app/services/chat_executor.py

from __future__ import annotations

from app.core.config import get_settings
from app.core.llm import build_llm
from app.agent.agent import run_optimized_agent
from app.tools.search import build_web_search_tool
from app.tools.scraper import fetch_url
from app.tools.image_tool import analyze_image
from app.tools.wikipedia_tool import wikipedia_search
from app.tools.arxiv_tool import arxiv_search
from app.utils.logger import get_logger

logger = get_logger(__name__)


def _build_executor(
    temperature: float,
    max_results: int,
    tool_names: list[str] | None = None,
    model_name: str | None = None,
    max_iterations: int = 5,
    compress_chars: int = 1200,
    tool_output_limit: int = 1500,
    force_tools: bool = False,
) -> dict:
    settings = get_settings()

    available = {
        "web_search": build_web_search_tool(settings, max_results),
        "fetch_url": fetch_url,
        "analyze_image": analyze_image,
        "wikipedia": wikipedia_search,
        "arxiv": arxiv_search,
    }

    if tool_names:
        tools = [available[name] for name in tool_names if name in available]
    else:
        tools = [available["web_search"], available["fetch_url"], available["analyze_image"]]

    llm = build_llm(settings, temperature, tools, model_name=model_name)

    def run_fn(prompt: str, chat_history: str = "") -> tuple[str, list]:
        return run_optimized_agent(
            tools=tools,
            llm=llm,
            prompt=prompt,
            chat_history=chat_history,
            max_iterations=max_iterations,
            compress_chars=compress_chars,
            tool_output_limit=tool_output_limit,
            force_tools=force_tools,
        )

    return {
        "tools": tools,
        "llm": llm,
        "run_fn": run_fn,
        "max_iterations": max_iterations,
        "compress_chars": compress_chars,
        "tool_output_limit": tool_output_limit,
    }


def build_compressed_chat_executor(
    temperature: float,
    max_results: int,
    tool_names: list[str] | None = None,
    model_name: str | None = None,
    max_iterations: int = 5,
    compress_chars: int = 1200,
    tool_output_limit: int = 1500,
    force_tools: bool = False,
) -> dict:
    logger.info("Building chat executor (max_iterations=%d)", max_iterations)
    return _build_executor(temperature, max_results, tool_names, model_name, max_iterations, compress_chars, tool_output_limit, force_tools)


def build_optimized_executor(
    temperature: float,
    max_results: int,
    tool_names: list[str] | None = None,
    model_name: str | None = None,
    max_iterations: int = 8,
    compress_chars: int = 1200,
    tool_output_limit: int = 1500,
    force_tools: bool = False,
) -> dict:
    logger.info("Building optimized executor (max_iterations=%d)", max_iterations)
    return _build_executor(temperature, max_results, tool_names, model_name, max_iterations, compress_chars, tool_output_limit, force_tools)