13ba1457b0
The embedding-service exposes endpoints at root level (/chunk, /embed, /extract-pdf, /rerank) not under /api/v1/. Fix the RAG service's embedding client to use the correct paths. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
124 lines
4.0 KiB
Python
124 lines
4.0 KiB
Python
import logging
|
|
from typing import Optional
|
|
|
|
import httpx
|
|
|
|
from config import settings
|
|
|
|
logger = logging.getLogger("rag-service.embedding")
|
|
|
|
_TIMEOUT = httpx.Timeout(timeout=120.0, connect=10.0)
|
|
|
|
|
|
class EmbeddingClient:
|
|
"""HTTP client for the embedding-service (port 8087)."""
|
|
|
|
def __init__(self) -> None:
|
|
self._base_url: str = settings.EMBEDDING_SERVICE_URL.rstrip("/")
|
|
|
|
def _url(self, path: str) -> str:
|
|
return f"{self._base_url}{path}"
|
|
|
|
# ------------------------------------------------------------------
|
|
# Embeddings
|
|
# ------------------------------------------------------------------
|
|
|
|
async def generate_embeddings(self, texts: list[str]) -> list[list[float]]:
|
|
"""
|
|
Send a batch of texts to the embedding service and return a list of
|
|
embedding vectors.
|
|
"""
|
|
async with httpx.AsyncClient(timeout=_TIMEOUT) as client:
|
|
response = await client.post(
|
|
self._url("/embed"),
|
|
json={"texts": texts},
|
|
)
|
|
response.raise_for_status()
|
|
data = response.json()
|
|
return data.get("embeddings", [])
|
|
|
|
async def generate_single_embedding(self, text: str) -> list[float]:
|
|
"""Convenience wrapper for a single text."""
|
|
results = await self.generate_embeddings([text])
|
|
if not results:
|
|
raise ValueError("Embedding service returned empty result")
|
|
return results[0]
|
|
|
|
# ------------------------------------------------------------------
|
|
# Reranking
|
|
# ------------------------------------------------------------------
|
|
|
|
async def rerank_documents(
|
|
self,
|
|
query: str,
|
|
documents: list[str],
|
|
top_k: int = 10,
|
|
) -> list[dict]:
|
|
"""
|
|
Ask the embedding service to re-rank documents for a given query.
|
|
Returns a list of {index, score, text}.
|
|
"""
|
|
async with httpx.AsyncClient(timeout=_TIMEOUT) as client:
|
|
response = await client.post(
|
|
self._url("/rerank"),
|
|
json={
|
|
"query": query,
|
|
"documents": documents,
|
|
"top_k": top_k,
|
|
},
|
|
)
|
|
response.raise_for_status()
|
|
data = response.json()
|
|
return data.get("results", [])
|
|
|
|
# ------------------------------------------------------------------
|
|
# Chunking
|
|
# ------------------------------------------------------------------
|
|
|
|
async def chunk_text(
|
|
self,
|
|
text: str,
|
|
strategy: str = "recursive",
|
|
chunk_size: int = 512,
|
|
overlap: int = 50,
|
|
) -> list[str]:
|
|
"""
|
|
Ask the embedding service to chunk a long text.
|
|
Returns a list of chunk strings.
|
|
"""
|
|
async with httpx.AsyncClient(timeout=_TIMEOUT) as client:
|
|
response = await client.post(
|
|
self._url("/chunk"),
|
|
json={
|
|
"text": text,
|
|
"strategy": strategy,
|
|
"chunk_size": chunk_size,
|
|
"overlap": overlap,
|
|
},
|
|
)
|
|
response.raise_for_status()
|
|
data = response.json()
|
|
return data.get("chunks", [])
|
|
|
|
# ------------------------------------------------------------------
|
|
# PDF extraction
|
|
# ------------------------------------------------------------------
|
|
|
|
async def extract_pdf(self, pdf_bytes: bytes) -> str:
|
|
"""
|
|
Send raw PDF bytes to the embedding service for text extraction.
|
|
Returns the extracted text as a string.
|
|
"""
|
|
async with httpx.AsyncClient(timeout=_TIMEOUT) as client:
|
|
response = await client.post(
|
|
self._url("/extract-pdf"),
|
|
files={"file": ("document.pdf", pdf_bytes, "application/pdf")},
|
|
)
|
|
response.raise_for_status()
|
|
data = response.json()
|
|
return data.get("text", "")
|
|
|
|
|
|
# Singleton
|
|
embedding_client = EmbeddingClient()
|