Allow configuration of log level in REST API via ENV (#541)

* configure log level via env. adjust debug messages * pin faiss version
deepset-ai · Nov 4, 2020 · 46fac41 · 46fac41
1 parent df13a68
commit 46fac41
Show file tree

Hide file tree

Showing 5 changed files with 8 additions and 7 deletions.
diff --git a/haystack/finder.py b/haystack/finder.py
@@ -51,6 +51,8 @@ def get_answers(self, question: str, top_k_reader: int = 1, top_k_retriever: int
 
         # 1) Apply retriever(with optional filters) to get fast candidate documents
         documents = self.retriever.retrieve(question, filters=filters, top_k=top_k_retriever, index=index)
+        logger.info(f"Got {len(documents)} candidates from retriever")
+        logger.debug(f"Retrieved document IDs: {[doc.id for doc in documents]}")
 
         if len(documents) == 0:
             logger.info("Retriever did not return any documents. Skipping reader ...")

diff --git a/haystack/retriever/sparse.py b/haystack/retriever/sparse.py
@@ -57,8 +57,6 @@ def retrieve(self, query: str, filters: dict = None, top_k: int = 10, index: str
             index = self.document_store.index
 
         documents = self.document_store.query(query, filters, top_k, self.custom_query, index)
-        logger.info(f"Got {len(documents)} candidates from retriever")
-
         return documents
 
 
@@ -73,8 +71,6 @@ def retrieve(self, query: str, filters: dict = None, top_k: int = 10, index: str
             index = self.document_store.index
         documents = self.document_store.query(query=None, filters=filters, top_k=top_k,
                                               custom_query=self.custom_query, index=index)
-        logger.info(f"Got {len(documents)} candidates from retriever")
-
         return documents
 
 # TODO make Paragraph generic for configurable units of text eg, pages, paragraphs, or split by a char_limit

diff --git a/requirements.txt b/requirements.txt
@@ -16,7 +16,7 @@ python-multipart
 python-docx
 sqlalchemy_utils
 # for using FAISS with GPUs, install faiss-gpu
-faiss-cpu; sys_platform != 'win32' and sys_platform != 'cygwin'
+faiss-cpu==1.6.3; sys_platform != 'win32' and sys_platform != 'cygwin'
 tika
 uvloop; sys_platform != 'win32' and sys_platform != 'cygwin'
 httptools

diff --git a/rest_api/config.py b/rest_api/config.py
@@ -59,5 +59,6 @@
     VALID_LANGUAGES = ast.literal_eval(VALID_LANGUAGES)
 
 # Monitoring
+LOG_LEVEL = os.getenv("LOG_LEVEL", "INFO")
 APM_SERVER = os.getenv("APM_SERVER", None)
 APM_SERVICE_NAME = os.getenv("APM_SERVICE_NAME", "haystack-backend")
diff --git a/rest_api/controller/search.py b/rest_api/controller/search.py
@@ -14,7 +14,7 @@
     RETRIEVER_TYPE, EMBEDDING_MODEL_PATH, USE_GPU, READER_MODEL_PATH, BATCHSIZE, CONTEXT_WINDOW_SIZE, \
     TOP_K_PER_CANDIDATE, NO_ANS_BOOST, MAX_PROCESSES, MAX_SEQ_LEN, DOC_STRIDE, CONCURRENT_REQUEST_PER_WORKER, \
     FAQ_QUESTION_FIELD_NAME, EMBEDDING_MODEL_FORMAT, READER_TYPE, READER_TOKENIZER, GPU_NUMBER, NAME_FIELD_NAME, \
-    VECTOR_SIMILARITY_METRIC, CREATE_INDEX
+    VECTOR_SIMILARITY_METRIC, CREATE_INDEX, LOG_LEVEL
 
 from rest_api.controller.request import Question
 from rest_api.controller.response import Answers, AnswersToIndividualQuestion
@@ -28,7 +28,9 @@
 from haystack.retriever.sparse import ElasticsearchRetriever, ElasticsearchFilterOnlyRetriever
 from haystack.retriever.dense import EmbeddingRetriever
 
-logger = logging.getLogger(__name__)
+logger = logging.getLogger('haystack')
+logger.setLevel(LOG_LEVEL)
+
 router = APIRouter()
 
 # Init global components: DocumentStore, Retriever, Reader, Finder