feat: add diff-based code modifications (#63)

* feat: add diff-based code modifications

* chore: update lockfile

* chore: fix type

* chore: add a shortcut for the diff command

* test: add more test coverage

* fix: improve diff-parser

* feat: add context flag

* test: improve test coverage

* docs: document new flag

* fix: should probably actually load the correct prompt template

* docs: note that diff mode is experimental

API.md

@@ -107,9 +107,11 @@ async function performAITask() {
         await runAIAssistedTask({
             path: '/path/to/project',
             task: 'Implement user authentication',
-            description: 'Add user login and registration functionality using JWT',
+            description: 'Add user login and registration functionality using JWT', // detailed task description
+            instructions: 'Use bcrypt for password hashing...', // additional instructions, pass an empty string to skip
+            context: ['src/api/auth.js', 'src/utils/auth.js'], // specify files or directories to include in the task context
             model: 'claude-3-5-sonnet-20240620',
-            dryRun: true,
+            dryRun: true, // Set to false to actually apply changes
         });
         console.log('AI-assisted task completed successfully!');
     } catch (error) {

README.md

@@ -111,6 +111,7 @@ While CodeWhisper excels at performing individual coding tasks and even large fe
 
 * 🧠 AI-powered task planning and code generation
 * 🔄 Full git integration for version control of AI-generated changes
+* 🔄 Support for diff-based code modifications to handle larger edits within output token limits
 * 🌍 Support for various models and LLM providers, such as Anthropic, OpenAI, and Groq
 * 🔐 Support for local models via Ollama
 * 🚀 Blazingly fast code processing with concurrent workers
@@ -160,6 +161,8 @@ codewhisper list-models
 # CodeWhisper will prompt you to select a model from the list of available models
 codewhisper task
 
+# To use the new experimental diff-based code modifications, you can use the following command:
+codewhisper task --diff
 
 # You can also specify a model directly
 # Claude-3.5 Sonnet
@@ -192,6 +195,7 @@ While CodeWhisper supports a variety of providers and models, my current recomme
 3. **GPT-4o-mini (OpenAI)**
    - Strong performance
    - Generates good quality plans and results
+   - Has a long max output length of 16384 tokens which can be used for complex tasks
 
 #### Experimental Support
 

USAGE.md

@@ -35,11 +35,18 @@ codewhisper task [options]
 Options:
 * `-p, --path <path>`: Path to the codebase (default: current directory)
 * `-m, --model <modelId>`: Specify the AI model to use (if not specified, CodeWhisper will prompt you to select a model from the list of available models)
+* `-t, --task <task>`: Short task title
+* `-d, --description <description>`: Detailed task description
+* `-i, --instructions <instructions>`: Additional instructions for the task
+* `-c, --context <paths...>`: Specify files or directories to include in the task context. Can be file paths, directory paths, or glob patterns. Multiple entries should be space-separated.
+* `-df, --diff`: Use diff format for file updates instead of full file content
 * `-g, --gitignore <path>`: Path to .gitignore file (default: .gitignore)
 * `-f, --filter <patterns...>`: File patterns to include (use glob patterns, e.g., "src/**/*.js")
 * `-e, --exclude <patterns...>`: File patterns to exclude (use glob patterns, e.g., "**/*.test.js")
 * `-s, --suppress-comments`: Strip comments from the code
 * `-l, --line-numbers`: Add line numbers to code blocks
+* `-cw, --context-window <number>`: Specify the context window for the AI model. Only applicable for Ollama models.
+* `-mt, --max-tokens <number>`: Specify the max output tokens for the AI model. Only applicable for Ollama models.
 * `--case-sensitive`: Use case-sensitive pattern matching
 * `--custom-ignores <patterns...>`: Additional patterns to ignore
 * `--cache-path <path>`: Custom path for the cache file
@@ -50,9 +57,6 @@ Options:
 * `--log-ai-interactions`: Enable logging of AI prompts, responses, and parsing results to a file (default: false)
 * `-max, --max-cost-threshold <number>`: Set a maximum cost threshold for AI operations in USD (e.g., 0.5 for $0.50)
 * `--auto-commit`: Automatically commit changes
-* `-t, --task <task>`: Short task title
-* `-d, --description <description>`: Detailed task description
-* `-i, --instructions <instructions>`: Additional instructions for the task
 
 #### Model Selection
 
@@ -319,10 +323,18 @@ codewhisper generate --model ollama:llama3.1:70b --context-window 131072 --max-t
 22. Run an AI-assisted task with detailed logging:
 
 ```bash
-codewhisper task -m claude-3-5-sonnet-20240620 --log-ai-interactions -t "Implement error handling" -d "Add comprehensive error handling to all API endpoints"
+# This will generate a log file (codewhisper-`<date>`.log) in the current directory, containing all AI prompts, responses, and parsing results.
+
+codewhisper task -m claude-3-5-sonnet-20240620 --log-ai-interactions -t "Implement error handling" -d "Add comprehensive error handling to all API endpoints" -i "some instructions"
 ```
 
-This will generate a log file (codewhisper-`<date>`.log) in the current directory, containing all AI prompts, responses, and parsing results.
+23. Run an AI-assisted task using diff-based updates:
+
+```bash
+# This command will use diff-based updates for modified files, which can help handle larger edits within token limits.
+
+codewhisper task --diff -m claude-3-5-sonnet-20240620 -t "Refactor authentication logic" -d "Update the user authentication system to use JWT tokens" -i "some instructions"
+```
 
 ## CI/CD Integration
 

package.json

@@ -97,9 +97,11 @@
     "@anthropic-ai/sdk": "0.25.0",
     "@inquirer/prompts": "5.3.6",
     "@octokit/rest": "21.0.1",
-    "ai": "3.2.42",
+    "@types/diff": "5.2.1",
+    "ai": "3.2.45",
     "chalk": "5.3.0",
     "commander": "12.1.0",
+    "diff": "5.2.0",
     "dotenv": "16.4.5",
     "fast-glob": "3.3.2",
     "fs-extra": "11.2.0",