawler,preprocessing,file-conversion,pdf,ocr,ray,onnx,beir,metrics,aws,preview]",
 ]
 all-gpu = [
   # beir is incompatible with faiss-gpu: https://github.com/beir-cellar/beir/issues/71
-  "farm-haystack[inference,docstores-gpu,audio,crawler,preprocessing,file-conversion,pdf,ocr,ray,onnx-gpu,metrics,aws]",
+  "farm-haystack[inference,docstores-gpu,audio,crawler,preprocessing,file-conversion,pdf,ocr,ray,onnx-gpu,metrics,aws,preview]",
 ]
 
 [project.scripts]

releasenotes/notes/preview-extra-6dfdca55d17cbc7f.yaml

@@ -0,0 +1,4 @@
+preview:
+  - |
+    Create a dedicated dependency list for the preview package, `farm-haystack[preview]`.
+    Using `haystack-ai` is still the recommended way to test Haystack 2.0.