src/infrastructure/adapters/together_adapter.py

import json
import logging
from typing import AsyncGenerator

from src.core.ports.llm_port import LlmPort
from src.core.config import settings

logger = logging.getLogger(__name__)

class TogetherAdapter(LlmPort):
    """
    Together AI adapter via OpenAI-compatible API.
    Free $25 credit on signup, then pay-as-you-go (very cheap).
    Available models (set TOGETHER_MODEL env var):
      meta-llama/Llama-3.3-70B-Instruct-Turbo   ← default, best quality
      mistralai/Mistral-Small-24B-Instruct-2501
      mistralai/Mixtral-8x7B-Instruct-v0.1
      mistralai/Mistral-7B-Instruct-v0.1
    Docs: https://docs.together.ai
    """
    BASE_URL = "https://api.together.xyz/v1"
    DEFAULT_MODEL = "meta-llama/Llama-3.3-70B-Instruct-Turbo"

    def __init__(self):
        self.llm = None
        key = settings.TOGETHER_API_KEY
        if key and key != "your-together-api-key-here":
            try:
                from langchain_openai import ChatOpenAI
                model = settings.TOGETHER_MODEL or self.DEFAULT_MODEL
                self.llm = ChatOpenAI(
                    api_key=key,
                    base_url=self.BASE_URL,
                    model=model,
                    temperature=0.2,
                    max_tokens=1024,
                )
                logger.info(f"✅ Together AI adapter initialized with model: {model}")
            except Exception as e:
                logger.error(f"Failed to initialize Together adapter: {e}")
        else:
            logger.warning("TOGETHER_API_KEY not set — Together AI adapter disabled.")

    def generate(self, prompt: str) -> str:
        if not self.llm:
            return "Together AI API key not configured."
        try:
            return self.llm.invoke(prompt).content
        except Exception as e:
            error_msg = str(e)
            if "429" in error_msg or "rate" in error_msg.lower():
                return "Together AI rate limit reached. Please try again shortly."
            logger.error(f"Together generate error: {e}")
            return f"Error generating response: {error_msg}"

    async def generate_stream(self, prompt: str) -> AsyncGenerator[str, None]:
        if not self.llm:
            yield f"data: {json.dumps({'token': 'Together AI API key not configured.'})}\n\n"
            yield "data: [DONE]\n\n"
            return
        try:
            for chunk in self.llm.stream(prompt):
                if hasattr(chunk, "content") and chunk.content:
                    yield f"data: {json.dumps({'token': chunk.content})}\n\n"
            yield "data: [DONE]\n\n"
        except Exception as e:
            error_msg = str(e)
            if "429" in error_msg or "rate" in error_msg.lower():
                msg = "Together AI rate limit reached. Please try again shortly."
            else:
                msg = f"Error: {error_msg}"
            yield f"data: {json.dumps({'token': msg})}\n\n"
            yield "data: [DONE]\n\n"