From 2a26723102b881989de01130d9c2c061c3305f7c Mon Sep 17 00:00:00 2001 From: AI Christianson Date: Wed, 18 Dec 2024 16:01:10 -0500 Subject: [PATCH] improve prompts --- CHANGELOG.md | 1 + ra_aid/prompts.py | 4 ++++ 2 files changed, 5 insertions(+) diff --git a/CHANGELOG.md b/CHANGELOG.md index d1062f8..19a7e86 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -8,6 +8,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0 ## [Unreleased] - Added monorepo_detected, existing_project_detected, and ui_detected tools so the agent can take specific actions. +- Prompt improvements for real-world projects. ## [0.6.3] diff --git a/ra_aid/prompts.py b/ra_aid/prompts.py index 11bafc0..ae7b19d 100644 --- a/ra_aid/prompts.py +++ b/ra_aid/prompts.py @@ -103,6 +103,10 @@ You have often been criticized for: - For tasks requiring UI changes, not researching existing UI libraries and conventions. - Not emitting enough research subtasks on changes on large projects, e.g. to discover testing or UI conventions, etc. - Doing one-shot tasks, which is good, but not compiling or testing your work when appropriate. +- Not finding *examples* of how to do similar things in the current codebase and emitting them with emit_key_snippets. +- Not finding unit tests because they are in slightly different locations than expected. +- Not handling real-world projects that often have inconsistencies and require more thorough research and pragmatism. +- Not finding *ALL* related files and snippets. You'll often be on the right path and give up/start implementing too quickly. """ # Planning stage prompt - guides task breakdown and implementation planning