feat(bugfixes): bug fixes for ingestion and summary (#72)

* handle exception * added prompt for summarization --------- Co-authored-by: Alain Krok <[email protected]> Co-authored-by: Dinesh Sajwan <[email protected]>
awslabs · Nov 2, 2023 · 338ac81 · 338ac81
1 parent e2085af
commit 338ac81
Show file tree

Hide file tree

Showing 6 changed files with 39 additions and 15 deletions.
diff --git a/lambda/aws-rag-appsync-stepfn-opensearch/s3_file_transformer/src/helpers/s3inmemoryloader.py b/lambda/aws-rag-appsync-stepfn-opensearch/s3_file_transformer/src/helpers/s3inmemoryloader.py
@@ -47,14 +47,18 @@ def load(self) -> str:
             obj = s3.Object(self.bucket, self.key)
             encodedpdf = obj.get()['Body'].read()
             pdfFile = PdfReader(BytesIO(encodedpdf))
+            # read pdf
+            raw_text = []
+            for page in pdfFile.pages:
+                raw_text.append(page.extract_text())
+            return '\n'.join(raw_text)
         except s3.meta.client.exceptions.NoSuchBucket as exception:
             logger.exception('NoSuchBucket')
             return ""
         except s3.meta.client.exceptions.NoSuchKey as exception:
             logger.exception('NoSuchKey')
             return ""
-        # read pdf
-        raw_text = []
-        for page in pdfFile.pages:
-            raw_text.append(page.extract_text())
-        return '\n'.join(raw_text)
+        except Exception as exception:
+            logger.exception(f"Reason: {exception}")
+            return ""
+
diff --git a/lambda/aws-summarization-appsync-stepfn/document_reader/s3inmemoryloader.py b/lambda/aws-summarization-appsync-stepfn/document_reader/s3inmemoryloader.py
@@ -48,14 +48,17 @@ def load(self) -> str:
             obj = s3.Object(self.bucket, self.key)
             encodedpdf = obj.get()['Body'].read()
             pdfFile = PdfReader(BytesIO(encodedpdf))
+            # read pdf
+            raw_text = []
+            for page in pdfFile.pages:
+                raw_text.append(page.extract_text())
+            return '\n'.join(raw_text)
         except s3.meta.client.exceptions.NoSuchBucket as exception:
             logger.exception('NoSuchBucket')
             return ""
         except s3.meta.client.exceptions.NoSuchKey as exception:
             logger.exception('NoSuchKey')
             return ""
-        # read pdf
-        raw_text = []
-        for page in pdfFile.pages:
-            raw_text.append(page.extract_text())
-        return '\n'.join(raw_text)
+        except Exception as exception:
+            logger.exception(f"Reason: {exception}")
+            return ""
diff --git a/lambda/aws-summarization-appsync-stepfn/summary_generator/lambda.py b/lambda/aws-summarization-appsync-stepfn/summary_generator/lambda.py
@@ -15,7 +15,7 @@
 
 from langchain.llms.bedrock import Bedrock
 from update_summary_status import updateSummaryJobStatus
-
+from langchain import PromptTemplate
 # external files
 from langchain.docstore.document import Document
 from langchain.chains.summarize import load_summarize_chain
@@ -38,6 +38,15 @@
 chain_type = os.environ["SUMMARY_LLM_CHAIN_TYPE"]
 
 aws_region = boto3.Session().region_name
+
+params = {
+        "max_tokens_to_sample": 4000,
+        "temperature": 0, 
+        "top_k": 250,
+        "top_p": 1,
+        "stop_sequences": ["\\n\\nHuman:"],
+    }
+
 bedrock_client = boto3.client(
         service_name='bedrock-runtime', 
         region_name=aws_region,
@@ -68,6 +77,8 @@ def handler(event, context: LambdaContext)-> dict:
     summary_llm = Bedrock(
         model_id="anthropic.claude-v2",
         client=bedrock_client,
+        model_kwargs=params,
+        streaming=False,
     )
 
     redis_host = os.environ.get("REDIS_HOST", "N/A")
@@ -133,10 +144,16 @@ def generate_summary(_summary_llm,chain_type,inputFile)-> str:
 
     logger.info(f" Using chain_type as {chain_type} for the document")    
     docs = [Document(page_content=inputFile)]
+    template = """\n\nHuman: Please read the text:\n{text}\n
+    Summarize the text in 300 words: 
+    \n\nAssistant:"""
+    prompt = PromptTemplate(template=template, input_variables=["text"])
+
     chain = load_summarize_chain(
                 _summary_llm, 
                 chain_type=chain_type, 
-                verbose=False
+                verbose=False,
+                prompt=prompt
                 )
     return chain.run(docs)
 
diff --git a/src/patterns/gen-ai/aws-qa-appsync-opensearch/index.ts b/src/patterns/gen-ai/aws-qa-appsync-opensearch/index.ts
@@ -386,7 +386,7 @@ export class QaAppsyncOpensearch extends Construct {
     }));
 
     const enableOperationalMetric = props.enableOperationalMetric || true;
-    const solution_id = 'QaAppsyncOpensearch_'+id;
+    const solution_id = 'genai_cdk_'+id;
 
     if (enableOperationalMetric) {
       question_answering_function.addEnvironment(

diff --git a/src/patterns/gen-ai/aws-rag-appsync-stepfn-opensearch/index.ts b/src/patterns/gen-ai/aws-rag-appsync-stepfn-opensearch/index.ts
@@ -546,7 +546,7 @@ export class RagAppsyncStepfnOpensearch extends Construct {
     }));
 
     const enableOperationalMetric = props.enableOperationalMetric || true;
-    const solution_id = 'RagAppsyncStepfnOpensearch_'+id;
+    const solution_id = 'genai_cdk_'+id;
 
     if (enableOperationalMetric) {
       embeddings_job_function.addEnvironment(

diff --git a/src/patterns/gen-ai/aws-summarization-appsync-stepfn/index.ts b/src/patterns/gen-ai/aws-summarization-appsync-stepfn/index.ts
@@ -485,7 +485,7 @@ export class SummarizationAppsyncStepfn extends Construct {
     );
 
     const enableOperationalMetric = props.enableOperationalMetric || true;
-    const solution_id = 'SummarizationAppsyncStepfn_'+id;
+    const solution_id = 'genai_cdk_'+id;
 
     if (enableOperationalMetric) {
       documentReaderLambda.addEnvironment(