Skip to main content
Project Progress Reports
Pre-Demo 1 Report
- Use CodeSearchNet dataset to create JSON files for code search
- Embed code comments for similarity search with UniXCoder
- Allow user input to search for most similar natural language query using vector search
- Rank top 10 most similar code segments based on embedding comparison score
- Ask LLM to re-rank the top 10 and get a new "most similar" ranking
Post-Demo 1 / Pre-Demo 2 Report
- Develop frontend for user input and top 10 display
- Change off of JSON to SQLite to allow for faster embedding and search
- Work on benchmarking for Dr. Phan's research
- Use Ollama to host LLMs locally for unlimited LLM querying
Post-Demo 2 / Pre-Final Demo Report
- Finish benchmarking to give accuracy score comparison for out methodology vs. UniXCoder
- Merge Ollama changes to work with benchmarking
- Update frontend for better display and more information
- Optimize embeddings for fastest search capabilities
Meeting Minutes
- Week 1 - 1 hour
- Week 2 - 1 hour
- Week 3 - 45 minutes
- Week 4 - 45 minutes
- Week 5 - 1 hour
- Week 6 - 30 minutes
- Week 7 - 30 minutes
- Week 8 - 30 minutes
- Week 9 - 30 minutes
- Week 10 - no meeting
- Week 11 - 15 minutes
- Week 12 - 10 minutes
- Week 13 - No meeting
- Week 14 - No meeting
