@@ -11,8 +11,7 @@ as an [_edit format_](https://aider.chat/docs/benchmarks.html#edit-formats).
1111
1212## Table of Contents
1313- [ What is CEDARScript?] ( #what-is-cedarscript )
14- - [ Installation] ( #how-to-use-it )
15- - [ Usage] ( #usage )
14+ - [ Installation] ( #installation )
1615- [ Why Use CEDARScript?] ( #why-use-cedarscript )
1716- [ Performance Comparison] ( #performance-comparison )
1817- [ Individual Test Analysis] ( #individual-test-analysis )
@@ -27,17 +26,23 @@ is a domain-specific language designed to improve how AI coding assistants inter
2726It provides a standardized way to express complex code modification and analysis operations, making it easier for
2827AI-assisted development tools to understand and execute these tasks.
2928
30- ## How to use it
29+ ## Installation
3130
32- 1 . [ Install Aider] ( https://aider.chat/docs/install.html ) , if you haven't.
31+ 1 . Install _ ** Aider with CEDARScript** _ via this command below:
32+ ``` shell
33+ python -m ensurepip --upgrade
34+ pip install --upgrade --force-reinstall \
35+ git+https://github.com/elifarley/aider@cedarscript \
36+ aider-chat
37+ ```
33382 . Now, simply use the [ ` --edit-format ` switch] ( https://aider.chat/docs/more/edit-formats.html ) and select ` cedarscript ` :
3439``` shell
35- aider --model gemini/gemini-1.5-flash-latest -- edit-format cedarscript
40+ aider --edit-format cedarscript
3641```
3742
3843## Why use CEDARScript?
3944
40- ` TL;DR ` : You can get higher success rates when compared to other edit formats.
45+ ` TL;DR ` : You can get higher success rates when refactoring large files, comparing to other edit formats.
4146
42471 . ** Higher Success Rates** : Significantly improves the performance of AI models in code refactoring tasks.
43482 . ** Cost-Effective Performance** : Enables more affordable models to compete with top-tier options.
@@ -63,17 +68,27 @@ CEDARScript has shown remarkable improvements in AI model performance for code r
6368This suggests that ** CEDARScript** can level the playing field, enabling more accessible AI models
6469to compete with and even _ exceed_ the capabilities of more expensive options in certain coding tasks.
6570
71+ ### The Gemini 1.5 Flash benchmark highlights
6672
67- ### Individual Test Analysis
73+ - 48% of tests (43 total) showed improvements
74+ - 103% increase in Pass 1 success rate (75 tests)
75+ - Test duration reduced by 93% (from 5:17:26 to 0:25:17)
76+ - Token efficiency greatly improved:
77+ - Sent tokens: -37% (7.59M)
78+ - Received tokens: -96% (180K)
79+ - Error reduction:
80+ - Error outputs: -94% (35 total)
81+ - Malformed outputs: -94% (6 cases)
82+ - Syntax errors: -85% (3 cases)
83+ - Indent errors eliminated (100% reduction)
6884
69- ** Overall Performance**
85+ <details >
86+ <summary >Delta...</summary >
87+
88+ ![ image] ( https://github.com/user-attachments/assets/86683a1b-2b64-49c9-89ff-eb18d3511ae7 )
89+ </details >
7090
71- | | Count | % |
72- | -------------| -------| -------|
73- | Improved | 39 | 43.8% |
74- | Worsened | 7 | 7.9% |
75- | Stable | 43 | 48.3% |
76- | Total tests | 89 | 100% |
91+ ### Individual Test Analysis
7792
7893<details >
7994<summary >Individual Test Diff</summary >
0 commit comments