feat: Add Turkish language support (turkce-lite/full/ultra)#179
Open
AlperGuven wants to merge 11 commits intoJuliusBrussee:mainfrom
Open
feat: Add Turkish language support (turkce-lite/full/ultra)#179AlperGuven wants to merge 11 commits intoJuliusBrussee:mainfrom
AlperGuven wants to merge 11 commits intoJuliusBrussee:mainfrom
Conversation
- Added `turkce-lite`, `turkce-full`, and `turkce-ultra` intensity levels to `skills/caveman/SKILL.md` - Added `/caveman turkce`, `/caveman türkçe`, "mağara adam modu", "az token lütfen", and "kısa konuş" triggers to the skill description and `hooks/caveman-mode-tracker.js` - Added Turkish benchmark prompts in `evals/prompts/tr.txt` and updated `evals/llm_run.py` to read all `.txt` prompt files - Updated README.md to include the Turkish mode documentation, examples, and token saving demonstration (using agglutinative language optimizations)
Add .eval-home to .gitignore to prevent accidental commits of evaluation home directories.
Remove the separate Turkish prompts file (tr.txt) and merge its contents into the existing English prompts file (en.txt). This simplifies the evaluation prompt structure by maintaining all prompts in one location. The commit_msg.txt file, which contained documentation for a previous Turkish feature addition, is also removed as it is no longer needed.
Remove token counts from Turkish example to focus on compression ratio. Update benchmark table with new token counts and recalculated averages.
Load environment variables from .env.local file to allow custom configuration. Set a separate HOME directory for Claude to avoid interfering with user's session. Simplify prompt loading by reading only the English prompts file.
Update SKILL.md to include new language variants (turkce and wenyan) in the mode switch command. This expands the caveman skill's stylistic options for users.
Add new section explaining Turkish mode functionality, including three compression levels (lite, full, ultra) with examples and benchmark results showing average 66% token savings.
Update README and SKILL documentation to provide more specific guidance for each Turkish caveman mode. Clarify the linguistic transformations applied in each level (Lite, Full, Ultra) and update examples to better illustrate the output style.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This branch is Feat/turkish mode commits ahead of JuliusBrussee:main — Contribute / Open Pull Request. Intention is to add Turkish language support to this package. Due to the agglutinative nature of Turkish, we used different strategies than those used for the English word “caveman,” and we also included the benchmark results for Claude.