app.py

from fastapi import FastAPI
from fastapi.responses import StreamingResponse
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel
from transformers import AutoTokenizer, AutoModelForCausalLM, TextIteratorStreamer
from typing import List, Optional
import torch
import asyncio
from threading import Thread

# ── APP SETUP ─────────────────────────────────────────
app = FastAPI(title="DevOS AI", description="AI coding agent by Cool Shot System")

app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# ── MODEL LOADING ─────────────────────────────────────
MODEL_NAME = "deepseek-ai/deepseek-coder-1.3b-instruct"

print(f"Loading model: {MODEL_NAME} ...")
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForCausalLM.from_pretrained(
    MODEL_NAME,
    torch_dtype=torch.float32,  # CPU-safe
    low_cpu_mem_usage=True,
)
model.eval()
print("Model loaded ✓")

# ── SCHEMAS ───────────────────────────────────────────
class CodeRequest(BaseModel):
    code: str
    language: str = "python"
    max_tokens: int = 128

class ChatMessage(BaseModel):
    role: str       # "user" or "assistant"
    content: str

class ChatRequest(BaseModel):
    messages: List[ChatMessage]
    system: Optional[str] = ""
    max_tokens: int = 1024

# ── HELPERS ───────────────────────────────────────────
def build_prompt(messages: List[ChatMessage], system: str = "") -> str:
    prompt = system.strip() + "\n\n" if system and system.strip() else ""
    for msg in messages[-10:]:  # last 10 messages for context window
        role_label = "User" if msg.role == "user" else "DevOS AI"
        prompt += f"{role_label}: {msg.content.strip()}\n"
    prompt += "DevOS AI:"
    return prompt

# ── ROUTES ────────────────────────────────────────────

@app.get("/")
def root():
    return {
        "status": "DevOS AI is running",
        "model": MODEL_NAME,
        "endpoints": ["/complete", "/chat", "/stream"]
    }

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


# ── /complete — inline code completion ────────────────
@app.post("/complete")
def complete_code(request: CodeRequest):
    prompt = f"Continue the following {request.language} code:\n{request.code}"

    inputs = tokenizer(
        prompt,
        return_tensors="pt",
        truncation=True,
        max_length=2048
    )

    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_new_tokens=request.max_tokens,
            temperature=0.2,
            do_sample=True,
            pad_token_id=tokenizer.eos_token_id,
            eos_token_id=tokenizer.eos_token_id,
        )

    generated = tokenizer.decode(outputs[0], skip_special_tokens=True)
    suggestion = generated[len(prompt):].strip()

    return {"suggestion": suggestion}


# ── /chat — full conversation, single response ─────────
@app.post("/chat")
def chat(request: ChatRequest):
    prompt = build_prompt(request.messages, request.system)

    inputs = tokenizer(
        prompt,
        return_tensors="pt",
        truncation=True,
        max_length=2048
    )

    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_new_tokens=request.max_tokens,
            temperature=0.4,
            do_sample=True,
            pad_token_id=tokenizer.eos_token_id,
            eos_token_id=tokenizer.eos_token_id,
            repetition_penalty=1.1,
        )

    generated = tokenizer.decode(outputs[0], skip_special_tokens=True)
    reply = generated[len(prompt):].strip()

    return {"reply": reply}


# ── /stream — streaming response (SSE) ────────────────
@app.post("/stream")
async def stream_chat(request: ChatRequest):
    prompt = build_prompt(request.messages, request.system)

    inputs = tokenizer(
        prompt,
        return_tensors="pt",
        truncation=True,
        max_length=2048
    )

    streamer = TextIteratorStreamer(
        tokenizer,
        skip_prompt=True,
        skip_special_tokens=True
    )

    generation_kwargs = dict(
        **inputs,
        max_new_tokens=request.max_tokens,
        temperature=0.4,
        do_sample=True,
        pad_token_id=tokenizer.eos_token_id,
        eos_token_id=tokenizer.eos_token_id,
        repetition_penalty=1.1,
        streamer=streamer,
    )

    # Run generation in background thread so we can stream
    thread = Thread(target=model.generate, kwargs=generation_kwargs)
    thread.start()

    async def token_generator():
        for token in streamer:
            if token:
                # SSE format
                yield f"data: {token}\n\n"
            await asyncio.sleep(0)  # yield control to event loop
        yield "data: [DONE]\n\n"

    return StreamingResponse(
        token_generator(),
        media_type="text/event-stream",
        headers={
            "Cache-Control": "no-cache",
            "X-Accel-Buffering": "no",
            "Connection": "keep-alive",
        }
    )