[MRG] Add RAG for the code analysis

[huangyz0918]

User description

Closes #127

---

PR Type

enhancement, dependencies

---

Description

• Implemented a new Memory class in mle/utils/memory.py for managing memory and external knowledge using ChromaDB.
• Integrated Memory initialization in the new function of mle/cli.py.
• Updated mle/utils/__init__.py to include the memory module.
• Added chromadb and onnxruntime to requirements.txt to support the new memory management functionality.

---

Changes walkthrough 📝

	Relevant files
Enhancement	cli.py Initialize and import `Memory` in CLI                                        mle/cli.py Imported Memory from mle.utils. Initialized Memory in the new function. +4/-0      __init__.py Include `memory` module in utils package                                  mle/utils/init.py Added import statement for memory. +1/-0      memory.py Implement `Memory` class with ChromaDB integration              mle/utils/memory.py Implemented Memory class for memory and external knowledge management. Integrated with ChromaDB for persistent storage. Provided methods for adding, querying, peeking, getting, deleting, counting, and resetting memory records. +156/-0
Dependencies	requirements.txt Add `chromadb` and `onnxruntime` to dependencies                  requirements.txt Added chromadb and onnxruntime as dependencies. +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 Reviewer Guide 🔍

⏱️ Estimated effort to review: 3 🔵🔵🔵⚪⚪

🧪 No relevant tests

🔒 No security concerns identified

⚡ Key issues to review Error Handling The Memory class methods do not include error handling for database operations which might raise exceptions during execution. Consider adding try-except blocks to handle potential exceptions gracefully. Security Concern The method reset in the Memory class allows resetting the memory, which can be a critical operation. It's mentioned to control this via an environment variable ALLOW_RESET, but this is not implemented in the code provided. Ensure to implement this check to prevent unauthorized resets.

[github-actions]

PR Code Suggestions ✨

Category	Suggestion	Score
Enhancement	Add safety checks before deleting a collection Modify the delete method to handle the case where the collection does not exist, preventing potential errors or exceptions when attempting to delete a non-existent collection. mle/utils/memory.py [139] -return self.client.delete_collection(name=collection_name) +if self.client.has_collection(name=collection_name): + return self.client.delete_collection(name=collection_name) +else: + logger.warning(f"Collection {collection_name} does not exist.") Suggestion importance[1-10]: 9 Why: The suggestion prevents errors by checking if a collection exists before attempting to delete it, which enhances the robustness of the code.	9
Possible issue	Add error handling during the initialization of the Memory class Ensure that the Memory class is properly initialized with error handling for missing or incorrect configuration data. This can prevent runtime errors if the configuration file or expected keys are missing. mle/cli.py [134] -Memory(project_dir) +try: + memory_instance = Memory(project_dir) +except KeyError as e: + logger.error(f"Configuration key missing: {e}") +except FileNotFoundError: + logger.error("Configuration file not found") Suggestion importance[1-10]: 8 Why: This suggestion adds error handling for potential issues during the initialization of the Memory class, which is crucial for preventing runtime errors due to missing configuration data.	8
Possible bug	Improve type safety in platform comparison Replace the direct comparison of data['platform'] with OpenAIModel with a more robust type checking or conversion to ensure that the comparison is valid and does not raise a TypeError. mle/utils/memory.py [36] -if data['platform'] == OpenAIModel: +if isinstance(data.get('platform'), OpenAIModel): Suggestion importance[1-10]: 7 Why: The suggestion improves type safety by using isinstance, which is a more robust way to check types and prevents potential TypeError during comparison.	7
Performance	Optimize UUID generation for efficiency Use a more efficient method for generating UUIDs in bulk by using list comprehension directly in the uuid.uuid4() call, which reduces the overhead of multiple function calls. mle/utils/memory.py [70] -ids = [str(uuid.uuid4()) for _ in range(len(queries))] +ids = [uuid.uuid4().hex for _ in queries] Suggestion importance[1-10]: 6 Why: This suggestion slightly optimizes the UUID generation process, but the performance gain is minor and not critical for functionality.	6

[huangyz0918]

Introducing the memory class for managing the vector storage, but how to fully leverage the sourcecode, how to draw the summary should put into a different PR