pdd-cli 0.0.23__py3-none-any.whl → 0.0.24__py3-none-any.whl

pdd/__init__.py

@@ -1 +1 @@
-__version__ = "0.0.23"
+__version__ = "0.0.24"

pdd/bug_main.py

@@ -3,6 +3,7 @@ import sys
 from typing import Tuple, Optional
 import click
 from rich import print as rprint
+from pathlib import Path
 
 from .construct_paths import construct_paths
 from .bug_to_unit_test import bug_to_unit_test
@@ -74,9 +75,22 @@ def bug_main(
 
         # Save results if output path is provided
         if output_file_paths.get("output"):
+            output_path = output_file_paths["output"]
+            # Additional check to ensure the path is not empty
+            if not output_path or output_path.strip() == '':
+                # Use a default output path in the current directory
+                output_path = f"test_{Path(code_file).stem}_bug.{language.lower()}"
+                if not ctx.obj.get('quiet', False):
+                    rprint(f"[yellow]Warning: Empty output path detected. Using default: {output_path}[/yellow]")
+                output_file_paths["output"] = output_path
+            
             # Create directory if it doesn't exist
-            os.makedirs(os.path.dirname(output_file_paths["output"]), exist_ok=True)
-            with open(output_file_paths["output"], 'w') as f:
+            dir_path = os.path.dirname(output_path)
+            if dir_path:  # Only create directory if there's a directory part in the path
+                os.makedirs(dir_path, exist_ok=True)
+            
+            # Write the file
+            with open(output_path, 'w') as f:
                 f.write(unit_test)
 
         # Provide user feedback

pdd/data/llm_model.csv

@@ -1,17 +1,17 @@
 provider,model,input,output,coding_arena_elo,base_url,api_key,counter,encoder,max_tokens,max_completion_tokens,structured_output
-OpenAI,"gpt-4o-mini",0.15,0.60,1246,,OPENAI_API_KEY,tiktoken,o200k_base,,16384,True
-OpenAI,"grok-2-1212",2,10,1255,"https://api.x.ai/v1",XAI_API_KEY,tiktoken,o200k_base,4096,,False
+OpenAI,"gpt-4.1-nano",0.1,0.40,1246,,OPENAI_API_KEY,tiktoken,o200k_base,,32768,True
+OpenAI,"grok-3-beta",3,15,1255,"https://api.x.ai/v1",XAI_API_KEY,tiktoken,o200k_base,131072,,False
 Anthropic,"claude-3-5-haiku-20241022",1,5,1259,,ANTHROPIC_API_KEY,anthropic,claude-3-sonnet-20240229,8192,,False
 OpenAI,"deepseek-coder",0.14,0.28,1279,https://api.deepseek.com/beta,DEEPSEEK_API_KEY,autotokenizer,deepseek-coder-7b-instruct-v1.5,8192,,False
-Google,"gemini-2.0-flash-thinking-exp-01-21",.1,.4,1291,,GOOGLE_API_KEY,,,8192,,False
-GoogleVertexAI,"gemini-2.0-pro-exp-02-05",1.25,5,1299,,VERTEX_AI_API_KEY,,,8192,,False
+Google,"gemini-2.5-flash-preview-04-17",.15,3.5,1291,,GOOGLE_API_KEY,,,65535,,False
+GoogleVertexAI,"gemini-2.5-pro-exp-03-25",1.25,10,1299,,VERTEX_AI_API_KEY,,,65535,,False
 Anthropic,claude-3-7-sonnet-20250219,3,15,1312,,ANTHROPIC_API_KEY,anthropic,claude-3-sonnet-20240229,64000,,False
-Google,gemini-2.5-pro-exp-03-25,1.25,5,1313,,GOOGLE_API_KEY,,,8192,,False
+Google,gemini-2.5-pro-exp-03-25,1.25,10,1313,,GOOGLE_API_KEY,,,65535,,False
 OpenAI,"deepseek-r1-distill-llama-70b-specdec",5,5,1314,https://api.groq.com/openai/v1,GROQ_API_KEY,autotokenizer,deepseek-coder-7b-instruct-v1.5,16384,,False
 Ollama,"deepseek-r1:70b-llama-distill-q8_0",0.0,0.0,1315,,PWD,,,,,False
 Ollama,deepseek-r1:32b-qwen-distill-fp16,0.0,0.0,1316,,PWD,,,,,False
-OpenAI,"o3-mini",1.1,4.4,1319,,OPENAI_API_KEY,tiktoken,o200k_base,,100000,True
-OpenAI,"o1-2024-12-17",15,60,1331,,OPENAI_API_KEY,tiktoken,o200k_base,,32768,True
-OpenAI,"gpt-4o-2024-11-20",2.5,10,1332,,OPENAI_API_KEY,tiktoken,o200k_base,,16384,True
+OpenAI,"o4-mini",1.1,4.4,1319,,OPENAI_API_KEY,tiktoken,o200k_base,,100000,True
+OpenAI,"o3",10,40,1331,,OPENAI_API_KEY,tiktoken,o200k_base,,100000,True
+OpenAI,"gpt-4.1",2,8,1332,,OPENAI_API_KEY,tiktoken,o200k_base,,32768,True
 OpenAI,"deepseek-reasoner",0.55,2.19,1336,https://api.deepseek.com/beta,DEEPSEEK_API_KEY,autotokenizer,deepseek-coder-7b-instruct-v1.5,8192,,False
 Fireworks,accounts/fireworks/models/deepseek-r1,3,8,1338,,FIREWORKS_API_KEY,,,8192,,False

pdd/fix_verification_errors.py

@@ -0,0 +1,259 @@
+import re
+from typing import Dict, Any
+from rich import print as rprint
+from rich.markdown import Markdown
+from .load_prompt_template import load_prompt_template
+from .llm_invoke import llm_invoke
+
+def fix_verification_errors(
+    program: str,
+    prompt: str,
+    code: str,
+    output: str,
+    strength: float,
+    temperature: float = 0.0,
+    verbose: bool = False,
+) -> Dict[str, Any]:
+    """
+    Identifies and fixes issues in a code module based on verification output.
+
+    Args:
+        program: The program code that ran the code module.
+        prompt: The prompt used to generate the code module.
+        code: The code module to be fixed.
+        output: The output logs from the program run during verification.
+        strength: The strength (0-1) for the LLM model selection.
+        temperature: The temperature for the LLM model. Defaults to 0.
+        verbose: If True, prints detailed execution information. Defaults to False.
+
+    Returns:
+        A dictionary containing:
+        - 'explanation': A string with verification details and fix explanation
+                         in XML format, or None if no issues were found.
+        - 'fixed_program': The potentially fixed program code string.
+        - 'fixed_code': The potentially fixed code module string.
+        - 'total_cost': The total cost incurred from LLM calls.
+        - 'model_name': The name of the LLM model used.
+        - 'verification_issues_count': The number of issues found during verification.
+    """
+    total_cost = 0.0
+    model_name = None
+    verification_issues_count = 0
+    verification_details = None
+    fix_explanation = None
+    fixed_program = program
+    fixed_code = code
+    final_explanation = None
+
+    if not all([program, prompt, code, output]):
+        rprint("[bold red]Error:[/bold red] Missing one or more required inputs (program, prompt, code, output).")
+        return {
+            "explanation": None,
+            "fixed_program": program,
+            "fixed_code": code,
+            "total_cost": 0.0,
+            "model_name": None,
+            "verification_issues_count": 0,
+        }
+    if not (0.0 <= strength <= 1.0):
+        rprint(f"[bold red]Error:[/bold red] Strength must be between 0.0 and 1.0, got {strength}.")
+        return {
+            "explanation": None,
+            "fixed_program": program,
+            "fixed_code": code,
+            "total_cost": 0.0,
+            "model_name": None,
+            "verification_issues_count": 0,
+        }
+
+    if verbose:
+        rprint("[blue]Loading prompt templates...[/blue]")
+    try:
+        find_errors_prompt_template = load_prompt_template("find_verification_errors_LLM")
+        fix_errors_prompt_template = load_prompt_template("fix_verification_errors_LLM")
+        if not find_errors_prompt_template or not fix_errors_prompt_template:
+            raise ValueError("One or both prompt templates could not be loaded.")
+    except Exception as e:
+        rprint(f"[bold red]Error loading prompt templates:[/bold red] {e}")
+        return {
+            "explanation": None,
+            "fixed_program": program,
+            "fixed_code": code,
+            "total_cost": total_cost,
+            "model_name": model_name,
+            "verification_issues_count": verification_issues_count,
+        }
+    if verbose:
+        rprint("[green]Prompt templates loaded successfully.[/green]")
+
+    if verbose:
+        rprint(f"\n[blue]Step 2: Running verification check (Strength: {strength}, Temp: {temperature})...[/blue]")
+
+    verification_input_json = {
+        "program": program,
+        "prompt": prompt,
+        "code": code,
+        "output": output,
+    }
+
+    try:
+        verification_response = llm_invoke(
+            prompt=find_errors_prompt_template,
+            input_json=verification_input_json,
+            strength=strength,
+            temperature=temperature,
+            verbose=False,
+        )
+        total_cost += verification_response.get('cost', 0.0)
+        model_name = verification_response.get('model_name', model_name)
+        verification_result = verification_response.get('result', '')
+
+        if verbose:
+            rprint(f"[cyan]Verification LLM call complete.[/cyan]")
+            rprint(f"  [dim]Model Used:[/dim] {verification_response.get('model_name', 'N/A')}")
+            rprint(f"  [dim]Cost:[/dim] ${verification_response.get('cost', 0.0):.6f}")
+
+    except Exception as e:
+        rprint(f"[bold red]Error during verification LLM call:[/bold red] {e}")
+        return {
+            "explanation": None,
+            "fixed_program": program,
+            "fixed_code": code,
+            "total_cost": total_cost,
+            "model_name": model_name,
+            "verification_issues_count": verification_issues_count,
+        }
+
+    if verbose:
+        rprint("\n[blue]Verification Result:[/blue]")
+        rprint(Markdown(verification_result))
+
+    issues_found = False
+    try:
+        count_match = re.search(r"<issues_count>(\d+)</issues_count>", verification_result)
+        if count_match:
+            verification_issues_count = int(count_match.group(1))
+        else:
+            rprint("[yellow]Warning:[/yellow] Could not find <issues_count> tag in verification result. Assuming 0 issues.")
+            verification_issues_count = 0
+
+        if verification_issues_count > 0:
+            details_match = re.search(r"<details>(.*?)</details>", verification_result, re.DOTALL)
+            if details_match:
+                verification_details = details_match.group(1).strip()
+                if verification_details:
+                    issues_found = True
+                    if verbose:
+                        rprint(f"\n[yellow]Found {verification_issues_count} potential issues. Proceeding to fix step.[/yellow]")
+                else:
+                    rprint("[yellow]Warning:[/yellow] <issues_count> is > 0, but <details> tag is empty. Treating as no issues found.")
+                    verification_issues_count = 0
+            else:
+                rprint("[yellow]Warning:[/yellow] <issues_count> is > 0, but could not find <details> tag. Treating as no issues found.")
+                verification_issues_count = 0
+        else:
+            if verbose:
+                rprint("\n[green]No issues found during verification.[/green]")
+
+    except ValueError:
+        rprint("[bold red]Error:[/bold red] Could not parse integer value from <issues_count> tag.")
+        return {
+            "explanation": None,
+            "fixed_program": program,
+            "fixed_code": code,
+            "total_cost": total_cost,
+            "model_name": model_name,
+            "verification_issues_count": 0,
+        }
+    except Exception as e:
+        rprint(f"[bold red]Error parsing verification result:[/bold red] {e}")
+        return {
+            "explanation": None,
+            "fixed_program": program,
+            "fixed_code": code,
+            "total_cost": total_cost,
+            "model_name": model_name,
+            "verification_issues_count": 0,
+        }
+
+    if issues_found and verification_details:
+        if verbose:
+            rprint(f"\n[blue]Step 5: Running fix generation (Strength: {strength}, Temp: {temperature})...[/blue]")
+
+        fix_input_json = {
+            "program": program,
+            "prompt": prompt,
+            "code": code,
+            "output": output,
+            "issues": verification_details,
+        }
+
+        try:
+            fix_response = llm_invoke(
+                prompt=fix_errors_prompt_template,
+                input_json=fix_input_json,
+                strength=strength,
+                temperature=temperature,
+                verbose=False,
+            )
+            total_cost += fix_response.get('cost', 0.0)
+            model_name = fix_response.get('model_name', model_name)
+            fix_result = fix_response.get('result', '')
+
+            if verbose:
+                rprint(f"[cyan]Fix LLM call complete.[/cyan]")
+                rprint(f"  [dim]Model Used:[/dim] {fix_response.get('model_name', 'N/A')}")
+                rprint(f"  [dim]Cost:[/dim] ${fix_response.get('cost', 0.0):.6f}")
+                rprint("\n[blue]Fix Result:[/blue]")
+                rprint(Markdown(fix_result))
+
+            fixed_program_match = re.search(r"<fixed_program>(.*?)</fixed_program>", fix_result, re.DOTALL)
+            fixed_code_match = re.search(r"<fixed_code>(.*?)</fixed_code>", fix_result, re.DOTALL)
+            explanation_match = re.search(r"<explanation>(.*?)</explanation>", fix_result, re.DOTALL)
+
+            if fixed_program_match:
+                fixed_program = fixed_program_match.group(1).strip()
+                if verbose: rprint("[green]Extracted fixed program.[/green]")
+            else:
+                if verbose: rprint("[yellow]Warning:[/yellow] Could not find <fixed_program> tag in fix result. Using original program.")
+
+            if fixed_code_match:
+                fixed_code = fixed_code_match.group(1).strip()
+                if verbose: rprint("[green]Extracted fixed code module.[/green]")
+            else:
+                if verbose: rprint("[yellow]Warning:[/yellow] Could not find <fixed_code> tag in fix result. Using original code module.")
+
+            if explanation_match:
+                fix_explanation = explanation_match.group(1).strip()
+                if verbose: rprint("[green]Extracted fix explanation.[/green]")
+            else:
+                if verbose: rprint("[yellow]Warning:[/yellow] Could not find <explanation> tag in fix result.")
+                fix_explanation = "[Fix explanation not provided by LLM]"
+
+        except Exception as e:
+            rprint(f"[bold red]Error during fix LLM call or extraction:[/bold red] {e}")
+            if verification_details and fix_explanation is None:
+                fix_explanation = f"[Error during fix generation: {e}]"
+
+    if verbose:
+        rprint(f"\n[bold blue]Total Cost for fix_verification_errors run:[/bold blue] ${total_cost:.6f}")
+
+    if issues_found and verification_details and fix_explanation:
+        final_explanation = (
+            f"<verification_details>{verification_details}</verification_details>\n"
+            f"<fix_explanation>{fix_explanation}</fix_explanation>"
+        )
+    elif issues_found and verification_details:
+        final_explanation = (
+            f"<verification_details>{verification_details}</verification_details>\n"
+            f"<fix_explanation>[Fix explanation not available or extraction failed]</fix_explanation>"
+        )
+
+    return {
+        "explanation": final_explanation,
+        "fixed_program": fixed_program,
+        "fixed_code": fixed_code,
+        "total_cost": total_cost,
+        "model_name": model_name,
+        "verification_issues_count": verification_issues_count,
+    }