haiku.rag/src/haiku/rag/reranking/vllm.py
2025-10-23 13:30:21 +03:00

44 lines
1.5 KiB
Python

import httpx
from haiku.rag.config import Config
from haiku.rag.reranking.base import RerankerBase
from haiku.rag.store.models.chunk import Chunk
class VLLMReranker(RerankerBase):
def __init__(self, model: str):
self._model = model
self._base_url = Config.providers.vllm.rerank_base_url
async def rerank(
self, query: str, chunks: list[Chunk], top_n: int = 10
) -> list[tuple[Chunk, float]]:
if not chunks:
return []
# Prepare documents for reranking
documents = [chunk.content for chunk in chunks]
async with httpx.AsyncClient() as client:
response = await client.post(
f"{self._base_url}/v1/rerank",
json={"model": self._model, "query": query, "documents": documents},
headers={
"accept": "application/json",
"Content-Type": "application/json",
},
)
response.raise_for_status()
result = response.json()
# Extract scores and pair with chunks
scored_chunks = []
for item in result.get("results", []):
index = item["index"]
score = item["relevance_score"]
scored_chunks.append((chunks[index], score))
# Sort by score (descending) and return top_n
scored_chunks.sort(key=lambda x: x[1], reverse=True)
return scored_chunks[:top_n]