History to Cypher generation prompt #23
Conversation
WalkthroughThe pull request introduces several changes across multiple files. The Changes
Possibly related PRs
Suggested reviewers
Poem
Thank you for using CodeRabbit. We offer it for free to the OSS community and would appreciate your support in helping us grow. If you find it useful, would you consider giving us a shout-out on your favorite social media? 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments. CodeRabbit Commands (Invoked using PR comments)
Other keywords and placeholders
CodeRabbit Configuration File (
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Outside diff range and nitpick comments (5)
.gitignore (1)
.gitignore changes do not align with PR objectives
The modifications to the
.gitignorefile are unrelated to enhancing the Cypher generation prompt with a history of answers. These changes appear to address general development environment settings and data directories, not the specific goals of this PR.🔗 Analysis chain
Line range hint
1-10: Verify alignment with PR objectivesThe change to the .gitignore file, while beneficial, doesn't seem to align with the stated PR objectives of enhancing the Cypher generation prompt with a history of answers. Could you please clarify if there are additional changes related to the PR objectives that are not reflected in this file?
To check for any missing files or changes related to the PR objectives, you can run the following script:
Further verification required to align .gitignore changes with PR objectives
The initial script partially identified references to
cypher_generation. To ensure all relevant files and content related to Cypher generation prompts and history are captured, please run the following updated script:
Further investigation needed to ensure all PR-related changes are captured
To verify if there are additional modifications related to enhancing the Cypher generation prompt with a history of answers, please run the following script to list all files changed in this PR:
🏁 Scripts executed
The following scripts were executed for the analysis:
Script:
#!/bin/bash # Description: Check for files related to Cypher generation prompt or history of answers # Test: Search for files with names or content related to Cypher or prompt history echo "Files with names related to Cypher or prompt history:" fd -t f -e py -e js -e ts cypher prompt history echo "\nFiles containing content related to Cypher generation or prompt history:" rg -i "cypher generation|prompt history" --type py --type js --type tsLength of output: 664
Script:
#!/bin/bash # Description: Enhanced search for files related to Cypher generation prompt or history of answers # Search for files with names containing 'cypher', 'prompt', or 'history' echo "Files with names related to cypher, prompt, or history:" fd -t f -e py -e js -e ts "cypher|prompt|history" echo "\nFiles containing content related to Cypher generation or prompt history:" rg -i "cypher generation|prompt history" --type py --type js --type tsLength of output: 592
Script:
#!/bin/bash # Description: List all files changed in the current PR # Fetch the list of changed files echo "Files changed in this PR:" git diff --name-only origin/main...HEADLength of output: 275
pyproject.toml (2)
Line range hint
34-38: Dev dependency updates look good, but there's a 'jupyter' version inconsistency.The pinned versions for 'sphinx', 'sphinx-rtd-theme', and 'pandas' improve reproducibility. The addition of 'ipykernel' is consistent with Jupyter-related changes.
However, 'jupyter' appears in both main (^1.1.1) and dev (^1.0.0) dependencies with different versions. This inconsistency should be resolved.
Consider using the same version of 'jupyter' in both places, preferably the newer one (^1.1.1), or remove it from one location if it's not needed in both.
Line range hint
1-52: Overall, the changes align with the PR objective, but there are some dependency management concerns.The version bump and dependency updates generally support the enhancement of the Cypher generation prompt. However, there are a few points to address:
- The addition of 'jupyter' to main dependencies should be reconsidered.
- There's an inconsistency in the 'jupyter' versions between main and dev dependencies.
- The pinning of dev dependency versions is a positive change for reproducibility.
Please review these points and make necessary adjustments to ensure consistent and appropriate dependency management.
graphrag_sdk/steps/graph_query_step.py (1)
Line range hint
43-87: Consider enhancing the historical answer feature.The implementation successfully incorporates the
last_answerinto the Cypher generation process. However, to make this feature more robust and easier to use, consider the following suggestions:
Add a method to update
last_answerafter a successful query execution. This would allow the class to maintain its own history without relying on external management.Consider implementing a small queue or list of past answers instead of just the last one. This could provide more context for complex multi-turn interactions.
Here's a potential implementation for updating
last_answer:def update_last_answer(self, new_answer: str): self.last_answer = new_answer # In the run method, after successful execution: if context is not None: self.update_last_answer(context)This change would make the historical answer feature more self-contained and easier to manage.
graphrag_sdk/chat_session.py (1)
Line range hint
55-85: Consider future enhancements for robustnessThe changes successfully implement the basic functionality for incorporating answer history. For future enhancements, consider:
- Adding error handling for cases where
last_answermight be unexpectedlyNoneor in an invalid format.- Implementing a mechanism to limit the size of
last_answerto prevent potential memory issues in long chat sessions.- Adding documentation for the
last_answerattribute and its usage in the class docstring.- Considering thread safety if this class might be used in a multi-threaded environment.
Would you like assistance in implementing any of these enhancements?
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
⛔ Files ignored due to path filters (1)
poetry.lockis excluded by!**/*.lock
📒 Files selected for processing (6)
- .gitignore (1 hunks)
- examples/movies/demo-movies.ipynb (6 hunks)
- graphrag_sdk/chat_session.py (3 hunks)
- graphrag_sdk/fixtures/prompts.py (1 hunks)
- graphrag_sdk/steps/graph_query_step.py (3 hunks)
- pyproject.toml (2 hunks)
🔇 Additional comments (15)
.gitignore (1)
6-6: LGTM: Addition of.vscodeto .gitignoreAdding
.vscodeto the .gitignore file is a good practice. It prevents IDE-specific settings from being shared across the team, which can lead to conflicts and unnecessary changes in version control.pyproject.toml (2)
Line range hint
3-3: Version bump looks good.The increment from 0.2.0 to 0.2.1 is appropriate for a minor update or bug fix. This change follows semantic versioning principles.
28-28: Consider moving 'jupyter' to dev dependencies.The addition of 'jupyter' as a main dependency suggests it's required for core functionality. However, Jupyter is typically used as a development tool. Unless it's essential for runtime operations, consider moving it to the
[tool.poetry.group.dev.dependencies]section to keep the main dependencies lean.To check if Jupyter is used in the main code, run:
✅ Verification successful
'jupyter' is not used in the main codebase and can be moved to dev dependencies.
🏁 Scripts executed
The following scripts were executed for the analysis:
Script:
#!/bin/bash # Search for Jupyter imports in Python files rg --type python 'import\s+jupyter|from\s+jupyter\s+import'Length of output: 92
Script:
#!/bin/bash # Search for Jupyter imports in Python files by specifying .py extension rg -g '*.py' 'import\s+jupyter|from\s+jupyter\s+import'Length of output: 57
graphrag_sdk/steps/graph_query_step.py (3)
10-10: LGTM: New import added for history-aware prompt.The addition of
CYPHER_GEN_PROMPT_WITH_HISTORYis consistent with the PR objective of incorporating answer history into the Cypher generation process.
35-35: LGTM: Constructor updated to include optionallast_answerparameter.The addition of the
last_answerparameter with a default value ofNoneand its assignment to an instance variable is well-implemented. This change:
- Aligns with the PR objective of incorporating answer history.
- Maintains backward compatibility by making the parameter optional.
- Allows the
last_answerto be used in other methods of the class.Also applies to: 41-41
Line range hint
1-87: Summary of review for GraphQueryGenerationStep changesThe changes successfully implement the ability to incorporate historical answers into the Cypher query generation process, aligning well with the PR objectives. Key points:
- The new
last_answerparameter is correctly added and used.- The logic for choosing between different prompts based on the presence of
last_answeris implemented.- The core functionality of the class remains intact.
However, there are a few areas for improvement:
- Fix the indentation issue in the
runmethod.- Consider refactoring the prompt selection logic for better readability.
- Implement a mechanism to update
last_answerafter successful query execution.- Potentially expand the feature to maintain a small history of answers for more context.
Overall, the changes are a good start, but addressing these points would make the implementation more robust and maintainable.
graphrag_sdk/chat_session.py (4)
55-55: LGTM: Addition oflast_answerattributeThe introduction of the
last_answerattribute aligns well with the PR objective of incorporating answer history. Initializing it toNoneis appropriate for a new chat session.
84-84: LGTM: Updatinglast_answerattributeUpdating the
last_answerattribute with the newly generated answer is crucial for maintaining the answer history. This change is well-placed and completes the implementation of the feature.
85-85: LGTM: Improved readabilityMoving the
return answerstatement to a new line is a minor stylistic improvement that enhances code readability.
71-71: VerifyGraphQueryGenerationStepcompatibilityThe addition of
last_answeras a parameter toGraphQueryGenerationStepis in line with the PR objective. However, we should ensure that theGraphQueryGenerationStepclass has been updated to handle this new parameter.Please run the following script to verify the
GraphQueryGenerationStepclass implementation:examples/movies/demo-movies.ipynb (5)
32-32: Execution count updateThis change in execution count is a result of re-running the notebook and doesn't affect the functionality.
62-62: Execution count resetThis reset of the execution count to 1 is a result of re-running the notebook from the beginning and doesn't affect the functionality.
87-87: Execution count resetsThese resets of execution counts to 1 are consistent with re-running the notebook from the beginning and don't affect the functionality.
Also applies to: 110-110
140-143: Chat output and questions updated to incorporate answer historyThe changes in the chat output and questions align well with the PR objective of enhancing the Cypher generation prompt by incorporating a history of answers. The new output demonstrates:
- The ability to answer questions about movie directors (e.g., for "The Matrix" and "Side By Side").
- The capability to establish connections between directors and actors (e.g., Wachowskis and Keanu Reeves).
- The ability to reference previous answers (e.g., asking about the oldest director mentioned earlier).
These updates effectively showcase the improved functionality of referencing previous responses when generating Cypher queries.
Also applies to: 151-153
159-161: Development environment updatesThe changes in the kernel specification reflect updates to the development environment:
- The environment is now explicitly managed by Poetry.
- Python version has been upgraded from 3.11.9 to 3.12.3.
While these changes don't directly relate to the PR objective, they may impact the development and execution environment.
Please verify that all dependencies and code are compatible with Python 3.12.3. Run the following script to check for any potential compatibility issues:
Also applies to: 173-173
✅ Verification successful
Python Version Compatibility Verified
All dependencies and code are compatible with Python 3.12.3.
- Python version specified in
pyproject.toml: "^3.9.0" includes 3.12.3.- No CI/CD configurations require updates.
- No version-specific imports or syntax issues found.
🏁 Scripts executed
The following scripts were executed for the analysis:
Script:
#!/bin/bash # Description: Check for Python version compatibility issues # Test: Look for any version-specific imports or syntax rg --type python '(import.*python3\.\d+|f"{.*?=.*?}"|\*\*kwargs|yield from|async def|await)' # Test: Check if pyproject.toml specifies Python version rg --type toml 'python = ".*"' pyproject.toml # Test: Look for any CI/CD configs that might need updating fd '\.github|\.gitlab-ci\.yml|\.circleci' -x rg 'python'Length of output: 398
Script:
#!/bin/bash # Description: Check for Python version compatibility issues # Test: Look for any version-specific imports or syntax rg --type python '(import.*python3\.\d+|f".*?=.*?"|\*\*kwargs|yield from|async def|await)' # Test: Check if pyproject.toml specifies Python version rg --type toml 'python\s*=\s*".*"' pyproject.toml # Test: Look for any CI/CD configs that might need updating fd '\.github|\.gitlab-ci\.yml|\.circleci' --exec rg 'python'Length of output: 254
Script:
#!/bin/bash # Description: Check for Python version compatibility issues # Test: Look for any version-specific imports or syntax rg --type py '(import.*python3\.\d+|f".*?=.*?"|\*\*kwargs|yield from|async def|await)' # Test: Check if pyproject.toml specifies Python version rg --type toml 'python\s*=\s*".*"' pyproject.toml # Test: Look for any CI/CD configs that might need updating fd '\.github|\.gitlab-ci\.yml|\.circleci' --exec rg 'python'Length of output: 2849
| (CYPHER_GEN_PROMPT.format(question=question) | ||
| if self.last_answer is None | ||
| else CYPHER_GEN_PROMPT_WITH_HISTORY.format(question=question, last_answer=self.last_answer)) |
There was a problem hiding this comment.
Fix indentation and consider improving readability.
The conditional logic for choosing the appropriate prompt based on self.last_answer is correct. However, there are two issues to address:
- The indentation of the conditional statement is incorrect, which will lead to a syntax error.
- The inline formatting of the prompts might reduce readability, especially for more complex prompts.
To fix the indentation and improve readability, consider refactoring as follows:
if error is False:
if self.last_answer is None:
cypher_prompt = CYPHER_GEN_PROMPT.format(question=question)
else:
cypher_prompt = CYPHER_GEN_PROMPT_WITH_HISTORY.format(
question=question,
last_answer=self.last_answer
)
else:
cypher_prompt = CYPHER_GEN_PROMPT_WITH_ERROR.format(
question=question,
error=error
)This structure separates the logic more clearly and improves readability.
There was a problem hiding this comment.
Actionable comments posted: 3
🧹 Outside diff range and nitpick comments (5)
graphrag_sdk/orchestrator/steps/agent.py (1)
119-120: Improved chat session handling aligns with PR objectives.The changes to
agent.runand the subsequent handling of its return value align well with the PR objective of incorporating a history of answers. The explicit management of the chat session should allow for better continuity in conversations.However, consider adding error handling to account for potential changes in the
agent.runreturn value:try: response, chat_session = agent.run(self.properties.payload) runner.set_session(self.properties.session_id, chat_session) except ValueError as e: logger.error(f"Unexpected return value from agent.run: {e}") return AgentStepResult(AgentResponseCode.AGENT_ERROR, {"error": str(e)})This will make the code more robust against future changes or errors in the
agent.runmethod.graphrag_sdk/agents/kg_agent.py (3)
46-46: LGTM! Consider adding a docstring for the new attribute.The addition of
self.chat_sessionaligns well with the PR objective of incorporating a history of answers. This change allows the agent to maintain a persistent chat session across multiple interactions.Consider adding a docstring for the
chat_sessionattribute to improve code documentation. For example:self.chat_session = self._kg.chat_session() """The chat session associated with this agent's knowledge graph."""
127-127: LGTM! Update the docstring to reflect the new method signature.The removal of the
sessionparameter and the updated return type improve the method's consistency with the new chat session handling approach.Please update the method's docstring to accurately reflect the new signature and return value. For example:
def run(self, params: dict) -> tuple[str, GenerativeModelChatSession]: """ Ask the agent a question. Args: params (dict): The parameters for the agent, including the 'prompt' key. Returns: tuple[str, GenerativeModelChatSession]: A tuple containing the agent's response and the updated chat session. """
138-139: LGTM! Consider adding error handling for the 'prompt' key.The updated implementation correctly uses the instance's
chat_sessionand returns both the output and the updated chat session. This aligns well with the PR objective of incorporating a history of answers.Consider adding error handling for the case where the 'prompt' key is missing from the
paramsdictionary. For example:if "prompt" not in params: raise ValueError("The 'prompt' key is required in the params dictionary.") output = self.chat_session.send_message(params["prompt"]) return (output, self.chat_session.qa_chat_session)This will provide a more informative error message if the method is called without the required 'prompt' parameter.
graphrag_sdk/kg.py (1)
158-160: LGTM! Consider adding a comment for clarity.The changes look good and align with the PR objective of incorporating answer history. The method now correctly returns a tuple containing both the answer (or error message) and the
qa_chat_session.Consider adding a brief comment explaining why we're returning the
qa_chat_sessioneven when no answer is found. This could help future developers understand the purpose of this change. For example:# Return the qa_chat_session even when no answer is found to maintain conversation history return ("I am sorry, I could not find the answer to your question", qa_chat_session)
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
📒 Files selected for processing (7)
- examples/movies/demo-movies.ipynb (6 hunks)
- examples/trip/demo_orchestrator_trip.ipynb (7 hunks)
- graphrag_sdk/agents/agent.py (1 hunks)
- graphrag_sdk/agents/kg_agent.py (2 hunks)
- graphrag_sdk/fixtures/prompts.py (3 hunks)
- graphrag_sdk/kg.py (1 hunks)
- graphrag_sdk/orchestrator/steps/agent.py (1 hunks)
🚧 Files skipped from review as they are similar to previous changes (1)
- examples/movies/demo-movies.ipynb
🧰 Additional context used
🔇 Additional comments (10)
graphrag_sdk/orchestrator/steps/agent.py (1)
Line range hint
93-123: Verify method signature change.The AI-generated summary mentions a change in the
runmethod signature, removing thesessionparameter. However, this change is not visible in the provided code. Let's verify if this change has been implemented correctly.graphrag_sdk/agents/kg_agent.py (1)
Line range hint
1-153: Overall, the changes look good and align well with the PR objectives.The modifications to the
KGAgentclass successfully incorporate a history of answers by introducing a persistent chat session. The code structure and logic are sound, with only minor suggestions for improvements in documentation and error handling.Key points:
- The addition of
self.chat_sessionin the constructor enables persistent chat sessions.- The
runmethod has been simplified and now correctly utilizes the chat session.- The return value of the
runmethod now includes both the output and the updated chat session.These changes should effectively enhance the Cypher generation prompt by allowing reference to previous responses.
graphrag_sdk/agents/agent.py (1)
147-147: Approve the signature change with a suggestion for verification.The updated
runmethod signature improves encapsulation by managing theGenerativeModelChatSessioninternally. This change enhances the method's interface and provides more clarity on its return value.To ensure this change doesn't introduce any issues, please verify the following:
- All subclasses of
Agenthave been updated to implement the newrunmethod signature.- All calls to the
runmethod have been updated to handle the returned tuple instead of using a separatesessionparameter.Run the following script to identify potential areas that need updating:
Please review the output of this script and make any necessary updates to ensure consistency with the new
runmethod signature.examples/trip/demo_orchestrator_trip.ipynb (6)
66-66: LGTM: Execution count updateThe execution count for this cell has been updated from 2 to 3. This change doesn't affect the functionality of the code and is likely due to re-running the notebook.
251-251: LGTM: Execution count and file path updatesThe execution count for this cell has been updated from 3 to 4. Additionally, the file paths for loading JSON data have been updated to reflect a new directory structure:
-with open("data/cities.json") as f: +with open("examples/trip/data/cities.json") as f: -with open("data/restaurants.json") as f: +with open("examples/trip/data/restaurants.json") as f: -with open("data/attractions.json") as f: +with open("examples/trip/data/attractions.json") as f:These changes improve the organization of the project and use relative paths from the repository root, which is a good practice for maintainability and portability.
Also applies to: 274-278
363-363: LGTM: Execution count updateThe execution count for this cell has been updated from 4 to 5. This change doesn't affect the functionality of the code and is likely due to re-running the notebook.
393-393: LGTM: Execution count updateThe execution count for this cell has been updated from 5 to 6. This change doesn't affect the functionality of the code and is likely due to re-running the notebook.
412-500: LGTM: Updated output and new interactive cellsThe execution count for the existing cell has been updated from 6 to 7, and its output has been refreshed. Additionally, three new cells (execution counts 8, 9, and 10) have been added, demonstrating further interactions with the orchestrator:
- A query for the restaurant mentioned in the itinerary for the second day's dinner.
- A request to change the restaurant and update the itinerary.
- A query for the restaurant mentioned in the itinerary for the first day's dinner.
These new cells effectively showcase the orchestrator's ability to handle follow-up questions and maintain context from previous queries, which aligns well with the PR objective of enhancing the Cypher generation prompt by incorporating a history of answers. This addition improves the demonstration of the system's capabilities and provides a more interactive example for users.
Line range hint
506-520: LGTM: Updated Python version and kernel settingsThe notebook metadata has been updated with the following changes:
- Python version updated from 3.11.4 to 3.11.10.
- Kernel display name changed from "Python 3 (ipykernel)" to "Poetry Env2".
- Kernel name changed from "python3" to "my-poetry-env".
These updates indicate a switch to a Poetry-managed environment, which is a good practice for dependency management in Python projects. The minor Python version update shouldn't cause any compatibility issues.
graphrag_sdk/fixtures/prompts.py (1)
464-476: Approve: Addition of historical context to Cypher generationThe introduction of
CYPHER_GEN_PROMPT_WITH_HISTORYis a valuable enhancement. By considering the last answer, it has the potential to improve the relevance and continuity of generated Cypher queries across multiple interactions.
| if not cypher or len(cypher) == 0: | ||
| return ("I am sorry, I could not find the answer to your question", qa_chat_session) | ||
|
|
There was a problem hiding this comment.
💡 Codebase verification
Impact on Callers of ask Method
Several callers of the ask method are assigning its return value to a single variable, which may lead to issues since ask now returns a tuple (str, GenerativeModelChatSession).
tests/test_multi_agent.py:315: Assigns toresponsetests/test_kg_openai.py:86: Assigns toanswertests/test_kg_ollama.py:91: Assigns toanswertests/test_kg_gemini.py:92: Assigns toanswer
These assignments should be updated to handle the tuple return type appropriately.
🔗 Analysis chain
Verify the impact of the return type change.
The ask method now returns a tuple (str, GenerativeModelChatSession) instead of just a string. This change might affect other parts of the system that call this method.
Please run the following script to check for potential issues:
Please review the results to ensure that all callers of the ask method are updated to handle the new tuple return type correctly.
🏁 Scripts executed
The following scripts were executed for the analysis:
Script:
#!/bin/bash
# Description: Find all occurrences of the 'ask' method being called and verify they handle the new return type correctly.
# Search for calls to the 'ask' method
echo "Searching for calls to the 'ask' method:"
rg --type python -n '\.ask\(' --glob '!graphrag_sdk/kg.py'
# Search for assignments of the 'ask' method result
echo "\nSearching for assignments of the 'ask' method result:"
rg --type python -n '=.*\.ask\(' --glob '!graphrag_sdk/kg.py'
# Note: Manual review of the results is required to ensure proper handling of the new return type.
Length of output: 410
Script:
#!/bin/bash
# Description: Find all occurrences of the 'ask' method being called and verify they handle the new return type correctly.
# Search for calls to the 'ask' method
echo "Searching for calls to the 'ask' method:"
rg '\.ask\(' --glob '*.py' -n
# Search for assignments of the 'ask' method result
echo "\nSearching for assignments of the 'ask' method result:"
rg '=.*\.ask\(' --glob '*.py' -n
# Note: Manual review of the results is required to ensure proper handling of the new return type.
Length of output: 1204
| First, determine if the last answer provided to the user is relevant to the current question. If it is relevant, you may consider incorporating information from it into the query. If it is not relevant, ignore it and generate the query based solely on the question. | ||
|
|
There was a problem hiding this comment.
🛠️ Refactor suggestion
Suggestion: Provide guidance for determining relevance
The prompt instructs to determine if the last answer is relevant but doesn't provide specific criteria for this determination. This could lead to inconsistent results across different executions. Consider adding guidelines or heuristics for assessing relevance, such as keyword matching or semantic similarity thresholds.
💡 Codebase verification
Sanitization Missing for last_answer
The last_answer variable is incorporated into prompts without any sanitization or validation. This exposes the application to potential prompt injection attacks. It's crucial to implement proper sanitization or validation for last_answer before including it in prompts to ensure security.
🔗 Analysis chain
Potential security concern: Sanitize last_answer to prevent prompt injection
While incorporating the last answer can enhance query relevance, it also introduces a potential security risk. If last_answer contains untrusted user input, it could lead to prompt injection attacks. Consider implementing a sanitization or validation step for last_answer before including it in the prompt.
To verify the potential security risk, let's check if there's any sanitization of user input in the codebase:
🏁 Scripts executed
The following scripts were executed for the analysis:
Script:
#!/bin/bash
# Search for input sanitization or validation related to last_answer
rg -i '(sanitize|validate|escape).*last_answer' --type python
Length of output: 94
Script:
#!/bin/bash
# Search for input sanitization or validation related to last_answer
rg -i '(sanitize|validate|escape).*last_answer' --type py
Length of output: 59
Script:
#!/bin/bash
# Search for all occurrences of last_answer in Python files to identify how it's processed
rg 'last_answer' --type py -A 5
Length of output: 2481
Add history of answers to Cypher generation prompt
Summary by CodeRabbit
New Features
.gitignoreto exclude the.vscodedirectory.Bug Fixes
Chores
pyproject.tomlto ensure compatibility and performance improvements.