Open Issues Need Help
View All on GitHub Improve Agent Robustness: Enhance Prompt Engineering to Reduce Invalid JSON and Hallucinations from Gemma Model about 2 months ago
AI Summary: Improve the prompt engineering for a React Native expense tracking app's AI agent (powered by Google's Gemma 27B model) to reduce invalid JSON outputs and hallucinations. This involves refining the system prompt in `services/aiService.js` by strengthening instructions, refining few-shot examples, and potentially simplifying tool descriptions to ensure the model consistently generates valid JSON and uses tools correctly.
Complexity:
4/5
bug good first issue prompt-engineering