Spaces:
Build error
Build error
Update app.py
Browse files
app.py
CHANGED
@@ -1,21 +1,25 @@
|
|
|
|
1 |
import dspy
|
2 |
import gradio as gr
|
3 |
import chromadb
|
4 |
import fitz # PyMuPDF
|
5 |
-
from sentence_transformers import SentenceTransformer
|
6 |
import json
|
7 |
-
import
|
8 |
from dspy import Example, MIPROv2, Evaluate, evaluate
|
|
|
|
|
9 |
|
10 |
-
#
|
11 |
-
HF_TOKEN = os.environ
|
12 |
-
|
13 |
-
# إعداد نموذج اللغة باستخدام Hugging Face Inference API
|
14 |
dspy.settings.configure(
|
15 |
-
lm=
|
|
|
|
|
|
|
|
|
16 |
)
|
17 |
|
18 |
-
# إعداد قاعدة Chroma
|
19 |
client = chromadb.PersistentClient(path="./chroma_db")
|
20 |
col = client.get_or_create_collection(name="arabic_docs")
|
21 |
|
@@ -35,7 +39,7 @@ def process_pdf(pdf_bytes):
|
|
35 |
|
36 |
# إدخال البيانات في Chroma
|
37 |
def ingest(pdf_file):
|
38 |
-
pdf_bytes = pdf_file
|
39 |
texts = process_pdf(pdf_bytes)
|
40 |
embeddings = embedder.encode(texts, show_progress_bar=True)
|
41 |
for i, (chunk, emb) in enumerate(zip(texts, embeddings)):
|
@@ -72,7 +76,7 @@ def answer(question):
|
|
72 |
out = model(question)
|
73 |
return out.answer
|
74 |
|
75 |
-
# تحميل
|
76 |
def load_dataset(path):
|
77 |
with open(path, "r", encoding="utf-8") as f:
|
78 |
return [Example(**json.loads(l)).with_inputs("question") for l in f]
|
@@ -89,7 +93,7 @@ def optimize(train_file, val_file):
|
|
89 |
|
90 |
# واجهة Gradio
|
91 |
with gr.Blocks() as demo:
|
92 |
-
gr.Markdown("## 🧠 نظام RAG عربي باستخدام DSPy + ChromaDB +
|
93 |
|
94 |
with gr.Tab("📥 تحميل وتخزين"):
|
95 |
pdf_input = gr.File(label="ارفع ملف PDF", type="binary")
|
|
|
1 |
+
import os
|
2 |
import dspy
|
3 |
import gradio as gr
|
4 |
import chromadb
|
5 |
import fitz # PyMuPDF
|
|
|
6 |
import json
|
7 |
+
from sentence_transformers import SentenceTransformer
|
8 |
from dspy import Example, MIPROv2, Evaluate, evaluate
|
9 |
+
from litellm import completion # Ensure LiteLLM is installed
|
10 |
+
from dspy.lm import LiteLLM
|
11 |
|
12 |
+
# إعداد نموذج اللغة باستخدام LiteLLM + Hugging Face
|
13 |
+
HF_TOKEN = os.environ.get("HF_TOKEN")
|
|
|
|
|
14 |
dspy.settings.configure(
|
15 |
+
lm=LiteLLM(
|
16 |
+
model="HuggingFaceH4/zephyr-7b-beta",
|
17 |
+
api_base="https://api-inference.huggingface.co/v1",
|
18 |
+
api_key=HF_TOKEN
|
19 |
+
)
|
20 |
)
|
21 |
|
22 |
+
# إعداد قاعدة بيانات Chroma
|
23 |
client = chromadb.PersistentClient(path="./chroma_db")
|
24 |
col = client.get_or_create_collection(name="arabic_docs")
|
25 |
|
|
|
39 |
|
40 |
# إدخال البيانات في Chroma
|
41 |
def ingest(pdf_file):
|
42 |
+
pdf_bytes = pdf_file
|
43 |
texts = process_pdf(pdf_bytes)
|
44 |
embeddings = embedder.encode(texts, show_progress_bar=True)
|
45 |
for i, (chunk, emb) in enumerate(zip(texts, embeddings)):
|
|
|
76 |
out = model(question)
|
77 |
return out.answer
|
78 |
|
79 |
+
# تحميل بيانات التدريب والتحقق
|
80 |
def load_dataset(path):
|
81 |
with open(path, "r", encoding="utf-8") as f:
|
82 |
return [Example(**json.loads(l)).with_inputs("question") for l in f]
|
|
|
93 |
|
94 |
# واجهة Gradio
|
95 |
with gr.Blocks() as demo:
|
96 |
+
gr.Markdown("## 🧠 نظام RAG عربي باستخدام DSPy + ChromaDB + Hugging Face")
|
97 |
|
98 |
with gr.Tab("📥 تحميل وتخزين"):
|
99 |
pdf_input = gr.File(label="ارفع ملف PDF", type="binary")
|