app/main.py

# app/main.py

import os
import json
import logging
import asyncio
from typing import Optional

import uvicorn
from fastapi import FastAPI, HTTPException
from fastapi.responses import StreamingResponse
from pydantic import BaseModel

from .model import PlutusModel, SummaryModel
from .recommender import Recommender


logging.basicConfig(level=logging.INFO)
logger = logging.getLogger("plutus.api")

_CACHE_DIR = os.getenv("HF_HOME", "/home/user/app")

DEFAULT_RECOMMEND_JSON = os.getenv(_CACHE_DIR, "recommend.json")
RECOMMEND_INDEX_PATH = os.path.join(_CACHE_DIR, "plutus_recommend_index.faiss")
RECOMMEND_META_PATH = os.path.join(_CACHE_DIR, "plutus_recommend_meta.json")





class GenerateCache:
    last_query: Optional[str] = None
    last_topic: Optional[str] = None
    last_personality: Optional[str] = None
    last_level: Optional[str] = None
    last_output: Optional[str] = None


GEN_CACHE = GenerateCache()




logger.info("Loading shared Plutus LLM and recommender...")

plutus_model = PlutusModel()
summary_model = SummaryModel()

recommender = Recommender(
    recommend_json_path=DEFAULT_RECOMMEND_JSON,
    index_path=RECOMMEND_INDEX_PATH,
    meta_path=RECOMMEND_META_PATH
)

app = FastAPI(title="Plutus Learner API")




class GenerateRequest(BaseModel):
    personality: str
    level: str
    topic: str
    query: str
    max_new_tokens: int = 700
    temperature: float = 0.5
    top_p: float = 0.9


class RecommendRequest(BaseModel):
    top_k: int = 5


class SummaryRequest(BaseModel):
    top_k: int = 5




@app.get("/health")
async def health():
    return {
        "status": "ok",
        "device": plutus_model.device
    }




@app.post("/generate")
async def generate(req: GenerateRequest):

    prompt = plutus_model.create_prompt(
        req.personality,
        req.level,
        req.topic,
        req.query
    )

    async def event_generator():
        full_text = ""

        for chunk in plutus_model.generate(
            prompt,
            max_new_tokens=req.max_new_tokens,
            temperature=req.temperature,
            top_p=req.top_p
        ):
            full_text += chunk + "\n"

            yield f"data: {json.dumps({'text': chunk})}\n\n"
            await asyncio.sleep(0)

        # Cache final result
        GEN_CACHE.last_query = req.query
        GEN_CACHE.last_topic = req.topic
        GEN_CACHE.last_personality = req.personality
        GEN_CACHE.last_level = req.level
        GEN_CACHE.last_output = full_text.strip()

        yield "data: [DONE]\n\n"

    return StreamingResponse(
        event_generator(),
        media_type="text/event-stream"
    )



@app.post("/recommend")
async def recommend(req: RecommendRequest):

    if GEN_CACHE.last_query is None:
        raise HTTPException(400, "No query found. Call /generate first.")

    results = recommender.recommend_for_query(
        query=GEN_CACHE.last_query,
        top_k=req.top_k,
        topic_boost=GEN_CACHE.last_topic
    )

    return {
        "query": GEN_CACHE.last_query,
        "results": [
            {"topic": r["topic"], "type": r["type"], "url": r["url"]}
            for r in results
        ]
    }




@app.post("/summary")
async def summary(req: SummaryRequest):

    if GEN_CACHE.last_output is None:
        raise HTTPException(400, "No generate output found. Call /generate first.")

    recs = recommender.recommend_for_query(
        query=GEN_CACHE.last_query,
        top_k=req.top_k,
        topic_boost=GEN_CACHE.last_topic
    )

    async def event_generator():
        for chunk in summary_model.summarize_text(
            full_teaching=GEN_CACHE.last_output,
            topic=GEN_CACHE.last_topic,
            level=GEN_CACHE.last_level,
            recommended=recs,
            max_new_tokens=300
        ):
            yield f"data: {json.dumps({'summary': chunk})}\n\n"
            await asyncio.sleep(0)

        yield "data: [DONE]\n\n"

    return StreamingResponse(
        event_generator(),
        media_type="text/event-stream"
    )




@app.post("/admin/build_index")
async def build_index(force: bool = False):
    recommender.build_index(force=force)
    return {"indexed": len(recommender.meta)}


if __name__ == "__main__":
    uvicorn.run("app.main:app", host="0.0.0.0", port=7860)