analyzer.py

import openai
import os
import json
import re

def analyze_code(code: str) -> str:
    """
    Uses OpenAI's GPT-4.1 mini model to analyze the given code.
    Returns the analysis as a string.
    """
    from openai import OpenAI
    client = OpenAI(api_key=os.getenv("modal_api"))
    client.base_url = os.getenv("base_url")
    system_prompt = (
        "You are a highly precise and strict JSON generator. Analyze the code given to you. "
        "Your ONLY output must be a valid JSON object with the following keys: 'strength', 'weaknesses', 'speciality', 'relevance rating'. "
        "Do NOT include any explanation, markdown, or text outside the JSON. Do NOT add any commentary, preamble, or postscript. "
        "If you cannot answer, still return a valid JSON with empty strings for each key. "
        "Example of the ONLY valid output:\n"
        "{\n  'strength': '...', \n  'weaknesses': '...', \n  'speciality': '...', \n  'relevance rating': '...'\n}"
    )
    response = client.chat.completions.create(
        model="Orion-zhen/Qwen2.5-Coder-7B-Instruct-AWQ",  # Updated model
        messages=[
            {"role": "system", "content": system_prompt},
            {"role": "user", "content": code}
        ],
        max_tokens=512,
        temperature=0.7
    )
    return response.choices[0].message.content

def parse_llm_json_response(response: str):
    try:
        # Extract only the substring between the first '{' and the last '}'
        print("DEBUGGGGG ::: ",response)
        start = response.find('{')
        end = response.rfind('}')
        if start != -1 and end != -1 and end > start:
            json_str = response[start:end+1]
        else:
            json_str = response
        # Replace single quotes with double quotes for JSON keys/values
        json_str = re.sub(r"(?<!\\)'", '"', json_str)
        return json.loads(json_str)
    except Exception as e:
        print("DEBUGGGGG error ::: ", e)
        return {"error": f"Failed to parse JSON: {e}", "raw": response}

def combine_repo_files_for_llm(repo_dir="repo_files", output_file="combined_repo.txt"):
    """
    Combines all .py and .md files in the given directory (recursively) into a single text file.
    Returns the path to the combined file.
    """
    combined_content = []
    seen_files = set()
    # Priority files
    priority_files = ["app.py", "README.md"]
    for pf in priority_files:
        pf_path = os.path.join(repo_dir, pf)
        if os.path.isfile(pf_path):
            try:
                with open(pf_path, "r", encoding="utf-8") as f:
                    combined_content.append(f"\n# ===== File: {pf} =====\n")
                    combined_content.append(f.read())
                seen_files.add(os.path.abspath(pf_path))
            except Exception as e:
                combined_content.append(f"\n# Could not read {pf_path}: {e}\n")
    # All other .py and .md files
    for root, _, files in os.walk(repo_dir):
        for file in files:
            if file.endswith(".py") or file.endswith(".md"):
                file_path = os.path.join(root, file)
                abs_path = os.path.abspath(file_path)
                if abs_path in seen_files:
                    continue
                try:
                    with open(file_path, "r", encoding="utf-8") as f:
                        combined_content.append(f"\n# ===== File: {file} =====\n")
                        combined_content.append(f.read())
                    seen_files.add(abs_path)
                except Exception as e:
                    combined_content.append(f"\n# Could not read {file_path}: {e}\n")
    with open(output_file, "w", encoding="utf-8") as out_f:
        out_f.write("\n".join(combined_content))
    return output_file

def analyze_code_chunk(code: str) -> str:
    """
    Analyzes a code chunk and returns a JSON summary for that chunk.
    """
    from openai import OpenAI
    client = OpenAI(api_key=os.getenv("modal_api"))
    client.base_url = os.getenv("base_url")
    chunk_prompt = (
        "You are a highly precise and strict JSON generator. Analyze the following code chunk. "
        "Your ONLY output must be a valid JSON object with the following keys: 'strength', 'weaknesses', 'speciality', 'relevance rating'. "
        "All property names and string values MUST use double quotes (\"). Do NOT use single quotes. "
        "Do NOT include any explanation, markdown, or text outside the JSON. Do NOT add any commentary, preamble, or postscript. "
        "If you cannot answer, still return a valid JSON with empty strings for each key. "
        "Example of the ONLY valid output:\n"
        '{\n  "strength": "...", \n  "weaknesses": "...", \n  "speciality": "...", \n  "relevance rating": "..."\n}'
    )
    response = client.chat.completions.create(
        model="Orion-zhen/Qwen2.5-Coder-7B-Instruct-AWQ",
        messages=[
            {"role": "system", "content": chunk_prompt},
            {"role": "user", "content": code}
        ],
        max_tokens=512,
        temperature=0.7
    )
    return response.choices[0].message.content

def aggregate_chunk_analyses(chunk_jsons: list) -> str:
    """
    Aggregates a list of chunk JSONs into a single JSON summary using the LLM.
    """
    from openai import OpenAI
    client = OpenAI(api_key=os.getenv("modal_api"))
    client.base_url = os.getenv("base_url")
    aggregation_prompt = (
        "You are a highly precise and strict, code analyzer and JSON generator. You are given a list of JSON analyses of code chunks. "
        "Aggregate these into a SINGLE overall JSON summary with the same keys: 'strength', 'weaknesses', 'speciality', 'relevance rating'. "
        "All property names and string values MUST use double quotes (\"). Do NOT use single quotes. "
        "Summarize and combine the information from all chunks. Do NOT include any explanation, markdown, or text outside the JSON. "
        "If a key is missing in all chunks, use an empty string. "
        "Example of the ONLY valid output:\n"
        '{\n  "strength": "...", \n  "weaknesses": "...", \n  "speciality": "...", \n  "relevance rating": "..."\n}'
    )
    user_content = "Here are the chunk analyses:\n" + "\n".join(chunk_jsons)
    response = client.chat.completions.create(
        model="Orion-zhen/Qwen2.5-Coder-7B-Instruct-AWQ",
        messages=[
            {"role": "system", "content": aggregation_prompt},
            {"role": "user", "content": user_content}
        ],
        max_tokens=512,
        temperature=0.3
    )
    return response.choices[0].message.content

def analyze_combined_file(output_file="combined_repo.txt"):
    """
    Reads the combined file, splits it into 500-line chunks, analyzes each chunk, and aggregates the LLM's output into a final summary.
    Returns the chunk JSONs (for debugging) and the aggregated analysis as a string.
    """
    try:
        with open(output_file, "r", encoding="utf-8") as f:
            lines = f.readlines()
        chunk_size = 500
        chunk_jsons = []
        for i in range(0, len(lines), chunk_size):
            chunk = "".join(lines[i:i+chunk_size])
            analysis = analyze_code_chunk(chunk)
            chunk_jsons.append(analysis)
        final_summary = aggregate_chunk_analyses(chunk_jsons)
        debug_output = (
            "==== Chunk JSON Outputs ===="
            + "\n\n".join([f"Chunk {i+1} JSON:\n{chunk_jsons[i]}" for i in range(len(chunk_jsons))])
            + "\n\n==== Final Aggregated Summary ===="
            + f"\n{final_summary}"
        )
        return debug_output
    except Exception as e:
        return f"Error analyzing combined file: {e}"