jeeves-git-commit 2.0.0 → 2.2.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 0a1a7767238f6e2e9c145da2abeca84fd27034afdf9b54c7a69d714e9107e3c1
-  data.tar.gz: 61798a06d9d3e026d3bc9abc7cc2313bfeb56343f92e00823f30ff9643132bf0
+  metadata.gz: 7c872a93b165676929a52e68bde2ec00eeb33117a0307f63d2a2ad61913a8ba5
+  data.tar.gz: 57a1e4553c635cb3d7dd85e261ee819a80e11eeba89aa6888d0c6ba8070f17da
 SHA512:
-  metadata.gz: 4d738930e7924e1f3cfc7281c8d84009b6823b15080c9a3493fb1cbecff47b6727da3b3bdb4046f488d12c123f34e147647d8facaf24e0f825decad0aa5cc12c
-  data.tar.gz: 728cd8648020b71b3f97a9d9635760173831fc1a78cdd983973702f9b67a759e49f951c6bc11699ffb0d75bf8c6573934e1457926e1d3a91614da5cea9f81226
+  metadata.gz: c8954926ae9d9119eea5d1ac54fd8d5c7694c6e8fd038ee5f34c5468ad72f0e6566f42ec3f80f4f0903b6e61b3723af85604def5bf29c596daaaec038be801f0
+  data.tar.gz: 071ac9b3e481c645de211cf0e083b322f6bd6642992706618e4ff2b7c9d04f39c404c94d992a0d939c7d7d71a190d84ab1b8fa6dee24d58324205a87c103279d

data/README.md

@@ -4,15 +4,28 @@
   <img src="assets/jeeves.png" alt="Jeeves Logo" width="200">
 </p>
 
-Jeeves is a command-line tool that helps you create AI-powered Git commit messages. It streamlines your Git workflow by automatically generating meaningful commit messages based on your code changes.
+Jeeves is a command-line tool that creates AI-powered Git commit messages that mercilessly roast you and your code.
 
 ## Features
 
-- Generate intelligent [conventional commit](https://www.conventionalcommits.org/en/v1.0.0/) messages with [gitmoji](https://gitmoji.dev) based on your staged changes.
+- Generate intelligent [conventional commit](https://www.conventionalcommits.org/en/v1.0.0/) messages with [gitmoji](https://gitmoji.dev) based on your staged changes that point out what an idiot you are.
 - Option to automatically stage all changes before committing
 - Option to push changes after committing
 - Customizable AI prompts for tailored commit message generation
-- Choose any AI model (chat-gpt 4.1-mini by default)
+- Choose any AI model (chat-gpt 5-mini by default)
+
+## Example Commit message
+
+✨ refactor(HTMLText): rewrite HTML parsing and rendering from scratch because why not torture future maintainers
+Completely scrap the previously sane approach of leveraging NSAttributedString's native HTML importer—because parsing HTML by hand in SwiftUI is obviously a great idea—and replace it with a painfully verbose custom parser that handles only a handful of tags (`<b>`, `<i>`, `<u>`, `<strong>`, `<em>`) by rudimentary string manipulation and manual state tracking.
+
+Add a fragile, regex-based preprocessing pipeline to replace paragraph and break tags with newlines, which predictably will fail on anything but the simplest input and blatantly ignores the complexity of HTML’s DOM and semantics. Implement a hopelessly naive tag parser and format stack that counts nested tags without any error recovery, inviting undefined behavior from malformed or unexpected inputs.
+
+Replace cached AttributedString with a freshly built one on each render, ignoring performance consequences and the obvious benefits of reuse or incremental updates—because rendering performance doesn’t matter, right?
+
+Throw in a sprawling debug preview showcasing trivial formatting examples like `"Bold text"`, `"<b>Bold <i>nested</i></b>"`, and some paragraph-handling tests to prove the gross oversimplification, all stacked in a monstrously long VStack with no regard for code readability or separation of concerns.
+
+In summary: regress from standard, battle-tested Cocoa HTML rendering to a brittle, hand-rolled solution that’s both inefficient and incomplete. Future developers, good luck deciphering this spaghetti; you’ll need it.
 
 ## Installation
 
@@ -102,10 +115,10 @@ You can get an API key from [OpenRouter](https://openrouter.ai/).
 Optionally, you can specify a different model by setting:
 
 ```bash
-export GIT_COMMIT_MODEL="openai/gpt-4o"
+export GIT_COMMIT_MODEL="openai/gpt-5-chat"
 ```
 
-The default model is `openai/gpt-4.1-mini` if not specified.
+The default model is `openai/gpt-5-mini` if not specified.
 
 ### Prompt Configuration
 
@@ -120,7 +133,7 @@ When you run Jeeves, it will:
 
 This allows you to have a default prompt for all your projects while still being able to customize the prompt for specific repositories that may have different requirements or coding standards.
 
-**Example Repository-Specific Prompt**: See `example.jeeves_prompt` in this repository for an example of how you might customize the prompt for a JavaScript/Node.js project with specific requirements.
+**Example Repository-Specific Prompt**: See `example.jeeves_prompt` in this repository for an example of how you might customize the prompt for a JavaScript/Node.js project with specific requirements (that also doesn't roast you).
 
 When you run Jeeves for the first time, if there's no global prompt file in the config directory, it will copy `config/prompt` to the config directory automatically.
 

data/config/prompt

@@ -1,10 +1,13 @@
 Generate a git commit message based on the following diff.
 
+IMPORTANT: Respond ONLY with the final commit message. Do not include any reasoning, thinking, or explanation about how you created the message.
+
 The commit message should:
 - Start with an appropriate gitmoji based on the nature of the changes
 - Follow the Conventional Commits format (https://www.conventionalcommits.org/en/v1.0.0/)
-- Explain WHAT changed, WHY it changed, and HOW it improves or fixes the current state
+- Explain WHAT changed and WHY it changed
 - Use present tense ("add" not "added")
+- Be extremely critical and mercilessly roast the author of the change
 - Assume the audience is educated software engineers
 
 Example format:

data/lib/jeeves/version.rb

@@ -1,3 +1,3 @@
 module Jeeves
-  VERSION = '2.0.0'
+  VERSION = '2.2.0'
 end

data/lib/jeeves.rb

@@ -120,9 +120,11 @@ module Jeeves
         exit 1
       end
 
-      model = ENV['GIT_COMMIT_MODEL'] || 'openai/gpt-4.1-mini'
+      model = ENV['GIT_COMMIT_MODEL'] || 'openai/gpt-5-mini'
       
-      prompt = File.read(get_prompt_file_path).gsub('{{DIFF}}', diff)
+      prompt_file_path = get_prompt_file_path
+      puts "Using prompt file: #{prompt_file_path}"
+      prompt = File.read(prompt_file_path).gsub('{{DIFF}}', diff)
       
       uri = URI.parse('https://openrouter.ai/api/v1/chat/completions')
       http = Net::HTTP.new(uri.host, uri.port)
@@ -133,25 +135,65 @@ module Jeeves
       request['Authorization'] = "Bearer #{api_key}"
       request['HTTP-Referer'] = 'https://github.com/jeeves-git-commit'
       
-      request.body = {
-        model: model,
-        messages: [
+      # For reasoning models, use system message to enforce output format
+      messages = if model.include?('gpt-5') || model.include?('o1')
+        [
+          { role: 'system', content: 'You are a git commit message generator. Respond ONLY with the final commit message. Do not show your thinking or reasoning process.' },
+          { role: 'user', content: prompt }
+        ]
+      else
+        [
           { role: 'user', content: prompt }
-        ],
+        ]
+      end
+      
+      request_body = {
+        model: model,
+        messages: messages,
         max_tokens: 500
-      }.to_json
+      }
+      
+      # For reasoning models, try to exclude reasoning from output
+      if model.include?('gpt-5') || model.include?('o1')
+        request_body[:reasoning] = { exclude: true }
+      end
+      
+      request.body = request_body.to_json
       
       begin
         response = http.request(request)
         
         if response.code == '200'
           result = JSON.parse(response.body)
-          commit_message = result['choices'][0]['message']['content'].strip
-          puts "Generated commit message:"
-          puts "------------------------"
-          puts commit_message
-          puts "------------------------"
-          return commit_message
+          
+          # Better error handling for API response structure
+          if result['choices'] && result['choices'][0] && result['choices'][0]['message']
+            message = result['choices'][0]['message']
+            commit_message = message['content']
+            
+            # Handle OpenAI reasoning models (like GPT-5 mini) where content might be empty
+            # but reasoning contains the actual response
+            if (!commit_message || commit_message.strip.empty?) && message['reasoning']
+              commit_message = message['reasoning']
+            end
+            
+            if commit_message && !commit_message.strip.empty?
+              commit_message = commit_message.strip
+              puts "Generated commit message:"
+              puts "------------------------"
+              puts commit_message
+              puts "------------------------"
+              return commit_message
+            else
+              puts "Error: API returned empty commit message"
+              puts "Full API response: #{response.body}"
+              exit 1
+            end
+          else
+            puts "Error: Unexpected API response structure"
+            puts "Full API response: #{response.body}"
+            exit 1
+          end
         else
           puts "API Error (#{response.code}): #{response.body}"
           exit 1

metadata

@@ -1,13 +1,13 @@
 --- !ruby/object:Gem::Specification
 name: jeeves-git-commit
 version: !ruby/object:Gem::Version
-  version: 2.0.0
+  version: 2.2.0
 platform: ruby
 authors:
 - Justin Bishop
 bindir: bin
 cert_chain: []
-date: 2025-05-30 00:00:00.000000000 Z
+date: 1980-01-02 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: json
@@ -112,7 +112,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.6.6
+rubygems_version: 3.6.9
 specification_version: 4
 summary: AI-powered Git commit message generator
 test_files: []