LLM Integration Architecture

def process(traceback_message: str, original_error_information: str, context: str) -> object:
    client = Groq(api_key=os.getenv("API"))
    
    chat_completion = client.chat.completions.create(
        messages=[...],  # See prompt engineering section
        model="llama3-70b-8192",
        response_format={"type": "json_object"}
    )
    
    return chat_completion.choices[0].message.content

{
    "role": "system",
    "content": "You are an expert software debugging assistant specializing in Python error analysis. Your task is to analyze error tracebacks and provide structured, actionable advice. Follow these steps precisely:"
}

{
    "role": "system",
    "content": """1. Analyze the provided error traceback and original error message.
2. Identify the source of the error within the repository structure.
3. Explain the error concisely in natural language.
4. Provide specific, actionable suggestions for resolving the error.
5. Format your response as a JSON object with the exact structure specified below."""
}

{
    "role": "system",
    "content": """Your response MUST be a valid JSON object with this exact structure:
{
  "where": {
    "repository_path": "<absolute path to repository>",
    "file_name": "<name of file containing error>",
    "line_number": "<line number where error occurred>"
  },
  "what": {
    "error_type": "<specific Python error type>",
    "description": "<concise explanation of error>"
  },
  "how": {
    "error_origination": "<line number where error originated>",
    "suggested_code_solution": "<code snippet to fix the error>"
  }
}"""
}

{
    "role": "system",
    "content": """Constraints:
- Provide ONLY the JSON object as your response. Do not include any other text.
- Ensure all JSON keys are exactly as specified.
- The 'suggested_code_solution' should be a valid Python code snippet without explanation.
- If multiple errors exist, focus on the most critical one that likely caused the others.
- Do not use placeholders in your response. Provide specific, contextual information based on the given error.
- Only fix the line from the line number given that is causing the error."""
}

{
    "role": "user",
    "content": f"Context: {context}\n\nTraceback message: {traceback_message}\n\nOriginal error message: {original_error_information}"
}

response_format={"type": "json_object"}

{
  "where": {
    "repository_path": "/Users/user/project",
    "file_name": "app.py",
    "line_number": "15"
  },
  "what": {
    "error_type": "ValueError",
    "description": "The function expected a positive integer but received a negative value"
  },
  "how": {
    "error_origination": "15",
    "suggested_code_solution": "if value < 0:\n    raise ValueError('Value must be positive')\nresult = process(value)"
  }
}

client = Groq(api_key=os.getenv("API"))

chat_completion = client.chat.completions.create(
    messages=[...],
    model="llama3-70b-8192",  # 70B parameter model with 8K context
    response_format={"type": "json_object"}
)

if os.path.exists(path):  # Only real files
    if is_project_file(path, project_root):  # Only project files
        result.append(read_file(path))

return list(dict.fromkeys(files))  # Remove duplicate file paths

from pydantic import BaseModel, Field
from pydantic_ai import Agent

class ZapperResult(BaseModel):
    path: str = Field("The path to the file where the error occurs")
    file: str = Field("The name of the file with the error")
    line: int = Field("Which line the error occured in", ge=0)
    what: str = Field("Explanation in natural language of what the error is")
    todo: str = Field("Describe steps to take in natural language in order to fix the error")

bug_zapper = Agent(
    model_name=os.getenv('MODEL'),
    api_key=os.getenv('API_KEY'),
    deps_type=FlagDependencies,
    result_type=ZapperResult,
    system_prompt=(
        'You are an expert software debugging assistant specializing in programming error analysis. '
        'Your task is to analyze error tracebacks and provide structured, actionable advice. '
    )
)

line: int = Field("Which line the error occured in", ge=0)
# Ensures line number is non-negative integer

result: ZapperResult = await bug_zapper.run(context)
print(result.path)  # IDE autocomplete works

@dataclass
class FlagDependencies:
    flag: str
    file_path: Union[str, Path]

response_format={"type": "json_object"}

- Do not use placeholders
- Do not include any other text
- Only fix the line from the line number given

Provide specific, contextual information based on the given error.

If multiple errors exist, focus on the most critical one that likely caused the others.

try:
    chat_completion = client.chat.completions.create(...)
except Exception as e:
    return {"error": str(e)}

import json
try:
    result = json.loads(response)
except json.JSONDecodeError:
    # Handle malformed response

import tiktoken
encoding = tiktoken.encoding_for_model("llama3")
if len(encoding.encode(context)) > 7000:
    # Truncate or summarize context

API=gsk_your_groq_api_key_here

from dotenv import load_dotenv
load_dotenv()

client = Groq(api_key=os.getenv("API"))