import json
import os
from typing import Dict, Any
from langchain_groq import ChatGroq
from langchain_core.prompts import ChatPromptTemplate
from models_schemas import AnalyzeResponse
from prompts import PROMPT_TEMPLATE, SCHEMA_JSON, PROMPT_SCORE

GROQ_MODEL_ID = os.getenv("GROQ_MODEL_ID", "deepseek-r1-distill-llama-70b")
TEMPERATURE = float(os.getenv("TEMPERATURE", "0.7"))

def _load_llm():
    return ChatGroq(
        model=GROQ_MODEL_ID,
        temperature=TEMPERATURE,
        max_tokens=None,
        timeout=None,
        max_retries=2,
    )

def _strip_think(text: str) -> str:
    if "</think>" in text:
        return text.split("</think>")[-1].strip()
    return text.strip()

def _force_schema_fields(d: Dict[str, Any]) -> Dict[str, Any]:
    keys = [
        "name","area","summary","skills","education","interview_questions",
        "strengths","areas_for_development","important_considerations",
        "final_recommendations","score"
    ]
    for k in keys:
        if k not in d:
            d[k] = [] if k in {
                "skills","interview_questions","strengths",
                "areas_for_development","important_considerations"
            } else (0.0 if k == "score" else "")
    return d

def analyze_cv_with_llm(cv_text: str, job_details: str) -> AnalyzeResponse:
    llm = _load_llm()
    prompt = ChatPromptTemplate.from_template(PROMPT_TEMPLATE)
    chain = prompt | llm
    output = chain.invoke({
        "schema": SCHEMA_JSON,
        "prompt_score": PROMPT_SCORE,
        "cv": cv_text,
        "job": job_details
    })

    raw = _strip_think(output.content)

    start = raw.find("{")
    end = raw.rfind("}")
    if start == -1 or end == -1:
        return AnalyzeResponse()

    json_str = raw[start:end+1]
    try:
        data = json.loads(json_str)
    except Exception:
        return AnalyzeResponse()

    data = _force_schema_fields(data)
    return AnalyzeResponse(**data)