gptdiff 0.1.24__tar.gz → 0.1.27__tar.gz

{gptdiff-0.1.24 → gptdiff-0.1.27}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: gptdiff
-Version: 0.1.24
+Version: 0.1.27
 Summary: A tool to generate and apply git diffs using LLMs
 Author: 255labs
 Classifier: License :: OSI Approved :: MIT License
@@ -82,13 +82,15 @@ done
 
 *Requires reasoning model*
 
-### Why GPTDiff?
+## Why Choose GPTDiff?
 
-- **Understands Your Code** - Describe changes in plain English
-- **Safe Modifications** - Keeps existing code working
-- **Auto-Fix** - `--apply` fixes mistakes in generated changes
-- **Works Instantly** - No complex setup needed
-- **Whole Project View** - Handles multiple files together
+- **Describe changes in plain English**  
+- **AI gets your whole project**  
+- **Auto-fixes conflicts**  
+- **Keeps code functional**  
+- **Fast setup, no fuss**  
+- **You approve every change**  
+- **Costs are upfront**
 
 ## Core Capabilities
 

{gptdiff-0.1.24 → gptdiff-0.1.27}/README.md

@@ -55,13 +55,15 @@ done
 
 *Requires reasoning model*
 
-### Why GPTDiff?
-
-- **Understands Your Code** - Describe changes in plain English
-- **Safe Modifications** - Keeps existing code working
-- **Auto-Fix** - `--apply` fixes mistakes in generated changes
-- **Works Instantly** - No complex setup needed
-- **Whole Project View** - Handles multiple files together
+## Why Choose GPTDiff?
+
+- **Describe changes in plain English**  
+- **AI gets your whole project**  
+- **Auto-fixes conflicts**  
+- **Keeps code functional**  
+- **Fast setup, no fuss**  
+- **You approve every change**  
+- **Costs are upfront**
 
 ## Core Capabilities
 

{gptdiff-0.1.24 → gptdiff-0.1.27}/gptdiff/applydiff.py

@@ -189,6 +189,7 @@ def parse_diff_per_file(diff_text):
 
     Note:
         Uses 'b/' prefix detection from git diffs to determine target paths
+        This doesn't work all the time and needs to be revised with stronger models
     """
     header_re = re.compile(r'^(?:diff --git\s+)?(a/[^ ]+)\s+(b/[^ ]+)\s*$', re.MULTILINE)
     lines = diff_text.splitlines()

{gptdiff-0.1.24 → gptdiff-0.1.27}/gptdiff/gptdiff.py

@@ -1,5 +1,6 @@
 #!/usr/bin/env python3
 from pathlib import Path
+from urllib.parse import urlparse
 import subprocess
 import hashlib
 import re
@@ -14,6 +15,7 @@ import pkgutil
 import contextvars
 from pkgutil import get_data
 import threading
+from threading import Lock
 
 import openai
 from openai import OpenAI
@@ -203,13 +205,26 @@ def load_prepend_file(file):
     with open(file, 'r') as f:
         return f.read()
 
-# Function to call GPT-4 API and calculate the cost
+def domain_for_url(base_url):
+    parsed = urlparse(base_url)
+    if parsed.netloc:
+        if parsed.username:
+            domain = parsed.hostname
+            if parsed.port:
+                domain += f":{parsed.port}"
+        else:
+            domain = parsed.netloc
+    else:
+        domain = base_url
+    return domain
+
 def call_llm_for_diff(system_prompt, user_prompt, files_content, model, temperature=0.7, max_tokens=30000, api_key=None, base_url=None):
     enc = tiktoken.get_encoding("o200k_base")
     
     # Use colors in print statements
     red = "\033[91m"
     green = "\033[92m"
+    blue = "\033[94m"
     reset = "\033[0m"
     start_time = time.time()
 
@@ -222,10 +237,13 @@ def call_llm_for_diff(system_prompt, user_prompt, files_content, model, temperat
     if 'gemini' in model:
         user_prompt = system_prompt + "\n" + user_prompt
 
+    input_content = system_prompt + "\n" + user_prompt + "\n" + files_content
+    token_count = len(enc.encode(input_content))
     messages = [
-        {"role": "system", "content": f"{green}{system_prompt}{reset}"},
+        {"role": "system", "content": system_prompt},
         {"role": "user", "content": user_prompt + "\n" + files_content},
     ]
+
     if VERBOSE:
         print(f"{green}Using {model}{reset}")
         print(f"{green}SYSTEM PROMPT{reset}")
@@ -233,7 +251,7 @@ def call_llm_for_diff(system_prompt, user_prompt, files_content, model, temperat
         print(f"{green}USER PROMPT{reset}")
         print(user_prompt, "+", len(enc.encode(files_content)), "tokens of file content")
     else:
-        print("Generating diff...")
+        print(f"Generating diff using model '{green}{model}{reset}' from '{blue}{domain_for_url(base_url)}{reset}' with {token_count} input tokens...")
 
     if not api_key:
         api_key = os.getenv('GPTDIFF_LLM_API_KEY')
@@ -248,9 +266,9 @@ def call_llm_for_diff(system_prompt, user_prompt, files_content, model, temperat
         temperature=temperature)
 
     if VERBOSE:
-        print("RESPONSE RAW-------------")
+        print("Debug: Raw LLM Response\n---")
         print(response.choices[0].message.content.strip())
-        print("/RESPONSE RAW-------------")
+        print("---")
     else:
         print("Diff generated.")
 
@@ -265,9 +283,6 @@ def call_llm_for_diff(system_prompt, user_prompt, files_content, model, temperat
     print("-" * 40)
 
     # Now, these rates are updated to per million tokens
-    cost_per_million_prompt_tokens = 30
-    cost_per_million_completion_tokens = 60
-    cost = (prompt_tokens / 1_000_000 * cost_per_million_prompt_tokens) + (completion_tokens / 1_000_000 * cost_per_million_completion_tokens)
 
     full_response = response.choices[0].message.content.strip()
     full_response, reasoning = swallow_reasoning(full_response)
@@ -279,7 +294,7 @@ def call_llm_for_diff(system_prompt, user_prompt, files_content, model, temperat
         toolbox.use(event)
     diff_response = diff_context.get()
 
-    return full_response, diff_response, prompt_tokens, completion_tokens, total_tokens, cost
+    return full_response, diff_response, prompt_tokens, completion_tokens, total_tokens
 
 # New API functions
 def build_environment(files_dict):
@@ -538,14 +553,23 @@ def smart_apply_patch(project_dir, diff_text, user_prompt, args):
     Attempt to apply a diff via smartapply: process each file concurrently using the LLM.
     """
     from pathlib import Path
+    start_time = time.time()
     parsed_diffs = parse_diff_per_file(diff_text)
     print("Found", len(parsed_diffs), "files in diff, processing smart apply concurrently:")
+    green = "\033[92m"
+    red = "\033[91m"
+    blue = "\033[94m"
+    reset = "\033[0m"
+
     if len(parsed_diffs) == 0:
         print(colorize_warning_warning("There were no entries in this diff. The LLM may have returned something invalid."))
         if args.beep:
             print("\a")
         return
     threads = []
+    success_files = []
+    failed_files = []
+    success_lock = Lock()
 
     def process_file(file_path, file_diff):
         full_path = Path(project_dir) / file_path
@@ -559,11 +583,14 @@ def smart_apply_patch(project_dir, diff_text, user_prompt, args):
                 print(colorize_warning_warning(f"File {file_path} not found - skipping deletion"))
             return
 
-        try:
-            original_content = full_path.read_text()
-        except (UnicodeDecodeError, IOError):
-            print(f"Skipping file {file_path} due to read error")
-            return
+        original_content = ""
+        if full_path.exists():
+            try:
+                original_content = full_path.read_text()
+            except (UnicodeDecodeError, IOError) as e:
+                print(f"Cannot read {file_path} due to {str(e)}, treating as new file")
+        else:
+            print(f"File {file_path} does not exist, treating as new file")
 
         # Use SMARTAPPLY-specific environment variables if set, otherwise fallback.
         smart_apply_model = os.getenv("GPTDIFF_SMARTAPPLY_MODEL")
@@ -586,9 +613,7 @@ def smart_apply_patch(project_dir, diff_text, user_prompt, args):
         else:
             base_url = os.getenv("GPTDIFF_LLM_BASE_URL", "https://nano-gpt.com/api/v1/")
 
-        print("-" * 40)
-        print("Running smartapply with", model, "on", file_path)
-        print("-" * 40)
+        print(f"Running smartapply in parallel using model '{green}{model}{reset}' from '{blue}{domain_for_url(base_url)}{reset}'...")
         try:
             updated_content = call_llm_for_apply_with_think_tool_available(
                 file_path, original_content, file_diff, model,
@@ -597,12 +622,18 @@ def smart_apply_patch(project_dir, diff_text, user_prompt, args):
                 max_tokens=args.max_tokens)
             if updated_content.strip() == "":
                 print("Cowardly refusing to write empty file to", file_path, "merge failed")
+                with success_lock:
+                    failed_files.append(file_path)
                 return
             full_path.parent.mkdir(parents=True, exist_ok=True)
             full_path.write_text(updated_content)
             print(f"\033[1;32mSuccessful 'smartapply' update {file_path}.\033[0m")
+            with success_lock:
+                success_files.append(file_path)
         except Exception as e:
             print(f"\033[1;31mFailed to process {file_path}: {str(e)}\033[0m")
+            with success_lock:
+                failed_files.append(file_path)
 
     for file_path, file_diff in parsed_diffs:
         thread = threading.Thread(target=process_file, args=(file_path, file_diff))
@@ -610,7 +641,17 @@ def smart_apply_patch(project_dir, diff_text, user_prompt, args):
         threads.append(thread)
     for thread in threads:
         thread.join()
-
+    elapsed = time.time() - start_time
+    minutes, seconds = divmod(int(elapsed), 60)
+    time_str = f"{minutes}m {seconds}s" if minutes else f"{seconds}s"
+    print(f"Smartapply successfully applied changes in {time_str}. Check the updated files to confirm.")
+    if failed_files:
+        print(f"\033[1;31mSmart apply completed in {time_str} with failures for {len(failed_files)} files:\033[0m")
+        for file in failed_files:
+            print(f"  - {file}")
+        print("Please check the errors above for details.")
+    else:
+        print(f"\033[1;32mSmart apply completed successfully in {time_str} for all {len(success_files)} files.\033[0m")
     if args.beep:
         print("\a")
 
@@ -710,9 +751,8 @@ def main():
         with open('prompt.txt', 'w') as f:
             f.write(full_prompt)
         print(f"Total tokens: {token_count:5d}")
-        print(f"\033[1;32mNot calling GPT-4.\033[0m")  # Green color for success message
+        print(f"\033[1;32mWrote full prompt to prompt.txt.\033[0m")
         print('Instead, wrote full prompt to prompt.txt. Use `xclip -selection clipboard < prompt.txt` then paste into chatgpt')
-        print(f"Total cost: ${0.0:.4f}")
         exit(0)
     else:
         # Validate API key presence before any API operations
@@ -729,7 +769,7 @@ def main():
                 print("Request canceled")
                 sys.exit(0)
         try:
-            full_text, diff_text, prompt_tokens, completion_tokens, total_tokens, cost = call_llm_for_diff(system_prompt, user_prompt, files_content, args.model, 
+            full_text, diff_text, prompt_tokens, completion_tokens, total_tokens = call_llm_for_diff(system_prompt, user_prompt, files_content, args.model, 
                                                                                                     temperature=args.temperature,
                                                                                                     api_key=os.getenv('GPTDIFF_LLM_API_KEY'),
                                                                                                     base_url=os.getenv('GPTDIFF_LLM_BASE_URL', "https://nano-gpt.com/api/v1/"),
@@ -741,35 +781,40 @@ def main():
             prompt_tokens = 0
             completion_tokens = 0
             total_tokens = 0
-            cost = 0
             print(f"Error in LLM response {e}")
 
     if(diff_text.strip() == ""):
-        print(f"\033[1;33mThere was no data in this diff. The LLM may have returned something invalid.\033[0m")
-        print("Unable to parse diff text. Full response:", full_text)
+        print(f"\033[1;33mWarning: No valid diff data was generated. This could be due to an unclear prompt or an invalid LLM response.\033[0m")
+        print("Suggested action: Refine your prompt or check the full response below for clues.")
+        print("Full LLM response:\n---\n" + full_text + "\n---")
         if args.beep:
-            print("\a")  # Terminal bell for completion notification
+            print("\a")
         return
 
     elif args.apply:
         print("\nAttempting apply with the following diff:")
-        print("\n<diff>")
         print(color_code_diff(diff_text))
-        print("\n</diff>")
-        print("Saved to patch.diff")
-        if apply_diff(project_dir, diff_text):
-            print(f"\033[1;32mPatch applied successfully with 'git apply'.\033[0m")
+        print("\033[94m**Attempting to apply patch using basic method...**\033[0m")
+        apply_result = apply_diff(project_dir, diff_text)
+        if apply_result:
+            print(f"\033[1;32mPatch applied successfully with basic apply.\033[0m")
         else:
-            print("Apply failed, attempting smart apply.")
+            print("\033[94m**Attempting smart apply with LLM...**\033[0m")
             smart_apply_patch(project_dir, diff_text, user_prompt, args)
 
     if args.beep:
-        print("\a")  # Terminal bell for completion notification
+        print("\a")
 
-    print(f"Prompt tokens: {prompt_tokens}")
-    print(f"Completion tokens: {completion_tokens}")
-    print(f"Total tokens: {total_tokens}")
-    #print(f"Total cost: ${cost:.4f}")
+    green = "\033[92m"
+    reset = "\033[0m"
+    if VERBOSE:
+        print("API Usage Details:")
+        print(f"- Prompt tokens: {prompt_tokens}")
+        print(f"- Completion tokens: {completion_tokens}")
+        print(f"- Total tokens: {total_tokens}")
+        print(f"- Model used: {green}{args.model}{reset}")
+    else:
+        print(f"API Usage: {total_tokens} tokens, Model used: {green}{args.model}{reset}")
 
 def swallow_reasoning(full_response: str) -> (str, str):
     """

{gptdiff-0.1.24 → gptdiff-0.1.27}/gptdiff/gptpatch.py

@@ -13,7 +13,7 @@ This tool uses the same patch-application logic as gptdiff.
 import sys
 import argparse
 from pathlib import Path
-from gptdiff.gptdiff import apply_diff, smart_apply_patch
+from gptdiff.gptdiff import apply_diff, smart_apply_patch, color_code_diff
 
 def parse_arguments():
     parser = argparse.ArgumentParser(
@@ -49,11 +49,14 @@ def parse_arguments():
         default=30000,
         help="Maximum tokens to use for LLM responses"
     )
+    parser.add_argument('--verbose', action='store_true', help='Enable verbose output with detailed information')
     parser.add_argument('--dumb', action='store_true', default=False, help='Attempt dumb apply before trying smart apply')
     return parser.parse_args()
 
 def main():
     args = parse_arguments()
+    import gptdiff.gptdiff as gd
+    gd.VERBOSE = args.verbose
     if args.diff:
         diff_text = args.diff
     else:
@@ -64,15 +67,21 @@ def main():
         diff_text = diff_path.read_text(encoding="utf8")
 
     project_dir = args.project_dir
+
+    if args.verbose:
+        print("\n\033[1;34mDiff to be applied:\033[0m")
+        print(color_code_diff(diff_text))
+        print("")
+
     if args.dumb:
         success = apply_diff(project_dir, diff_text)
         if success:
-            print("✅ Diff applied successfully.")
+            print("\033[1;32m✅ Diff applied successfully.\033[0m")
         else:
-            print("❌ Failed to apply diff using git apply. Attempting smart apply.")
+            print("\033[1;31m❌ Failed to apply diff using git apply. Attempting smart apply.\033[0m")
             smart_apply_patch(project_dir, diff_text, "", args)
     else:
         smart_apply_patch(project_dir, diff_text, "", args)
         
 if __name__ == "__main__":
-    main()
+    main()

{gptdiff-0.1.24 → gptdiff-0.1.27}/gptdiff.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: gptdiff
-Version: 0.1.24
+Version: 0.1.27
 Summary: A tool to generate and apply git diffs using LLMs
 Author: 255labs
 Classifier: License :: OSI Approved :: MIT License
@@ -82,13 +82,15 @@ done
 
 *Requires reasoning model*
 
-### Why GPTDiff?
+## Why Choose GPTDiff?
 
-- **Understands Your Code** - Describe changes in plain English
-- **Safe Modifications** - Keeps existing code working
-- **Auto-Fix** - `--apply` fixes mistakes in generated changes
-- **Works Instantly** - No complex setup needed
-- **Whole Project View** - Handles multiple files together
+- **Describe changes in plain English**  
+- **AI gets your whole project**  
+- **Auto-fixes conflicts**  
+- **Keeps code functional**  
+- **Fast setup, no fuss**  
+- **You approve every change**  
+- **Costs are upfront**
 
 ## Core Capabilities
 

{gptdiff-0.1.24 → gptdiff-0.1.27}/setup.py

@@ -2,7 +2,7 @@ from setuptools import setup, find_packages
 
 setup(
     name='gptdiff',
-    version='0.1.24',
+    version='0.1.27',
     description='A tool to generate and apply git diffs using LLMs',
     author='255labs',
     packages=find_packages(),  # Use find_packages() to automatically discover packages

{gptdiff-0.1.24 → gptdiff-0.1.27}/tests/test_applydiff.py

@@ -141,3 +141,31 @@ def test_new_file_creation_minimal_header_failure(tmp_project_dir_empty):
     content = new_file.read_text()
     assert content.strip() == expected_content.strip(), f"Expected file content:\n{expected_content}\nGot:\n{content}"
 
+@pytest.fixture
+def tmp_project_dir_with_gptdiff(tmp_path):
+    """
+    Create a temporary project directory with a gptdiff.py file containing four lines.
+    """
+    project_dir = tmp_path / "project"
+    project_dir.mkdir()
+    file = project_dir / "gptdiff.py"
+    file.write_text("#!/usr/bin/env python3\nfrom pathlib import Path\n# Line 3\n# Line 4\n")
+    return project_dir
+
+def test_apply_bad_diff_fails(tmp_project_dir_with_gptdiff):
+    """
+    Test that a diff is applied correctly to the file.
+    """
+    diff_text = """diff --git a/gptdiff/gptdiff.py b/gptdiff/gptdiff.py
+index 1234567..89abcde 100644
+--- a/gptdiff/gptdiff.py
++++ b/gptdiff/gptdiff.py
+@@ -1,4 +1,5 @@
+ #!/usr/bin/env python3
++from threading import Lock
+ from pathlib import Path"""
+
+    # Assume apply_diff is a function that applies the diff
+    from gptdiff.gptdiff import apply_diff
+    result = apply_diff(str(tmp_project_dir_with_gptdiff), diff_text)
+    assert result is False, "apply_diff should fail, needs smartapply"

{gptdiff-0.1.24 → gptdiff-0.1.27}/tests/test_parse_diff_per_file.py

@@ -126,6 +126,20 @@ diff --git a/file2.py b/file2.py
         self.assertIn("file1.py", paths)
         self.assertIn("file2.py", paths)
 
+def test_parse_diff_per_file_unconventional_header():
+    diff_text = """--- game.js
++++ game.js
+@@ -0,0 +1,3 @@
++let player = {
++    class: "Warrior",
++};
+"""
+    result = parse_diff_per_file(diff_text)
+    assert len(result) == 1, f"Expected one file patch, got {len(result)}"
+    file_path, patch = result[0]
+    assert file_path == "game.js", f"Expected file path 'game.js', got '{file_path}'"
+    assert "+++ game.js" in patch, "Expected patch to include '+++ game.js'"
+    assert "+let player" in patch, "Expected patch to include added lines"
 
 if __name__ == '__main__':
     unittest.main()

{gptdiff-0.1.24 → gptdiff-0.1.27}/tests/test_smartapply.py

@@ -244,3 +244,23 @@ def test_smartapply_complex_single_hunk(monkeypatch):
     assert "if not data:" in updated
     assert "temp = data * 2" not in updated
     assert "for x in data:" in updated
+
+def test_smartapply_new_file_with_incorrect_header(monkeypatch):
+    """Test that smartapply creates a new file from a diff with an incorrect '--- a/' header."""
+    diff_text = """
+diff --git a/game.js b/game.js
+--- a/game.js
+++++ b/game.js
+@@ -0,0 +1,3 @@
++let player = {
++    class: "Warrior",
++};
+"""
+    original_files = {}
+    expected_content = "let player = {\n    class: \"Warrior\",\n};"
+    def mock_call_llm(file_path, original_content, file_diff, model, api_key, base_url, extra_prompt=None, max_tokens=None):
+        return expected_content
+    monkeypatch.setattr('gptdiff.gptdiff.call_llm_for_apply', mock_call_llm)
+    updated_files = smartapply(diff_text, original_files)
+    assert "game.js" in updated_files, "The new file 'game.js' should be created"
+    assert updated_files["game.js"] == expected_content, "The file content should match the diff"