This is an automated email from the ASF dual-hosted git repository.
ferruzzi pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/airflow.git
The following commit(s) were added to refs/heads/main by this push:
new 4d0c7242bc Small refactor for example_bedrock_knowledge_base.py
(#39672)
4d0c7242bc is described below
commit 4d0c7242bcc8a403c03edd993b8b445a51720492
Author: D. Ferruzzi <ferru...@amazon.com>
AuthorDate: Thu May 16 14:43:48 2024 -0700
Small refactor for example_bedrock_knowledge_base.py (#39672)
- Renamed example_bedrock_retrieve_and_generate.py, a more accurate and
descriptive name
- Added an example for invoking Claude models and added that to the docs
---
.../operators/bedrock.rst | 34 ++++++++++++++--------
.../operators/opensearchserverless.rst | 2 +-
...py => example_bedrock_retrieve_and_generate.py} | 27 ++++++++++++-----
3 files changed, 43 insertions(+), 20 deletions(-)
diff --git a/docs/apache-airflow-providers-amazon/operators/bedrock.rst
b/docs/apache-airflow-providers-amazon/operators/bedrock.rst
index 1808f5138c..daf9301565 100644
--- a/docs/apache-airflow-providers-amazon/operators/bedrock.rst
+++ b/docs/apache-airflow-providers-amazon/operators/bedrock.rst
@@ -46,7 +46,10 @@ Invoke an existing Amazon Bedrock Model
To invoke an existing Amazon Bedrock model, you can use
:class:`~airflow.providers.amazon.aws.operators.bedrock.BedrockInvokeModelOperator`.
-Note that every model family has different input and output formats.
+Note that every model family has different input and output formats. Some
examples are included below, but
+for details on the different formats, see
+`Inference parameters for foundation models
<https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters.html>`__
+
For example, to invoke a Meta Llama model you would use:
.. exampleinclude:: /../../tests/system/providers/amazon/aws/example_bedrock.py
@@ -63,7 +66,14 @@ To invoke an Amazon Titan model you would use:
:start-after: [START howto_operator_invoke_titan_model]
:end-before: [END howto_operator_invoke_titan_model]
-For details on the different formats, see `Inference parameters for foundation
models
<https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters.html>`__
+To invoke a Claude V2 model using the Completions API you would use:
+
+.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_retrieve_and_generate.py
+ :language: python
+ :dedent: 4
+ :start-after: [START howto_operator_invoke_claude_model]
+ :end-before: [END howto_operator_invoke_claude_model]
+
.. _howto/operator:BedrockCustomizeModelOperator:
@@ -119,7 +129,7 @@ To create an Amazon Bedrock Knowledge Base, you can use
For more information on which models support embedding data into a vector
store, see
https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base-supported.html
-.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_knowledge_base.py
+.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_retrieve_and_generate.py
:language: python
:dedent: 4
:start-after: [START howto_operator_bedrock_create_knowledge_base]
@@ -132,7 +142,7 @@ Delete an Amazon Bedrock Knowledge Base
Deleting a Knowledge Base is a simple boto API call and can be done in a
TaskFlow task like the example below.
-.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_knowledge_base.py
+.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_retrieve_and_generate.py
:language: python
:start-after: [START howto_operator_bedrock_delete_knowledge_base]
:end-before: [END howto_operator_bedrock_delete_knowledge_base]
@@ -145,7 +155,7 @@ Create an Amazon Bedrock Data Source
To create an Amazon Bedrock Data Source, you can use
:class:`~airflow.providers.amazon.aws.operators.bedrock.BedrockCreateDataSourceOperator`.
-.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_knowledge_base.py
+.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_retrieve_and_generate.py
:language: python
:dedent: 4
:start-after: [START howto_operator_bedrock_create_data_source]
@@ -158,7 +168,7 @@ Delete an Amazon Bedrock Data Source
Deleting a Data Source is a simple boto API call and can be done in a TaskFlow
task like the example below.
-.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_knowledge_base.py
+.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_retrieve_and_generate.py
:language: python
:start-after: [START howto_operator_bedrock_delete_data_source]
:end-before: [END howto_operator_bedrock_delete_data_source]
@@ -171,7 +181,7 @@ Ingest data into an Amazon Bedrock Data Source
To add data from an Amazon S3 bucket into an Amazon Bedrock Data Source, you
can use
:class:`~airflow.providers.amazon.aws.operators.bedrock.BedrockIngestDataOperator`.
-.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_knowledge_base.py
+.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_retrieve_and_generate.py
:language: python
:dedent: 4
:start-after: [START howto_operator_bedrock_ingest_data]
@@ -191,7 +201,7 @@ would like to pass the results through an LLM in order to
generate a text respon
For more information on which models support retrieving information from a
knowledge base, see
https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base-supported.html
-.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_knowledge_base.py
+.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_retrieve_and_generate.py
:language: python
:dedent: 4
:start-after: [START howto_operator_bedrock_retrieve]
@@ -213,7 +223,7 @@ NOTE: Support for "external sources" was added in boto
1.34.90
Example using an Amazon Bedrock Knowledge Base:
-.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_knowledge_base.py
+.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_retrieve_and_generate.py
:language: python
:dedent: 4
:start-after: [START howto_operator_bedrock_knowledge_base_rag]
@@ -221,7 +231,7 @@ Example using an Amazon Bedrock Knowledge Base:
Example using a PDF file in an Amazon S3 Bucket:
-.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_knowledge_base.py
+.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_retrieve_and_generate.py
:language: python
:dedent: 4
:start-after: [START howto_operator_bedrock_external_sources_rag]
@@ -268,7 +278,7 @@ Wait for an Amazon Bedrock Knowledge Base
To wait on the state of an Amazon Bedrock Knowledge Base until it reaches a
terminal state you can use
:class:`~airflow.providers.amazon.aws.sensors.bedrock.BedrockKnowledgeBaseActiveSensor`
-.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_knowledge_base.py
+.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_retrieve_and_generate.py
:language: python
:dedent: 4
:start-after: [START howto_sensor_bedrock_knowledge_base_active]
@@ -282,7 +292,7 @@ Wait for an Amazon Bedrock ingestion job to finish
To wait on the state of an Amazon Bedrock data ingestion job until it reaches
a terminal state you can use
:class:`~airflow.providers.amazon.aws.sensors.bedrock.BedrockIngestionJobSensor`
-.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_knowledge_base.py
+.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_retrieve_and_generate.py
:language: python
:dedent: 4
:start-after: [START howto_sensor_bedrock_ingest_data]
diff --git
a/docs/apache-airflow-providers-amazon/operators/opensearchserverless.rst
b/docs/apache-airflow-providers-amazon/operators/opensearchserverless.rst
index f947c91e30..4fb0b4db93 100644
--- a/docs/apache-airflow-providers-amazon/operators/opensearchserverless.rst
+++ b/docs/apache-airflow-providers-amazon/operators/opensearchserverless.rst
@@ -46,7 +46,7 @@ Wait for an Amazon OpenSearch Serverless Collection to become
active
To wait on the state of an Amazon Bedrock customize model job until it reaches
a terminal state you can use
:class:`~airflow.providers.amazon.aws.sensors.bedrock.OpenSearchServerlessCollectionActiveSensor`
-.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_knowledge_base.py
+.. exampleinclude::
/../../tests/system/providers/amazon/aws/example_bedrock_retrieve_and_generate.py
:language: python
:dedent: 4
:start-after: [START howto_sensor_opensearch_collection_active]
diff --git
a/tests/system/providers/amazon/aws/example_bedrock_knowledge_base.py
b/tests/system/providers/amazon/aws/example_bedrock_retrieve_and_generate.py
similarity index 96%
rename from tests/system/providers/amazon/aws/example_bedrock_knowledge_base.py
rename to
tests/system/providers/amazon/aws/example_bedrock_retrieve_and_generate.py
index b0a333382a..c544cee7f8 100644
--- a/tests/system/providers/amazon/aws/example_bedrock_knowledge_base.py
+++ b/tests/system/providers/amazon/aws/example_bedrock_retrieve_and_generate.py
@@ -44,6 +44,7 @@ from airflow.providers.amazon.aws.operators.bedrock import (
BedrockCreateDataSourceOperator,
BedrockCreateKnowledgeBaseOperator,
BedrockIngestDataOperator,
+ BedrockInvokeModelOperator,
BedrockRaGOperator,
BedrockRetrieveOperator,
)
@@ -61,12 +62,14 @@ from airflow.utils.helpers import chain
from airflow.utils.trigger_rule import TriggerRule
from tests.system.providers.amazon.aws.utils import SystemTestContextBuilder
-###########################################################################################################
+#######################################################################
# NOTE:
-# The account running this test must first manually request access to the
`Titan Embeddings G1 - Text`
-# and `Anthropic Claude v2.0` foundation models via the Bedrock console.
Gaining access to the models
-# can take 24 hours from the time of request.
-###########################################################################################################
+# Access to the following foundation model must be requested via
+# the Amazon Bedrock console and may take up to 24 hours to apply:
+#######################################################################
+
+CLAUDE_MODEL_ID = "anthropic.claude-v2"
+TITAN_MODEL_ID = "amazon.titan-embed-text-v1"
# Externally fetched variables:
ROLE_ARN_KEY = "ROLE_ARN"
@@ -462,11 +465,20 @@ with DAG(
)
# [END howto_sensor_opensearch_collection_active]
+ PROMPT = "What color is an orange?"
+ # [START howto_operator_invoke_claude_model]
+ invoke_claude_completions = BedrockInvokeModelOperator(
+ task_id="invoke_claude_completions",
+ model_id=CLAUDE_MODEL_ID,
+ input_data={"max_tokens_to_sample": 4000, "prompt": f"\n\nHuman:
{PROMPT}\n\nAssistant:"},
+ )
+ # [END howto_operator_invoke_claude_model]
+
# [START howto_operator_bedrock_create_knowledge_base]
create_knowledge_base = BedrockCreateKnowledgeBaseOperator(
task_id="create_knowledge_base",
name=knowledge_base_name,
-
embedding_model_arn=f"arn:aws:bedrock:{region_name}::foundation-model/amazon.titan-embed-text-v1",
+
embedding_model_arn=f"arn:aws:bedrock:{region_name}::foundation-model/{TITAN_MODEL_ID}",
role_arn=test_context[ROLE_ARN_KEY],
storage_config={
"type": "OPENSEARCH_SERVERLESS",
@@ -522,7 +534,7 @@ with DAG(
task_id="knowledge_base_rag",
input="Who was the CEO of Amazon on 2022?",
source_type="KNOWLEDGE_BASE",
-
model_arn=f"arn:aws:bedrock:{region_name}::foundation-model/anthropic.claude-v2",
+
model_arn=f"arn:aws:bedrock:{region_name}::foundation-model/{CLAUDE_MODEL_ID}",
knowledge_base_id=create_knowledge_base.output,
)
# [END howto_operator_bedrock_knowledge_base_rag]
@@ -552,6 +564,7 @@ with DAG(
create_vector_index(index_name=index_name, collection_id=collection,
region=region_name),
copy_data_to_s3(bucket=bucket_name),
# TEST BODY
+ invoke_claude_completions,
create_knowledge_base,
await_knowledge_base,
create_data_source,
