Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,7 +1,159 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import gradio as gr
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
3 |
-
|
4 |
-
|
5 |
|
6 |
-
|
7 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from haystack.document_stores.faiss import FAISSDocumentStore
|
2 |
+
from haystack.nodes.retriever import EmbeddingRetriever
|
3 |
+
from haystack.nodes.ranker import BaseRanker
|
4 |
+
from haystack.pipelines import Pipeline
|
5 |
+
|
6 |
+
from haystack.document_stores.base import BaseDocumentStore
|
7 |
+
from haystack.schema import Document
|
8 |
+
|
9 |
+
from typing import Optional, List
|
10 |
+
|
11 |
import gradio as gr
|
12 |
+
import numpy as np
|
13 |
+
import requests
|
14 |
+
import os
|
15 |
+
|
16 |
+
RETRIEVER_URL = os.getenv("RETRIEVER_URL")
|
17 |
+
RANKER_URL = os.getenv("RANKER_URL")
|
18 |
+
HF_TOKEN = os.getenv("HF_TOKEN")
|
19 |
+
|
20 |
+
|
21 |
+
class Retriever(EmbeddingRetriever):
|
22 |
+
def __init__(
|
23 |
+
self,
|
24 |
+
document_store: Optional[BaseDocumentStore] = None,
|
25 |
+
top_k: int = 10,
|
26 |
+
batch_size: int = 32,
|
27 |
+
scale_score: bool = True,
|
28 |
+
):
|
29 |
+
self.document_store = document_store
|
30 |
+
self.top_k = top_k
|
31 |
+
self.batch_size = batch_size
|
32 |
+
self.scale_score = scale_score
|
33 |
+
|
34 |
+
def embed_queries(self, queries: List[str]) -> np.ndarray:
|
35 |
+
response = requests.post(
|
36 |
+
RETRIEVER_URL,
|
37 |
+
json={"queries": queries, "inputs": ""},
|
38 |
+
headers={"Authorization": f"Bearer {HF_TOKEN}"},
|
39 |
+
)
|
40 |
+
|
41 |
+
arrays = np.array(response.json())
|
42 |
+
|
43 |
+
return arrays
|
44 |
+
|
45 |
+
def embed_documents(self, documents: List[Document]) -> np.ndarray:
|
46 |
+
response = requests.post(
|
47 |
+
RETRIEVER_URL,
|
48 |
+
json={"documents": [d.to_dict() for d in documents], "inputs": ""},
|
49 |
+
headers={"Authorization": f"Bearer {HF_TOKEN}"},
|
50 |
+
)
|
51 |
+
|
52 |
+
arrays = np.array(response.json())
|
53 |
+
|
54 |
+
return arrays
|
55 |
+
|
56 |
+
|
57 |
+
class Ranker(BaseRanker):
|
58 |
+
def predict(
|
59 |
+
self, query: str, documents: List[Document], top_k: Optional[int] = None
|
60 |
+
) -> List[Document]:
|
61 |
+
documents = [d.to_dict() for d in documents]
|
62 |
+
for doc in documents:
|
63 |
+
doc["embedding"] = doc["embedding"].tolist()
|
64 |
+
|
65 |
+
response = requests.post(
|
66 |
+
RANKER_URL,
|
67 |
+
json={
|
68 |
+
"query": query,
|
69 |
+
"documents": documents,
|
70 |
+
"top_k": top_k,
|
71 |
+
"inputs": "",
|
72 |
+
},
|
73 |
+
headers={"Authorization": f"Bearer {HF_TOKEN}"},
|
74 |
+
).json()
|
75 |
+
|
76 |
+
if "error" in response:
|
77 |
+
raise Exception(response["error"])
|
78 |
+
|
79 |
+
return [Document.from_dict(d) for d in response]
|
80 |
+
|
81 |
+
def predict_batch(
|
82 |
+
self,
|
83 |
+
queries: List[str],
|
84 |
+
documents: List[List[Document]],
|
85 |
+
batch_size: Optional[int] = None,
|
86 |
+
top_k: Optional[int] = None,
|
87 |
+
) -> List[List[Document]]:
|
88 |
+
documents = [[d.to_dict() for d in docs] for docs in documents]
|
89 |
+
for docs in documents:
|
90 |
+
for doc in docs:
|
91 |
+
doc["embedding"] = doc["embedding"].tolist()
|
92 |
+
|
93 |
+
response = requests.post(
|
94 |
+
RANKER_URL,
|
95 |
+
json={
|
96 |
+
"queries": queries,
|
97 |
+
"documents": documents,
|
98 |
+
"batch_size": batch_size,
|
99 |
+
"top_k": top_k,
|
100 |
+
"inputs": "",
|
101 |
+
},
|
102 |
+
).json()
|
103 |
+
|
104 |
+
if "error" in response:
|
105 |
+
raise Exception(response["error"])
|
106 |
+
|
107 |
+
return [[Document.from_dict(d) for d in docs] for docs in response]
|
108 |
+
|
109 |
+
|
110 |
+
TOP_K = 2
|
111 |
+
BATCH_SIZE = 16
|
112 |
+
EXAMPLES = [
|
113 |
+
"There is a blue house on Oxford Street.",
|
114 |
+
"Paris is the capital of France.",
|
115 |
+
"The Eiffel Tower is in Paris.",
|
116 |
+
"The Louvre is in Paris.",
|
117 |
+
"London is the capital of England.",
|
118 |
+
"Cairo is the capital of Egypt.",
|
119 |
+
"The pyramids are in Egypt.",
|
120 |
+
"The Sphinx is in Egypt.",
|
121 |
+
]
|
122 |
+
|
123 |
+
if os.path.exists("faiss_document_store.db"):
|
124 |
+
os.remove("faiss_document_store.db")
|
125 |
+
|
126 |
+
document_store = FAISSDocumentStore(embedding_dim=384, return_embedding=True)
|
127 |
+
document_store.write_documents(
|
128 |
+
[Document(content=d, id=i) for i, d in enumerate(EXAMPLES)]
|
129 |
+
)
|
130 |
+
|
131 |
+
|
132 |
+
retriever = Retriever(document_store=document_store, top_k=TOP_K, batch_size=BATCH_SIZE)
|
133 |
+
document_store.update_embeddings(retriever=retriever)
|
134 |
+
ranker = Ranker()
|
135 |
+
|
136 |
+
pipe = Pipeline()
|
137 |
+
pipe.add_node(component=retriever, name="Retriever", inputs=["Query"])
|
138 |
+
pipe.add_node(component=ranker, name="Ranker", inputs=["Retriever"])
|
139 |
+
|
140 |
+
|
141 |
+
def run(query: str) -> dict:
|
142 |
+
output = pipe.run(query=query)
|
143 |
+
|
144 |
+
return (
|
145 |
+
f"Closest document(s): {[output['documents'][i].content for i in range(TOP_K)]}"
|
146 |
+
)
|
147 |
+
|
148 |
|
149 |
+
# warm up
|
150 |
+
run("What is the capital of France?")
|
151 |
|
152 |
+
gr.Interface(
|
153 |
+
fn=run,
|
154 |
+
inputs="text",
|
155 |
+
outputs="text",
|
156 |
+
title="Pipeline",
|
157 |
+
examples=["What is the capital of France?"],
|
158 |
+
description="A pipeline for retrieving and ranking documents.",
|
159 |
+
).launch()
|