feat: Adding BedrockEncoder

[theanupllm]

User description

#173 An implementation for Bedrock embedding models.

Now Semantic router will support generating embeddings using AWS Bedrock. It integrates the BedrockEncoder into the encoder modules and adds it to the AutoEncoder class. Additionally, it sets up unit tests for BedrockEncoder functionalities, error handling, and creates a Jupyter notebook documentation for demonstrating its usage. Boto3 has been added as a dependency for AWS operations related to Bedrock.

---

PR Type

enhancement, documentation

---

Description

• Added BedrockEncoder class for embedding generation using AWS Bedrock.
• Integrated BedrockEncoder into the encoder modules and added it to the AutoEncoder class.
• Set up unit tests for BedrockEncoder functionalities and error handling.
• Created a Jupyter notebook documentation for demonstrating the usage of BedrockEncoder.
• Added boto3 as a dependency for AWS operations related to Bedrock.

---

Changes walkthrough 📝

	Relevant files
Enhancement	__init__.py Integrate BedrockEncoder into Encoder Modules                        semantic_router/encoders/init.py Imported BedrockEncoder in the module. Added BedrockEncoder to the list of available encoders. Added conditional initialization for BedrockEncoder in AutoEncoder class. +4/-0      bedrock.py Implement BedrockEncoder with AWS Integration                        semantic_router/encoders/bedrock.py Implemented BedrockEncoder class with methods for initialization, client setup, and embedding generation. Added error handling and environment variable management for AWS credentials and configuration. +250/-0  schema.py Add BEDROCK Encoder Type                                                                  semantic_router/schema.py Added BEDROCK as a new encoder type in the EncoderType enum. +1/-0      defaults.py Set Default Bedrock Embedding Model                                            semantic_router/utils/defaults.py Added default embedding model for Bedrock encoder. +5/-0
Tests	test_bedrock.py Add Unit Tests for BedrockEncoder                                                tests/unit/encoders/test_bedrock.py Added unit tests for BedrockEncoder covering initialization, embedding generation, and error handling. +116/-0
Documentation	bedrock.ipynb Documentation for BedrockEncoder Usage                                      docs/encoders/bedrock.ipynb Created a comprehensive Jupyter notebook demonstrating the usage of BedrockEncoder. +1323/-0
Configuration changes	pyproject.toml Add Boto3 Dependency for Bedrock Support                                  pyproject.toml Added boto3 as an optional dependency under the 'bedrock' extra. +2/-0

---

💡 PR-Agent usage:
Comment /help on the PR to get a list of all available PR-Agent tools and their descriptions

[github-actions]

PR Description updated to latest commit (73cf6b8)

[github-actions]

PR Review 🔍

⏱️ Estimated effort to review [1-5]	4, because the PR introduces a new encoder class with a significant amount of new functionality, including AWS integration and error handling. The complexity of the AWS client initialization and the embedding generation logic requires careful review to ensure security and functionality.
🧪 Relevant tests	Yes
⚡ Possible issues	Possible Bug: The chunk_strings function in bedrock.py uses a hardcoded MAX_WORDS value which might not be suitable for all types of documents. This could lead to incomplete or incorrect embeddings for longer texts.
	Error Handling: The error handling in the __call__ method of BedrockEncoder might not catch all potential exceptions from the AWS SDK, leading to unhandled exceptions in some cases.
🔒 Security concerns	Sensitive information exposure: The code handles AWS credentials and session tokens, which are sensitive. While the PR retrieves these from environment variables or parameters, ensuring these are not logged or exposed through exceptions is crucial.

Code feedback:

---

relevant file	semantic_router/encoders/bedrock.py
suggestion	Consider parameterizing MAX_WORDS in the chunk_strings function or adjusting it based on the document length to handle longer texts more effectively. This change would make the function more flexible and capable of dealing with a variety of document sizes without risking data loss or inefficiency. [important]
relevant line	MAX_WORDS=20

---

relevant file	semantic_router/encoders/bedrock.py
suggestion	Implement more comprehensive exception handling around the AWS SDK calls in the __call__ method to catch specific exceptions like network issues or API limits. This would improve the robustness of the encoder and provide clearer error messages for troubleshooting. [important]
relevant line	response = self.client.invoke_model(

---

relevant file	semantic_router/encoders/bedrock.py
suggestion	Add a validation step to check the input_type in the __init__ method of BedrockEncoder. This ensures that only supported input types are used, which can prevent runtime errors and make the code safer and more predictable. [medium]
relevant line	self.input_type = input_type

---

relevant file	semantic_router/encoders/bedrock.py
suggestion	Consider using a more secure method to handle AWS credentials, such as AWS IAM roles for EC2 or other AWS services, which can avoid hardcoding or passing credentials directly and reduce the risk of credential leakage. [important]
relevant line	access_key_id: Optional[str] = None

---

[github-actions]

PR Code Suggestions ✨

Category	Suggestion	Score
Possible bug	Add error handling for JSON parsing in the __call__ method The call method should handle potential JSON parsing errors when calling json.loads to prevent unhandled exceptions if the response is not in the expected format. semantic_router/encoders/bedrock.py [206-207] -response_body = json.loads(response.get("body").read()) +try: + response_body = json.loads(response.get("body").read()) +except json.JSONDecodeError: + raise ValueError("Failed to parse JSON response from Bedrock service.") Suggestion importance[1-10]: 8 Why: This suggestion addresses a potential bug by adding error handling for JSON parsing, which is crucial for preventing runtime errors due to unexpected response formats. This improvement significantly enhances the robustness of the method.	8
Best practice	Use more specific exception handling for boto3 errors The exception handling in the _initialize_client method could be more specific by catching boto3 related exceptions instead of a general Exception. This would make the error handling more precise and informative. semantic_router/encoders/bedrock.py [137-140] -except Exception as err: +except boto3.exceptions.BotoCoreError as err: raise ValueError( - f"The Bedrock client failed to initialize. Error: {err}" + f"The Bedrock client failed to initialize due to a BotoCoreError. Error: {err}" ) from err Suggestion importance[1-10]: 7 Why: This suggestion improves error handling by specifying the type of exceptions related to boto3, which enhances the clarity and maintainability of the error handling process.	7
Enhancement	Allow customization of the MAX_WORDS parameter in the chunk_strings method The method chunk_strings uses a hardcoded MAX_WORDS value which limits flexibility. Consider making MAX_WORDS a parameter of the BedrockEncoder class or method to allow customization based on different use cases. semantic_router/encoders/bedrock.py [163] -def chunk_strings(strings, MAX_WORDS=20): +def chunk_strings(strings, max_words): ... Suggestion importance[1-10]: 6 Why: Making MAX_WORDS customizable is a valid enhancement that increases the flexibility of the chunk_strings method. This change allows the method to be more adaptable to various text sizes and use cases.	6
Maintainability	Reorder imports alphabetically to improve code readability It's recommended to maintain imports in alphabetical order for better readability and maintainability. The import of BedrockEncoder should be placed in the correct order. semantic_router/encoders/init.py [4] from semantic_router.encoders.bedrock import BedrockEncoder +from semantic_router.encoders.bm25 import BM25Encoder +from semantic_router.encoders.clip import CLIPEncoder +from semantic_router.encoders.cohere import CohereEncoder Suggestion importance[1-10]: 5 Why: The suggestion correctly identifies a minor maintainability improvement by reordering imports alphabetically. However, the existing and improved code snippets are identical, indicating a mistake in the suggestion's presentation.	5

[codecov]

Codecov Report

Attention: Patch coverage is 83.33333% with 15 lines in your changes are missing coverage. Please review.

Project coverage is 80.98%. Comparing base (8cf0234) to head (73cf6b8).

❗ Current head 73cf6b8 differs from pull request most recent head a4b13a6. Consider uploading reports for the commit a4b13a6 to get more accurate results

Files	Patch %	Lines
semantic_router/encoders/bedrock.py	84.70%	13 Missing ⚠️
semantic_router/encoders/__init__.py	33.33%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #277      +/-   ##
==========================================
+ Coverage   80.90%   80.98%   +0.08%     
==========================================
  Files          44       45       +1     
  Lines        2582     2672      +90     
==========================================
+ Hits         2089     2164      +75     
- Misses        493      508      +15     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[jamescalam]

lgtm! Just rerunning tests