app/services/lancer.py

"""
Lancer Deep Research Service
Integrates with Lancer Search API for AI-powered research
"""
import httpx
from typing import Optional, List, Dict, Any
from dataclasses import dataclass


LANCER_BASE_URL = "https://madras1-lancer.hf.space"


@dataclass
class SearchResult:
    """Individual search result from Lancer"""
    title: str
    url: str
    content: str
    score: float
    published_date: Optional[str] = None


@dataclass
class ResearchResponse:
    """Response from Lancer research/search"""
    query: str
    answer: Optional[str]
    results: List[SearchResult]
    citations: List[Dict[str, Any]]
    processing_time_ms: float
    raw_text: str  # Combined text for NER extraction


async def search(
    query: str,
    max_results: int = 10,
    freshness: str = "any"
) -> ResearchResponse:
    """
    Perform a search with AI synthesis using Lancer API.
    """
    try:
        async with httpx.AsyncClient(timeout=60.0) as client:
            response = await client.post(
                f"{LANCER_BASE_URL}/api/v1/search",
                json={
                    "query": query,
                    "max_results": max_results,
                    "freshness": freshness,
                    "include_answer": True
                }
            )
            
            if response.status_code != 200:
                raise Exception(f"Lancer API error: {response.status_code}")
            
            data = response.json()
            
            results = [
                SearchResult(
                    title=r.get("title", ""),
                    url=r.get("url", ""),
                    content=r.get("content", ""),
                    score=r.get("score", 0.0),
                    published_date=r.get("published_date")
                )
                for r in data.get("results", [])
            ]
            
            # Combine all text for NER
            raw_text = data.get("answer", "") or ""
            for r in results:
                raw_text += f"\n{r.title}. {r.content}"
            
            return ResearchResponse(
                query=data.get("query", query),
                answer=data.get("answer"),
                results=results,
                citations=data.get("citations", []),
                processing_time_ms=data.get("processing_time_ms", 0),
                raw_text=raw_text
            )
            
    except Exception as e:
        raise Exception(f"Lancer search failed: {str(e)}")


async def deep_research(
    query: str,
    max_dimensions: int = 5,
    max_sources_per_dim: int = 5
) -> ResearchResponse:
    """
    Perform deep multi-dimensional research using Lancer API.
    This provides richer, more comprehensive analysis.
    """
    try:
        async with httpx.AsyncClient(timeout=120.0) as client:
            response = await client.post(
                f"{LANCER_BASE_URL}/api/v1/research/deep",
                json={
                    "query": query,
                    "max_dimensions": max_dimensions,
                    "max_sources_per_dim": max_sources_per_dim,
                    "max_total_searches": 20
                }
            )
            
            if response.status_code != 200:
                raise Exception(f"Lancer API error: {response.status_code}")
            
            data = response.json()
            
            # Deep research returns a different format - adapt it
            results = []
            raw_text = ""
            
            # Extract from dimensions if present
            if "dimensions" in data:
                for dim in data["dimensions"]:
                    dim_name = dim.get("dimension", "")
                    raw_text += f"\n## {dim_name}\n"
                    for r in dim.get("results", []):
                        results.append(SearchResult(
                            title=r.get("title", ""),
                            url=r.get("url", ""),
                            content=r.get("content", ""),
                            score=r.get("score", 0.0)
                        ))
                        raw_text += f"{r.get('title', '')}. {r.get('content', '')}\n"
            
            # Add final report
            final_report = data.get("final_report", data.get("report", ""))
            if final_report:
                raw_text = final_report + "\n" + raw_text
            
            return ResearchResponse(
                query=query,
                answer=final_report,
                results=results,
                citations=data.get("citations", []),
                processing_time_ms=data.get("processing_time_ms", 0),
                raw_text=raw_text
            )
            
    except Exception as e:
        raise Exception(f"Lancer deep research failed: {str(e)}")


async def heavy_search(
    query: str,
    max_results: int = 5
) -> ResearchResponse:
    """
    Heavy search with full content scraping from sources.
    Slower but provides more context.
    """
    try:
        async with httpx.AsyncClient(timeout=90.0) as client:
            response = await client.post(
                f"{LANCER_BASE_URL}/api/v1/search/heavy",
                json={
                    "query": query,
                    "max_results": max_results,
                    "include_answer": True
                }
            )
            
            if response.status_code != 200:
                raise Exception(f"Lancer API error: {response.status_code}")
            
            data = response.json()
            
            results = [
                SearchResult(
                    title=r.get("title", ""),
                    url=r.get("url", ""),
                    content=r.get("content", ""),
                    score=r.get("score", 0.0)
                )
                for r in data.get("results", [])
            ]
            
            raw_text = data.get("answer", "") or ""
            for r in results:
                raw_text += f"\n{r.title}. {r.content}"
            
            return ResearchResponse(
                query=query,
                answer=data.get("answer"),
                results=results,
                citations=data.get("citations", []),
                processing_time_ms=data.get("processing_time_ms", 0),
                raw_text=raw_text
            )
            
    except Exception as e:
        raise Exception(f"Lancer heavy search failed: {str(e)}")