testing done on app.py and model.py

app.py

@@ -4,28 +4,32 @@ from model import RAGModel, load_configs
 
 
 def run_on_start():
-    global r
-    global configs
-    configs = load_configs(config_file="rag.configs.yml")
-    r = RAGModel(configs)
+    
+    if "configs" not in st.session_state:
+        st.session_state.configs = configs = load_configs(config_file="rag.configs.yml")
+    if "model" not in st.session_state:
+        st.session_state.model = RAGModel(configs)
+   
+run_on_start()
 
 
 def search(query):
     g = GoogleSearch(query)
     data = g.all_page_data
-    d = Document(data, min_char_len=configs["document"]["min_char_length"])
-    st.session_state.doc = d.doc()[0]
+    d = Document(data, min_char_len=st.session_state.configs["document"]["min_char_length"])
+    st.session_state.doc = d.doc()
 
 
-st.title("LLM powred Google search")
+st.title("LLeUUNDd Google search")
 
 if "messages" not in st.session_state:
-    run_on_start()
     st.session_state.messages = []
 
 if "doc" not in st.session_state:
     st.session_state.doc = None
 
+if "refresh" not in st.session_state:
+    st.session_state.refresh = True 
 
 for message in st.session_state.messages:
     with st.chat_message(message["role"]):
@@ -36,15 +40,17 @@ if prompt := st.chat_input("Search Here insetad of Google"):
     st.chat_message("user").markdown(prompt)
     st.session_state.messages.append({"role": "user", "content": prompt})
 
-    search(prompt)
-    s, u = SemanticSearch(
-        prompt,
+    if st.session_state.refresh:
+        st.session_state.refresh = False 
+        search(prompt)
+
+    s = SemanticSearch(
         st.session_state.doc,
-        configs["model"]["embeding_model"],
-        configs["model"]["device"],
+        st.session_state.configs["model"]["embeding_model"],
+        st.session_state.configs["model"]["device"],
     )
-    topk = s.semantic_search(query=prompt, k=32)
-    output = r.answer_query(query=prompt, topk_items=topk)
+    topk, u = s.semantic_search(query=prompt, k=32)
+    output = st.session_state.model.answer_query(query=prompt, topk_items=topk)
     response = output
     with st.chat_message("assistant"):
         st.markdown(response)

model.py

@@ -4,7 +4,7 @@ from transformers import BitsAndBytesConfig
 from transformers.utils import is_flash_attn_2_available
 import yaml
 import torch
-
+import nltk
 
 def load_configs(config_file: str) -> dict:
     with open(config_file, "r") as f:
@@ -35,13 +35,16 @@ class RAGModel:
 
     def create_prompt(self, query, topk_items: list[str]):
 
-        context = "_ " + "\n-".join(c for c in topk_items)
+        context =  "\n-".join(c for c in topk_items)
 
-        base_prompt = f"""Give time for yourself to read the context and then answer the query. 
+        base_prompt = f"""You are an alternate to goole search. Your job is to answer the user query in as detailed manner as possible.
+        you have access to the internet and other relevent data related to the user's question.
+        Give time for yourself to read the context and user query and extract relevent data and then answer the query.
+        make sure your answers is as detailed as posssbile. 
         Do not return thinking process, just return the answer.
-        If you do not find the answer, or if the query is offesnsive or in any other way harmfull just return "I'm not aware of it"
-        Now use the following context items to answer the user query.
-        context: {context}.
+        Give the output structured as a Wikipedia article.
+        Now use the following context items to answer the user query
+        context: {context}
         user query : {query} 
         """
 
@@ -56,16 +59,16 @@ class RAGModel:
 
         prompt = self.create_prompt(query, topk_items)
         input_ids = self.tokenizer(prompt, return_tensors="pt").to(self.device)
-        output = self.model.generate(**input_ids, max_new_tokens=512)
+        output = self.model.generate(**input_ids, temperature=0.7, max_new_tokens=512, do_sample=True)
         text = self.tokenizer.decode(output[0])
+        text = text.replace(prompt, "").replace("<bos>", "").replace("<eos>", "")
 
-        return text
 
+        return text
 
 if __name__ == "__main__":
-
     configs = load_configs(config_file="rag.configs.yml")
-    query = "what is computer vision"
+    query = "Explain F1 racing for a beginer"
     g = GoogleSearch(query)
     data = g.all_page_data
     d = Document(data, 512)

rag.configs.yml

@@ -1,8 +1,8 @@
 document:
-  min_char_length: 333 
+  min_char_length: 512 
 
 model: 
   embeding_model: all-mpnet-base-v2
-  genration_model: google/gemma-2b-it
-  device: mps 
+  genration_model: google/gemma-7b-it
+  device : cuda
   

requirments.txt

@@ -1 +1,76 @@
-beautifulsoup4=4.12.3
+
+beautifulsoup4==4.12.3
+accelerate==0.29.2
+altair==5.3.0
+attrs==23.2.0
+beautifulsoup4==4.12.3
+bitsandbytes==0.42.0
+blinker==1.7.0
+Brotli @ file:///Users/runner/miniforge3/conda-bld/brotli-split_1625213545710/work
+cachetools==5.3.3
+certifi @ file:///home/conda/feedstock_root/build_artifacts/certifi_1707022139797/work/certifi
+cffi @ file:///private/var/folders/k1/30mswbxs7r1g6zwn8y4fyt500000gp/T/abs_b4nang6w_y/croot/cffi_1700254307954/work
+charset-normalizer @ file:///home/conda/feedstock_root/build_artifacts/charset-normalizer_1698833585322/work
+click==8.1.7
+colorama @ file:///home/conda/feedstock_root/build_artifacts/colorama_1666700638685/work
+filelock @ file:///home/conda/feedstock_root/build_artifacts/filelock_1711394622191/work
+fsspec @ file:///home/conda/feedstock_root/build_artifacts/fsspec_1710808267764/work
+gitdb==4.0.11
+GitPython==3.1.43
+huggingface_hub @ file:///home/conda/feedstock_root/build_artifacts/huggingface_hub_1711986612800/work
+idna @ file:///home/conda/feedstock_root/build_artifacts/idna_1701026962277/work
+importlib_metadata @ file:///home/conda/feedstock_root/build_artifacts/importlib-metadata_1710971335535/work
+Jinja2 @ file:///home/conda/feedstock_root/build_artifacts/jinja2_1704966972576/work
+joblib==1.3.2
+jsonschema==4.21.1
+jsonschema-specifications==2023.12.1
+markdown-it-py==3.0.0
+MarkupSafe @ file:///private/var/folders/nz/j6p8yfhx1mv_0grj5xl4650h0000gp/T/abs_a84ni4pci8/croot/markupsafe_1704206002077/work
+mdurl==0.1.2
+mpmath @ file:///home/conda/feedstock_root/build_artifacts/mpmath_1678228039184/work
+networkx @ file:///home/conda/feedstock_root/build_artifacts/networkx_1698504735452/work
+nltk==3.8.1
+numpy @ file:///private/var/folders/k1/30mswbxs7r1g6zwn8y4fyt500000gp/T/abs_a51i_mbs7m/croot/numpy_and_numpy_base_1708638620867/work/dist/numpy-1.26.4-cp39-cp39-macosx_11_0_arm64.whl#sha256=829e20a6c33ce51c1a93497d06cb4af22d84caa54a431ea062765da3134e5287
+packaging @ file:///home/conda/feedstock_root/build_artifacts/packaging_1710075952259/work
+pandas==2.2.1
+pillow @ file:///private/var/folders/nz/j6p8yfhx1mv_0grj5xl4650h0000gp/T/abs_e02b4k5qik/croot/pillow_1707233036487/work
+protobuf==4.25.3
+psutil @ file:///private/var/folders/nz/j6p8yfhx1mv_0grj5xl4650h0000gp/T/abs_1310b568-21f4-4cb0-b0e3-2f3d31e39728k9coaga5/croots/recipe/psutil_1656431280844/work
+pyarrow==15.0.2
+pycparser @ file:///home/conda/feedstock_root/build_artifacts/pycparser_1711811537435/work
+pydeck==0.8.1b0
+Pygments==2.17.2
+PySocks @ file:///home/conda/feedstock_root/build_artifacts/pysocks_1661604839144/work
+python-dateutil==2.9.0.post0
+pytz==2024.1
+PyYAML==5.4.1
+referencing==0.34.0
+regex==2023.12.25
+requests @ file:///home/conda/feedstock_root/build_artifacts/requests_1684774241324/work
+rich==13.7.1
+rpds-py==0.18.0
+safetensors @ file:///private/var/folders/k1/30mswbxs7r1g6zwn8y4fyt500000gp/T/abs_09qdt_s9t7/croot/safetensors_1708633848061/work
+scikit-learn @ file:///private/var/folders/k1/30mswbxs7r1g6zwn8y4fyt500000gp/T/abs_60ynh176wd/croot/scikit-learn_1694789615217/work
+scipy @ file:///private/var/folders/k1/30mswbxs7r1g6zwn8y4fyt500000gp/T/abs_41w43uybvr/croot/scipy_1710947318888/work/dist/scipy-1.12.0-cp39-cp39-macosx_11_0_arm64.whl#sha256=73d83606c8528425eb69a034da182c70ebf79b1a85019adc1f5f32a1329c830c
+sentence-transformers @ file:///home/conda/feedstock_root/build_artifacts/sentence-transformers_1711454085860/work
+sentencepiece @ file:///private/var/folders/nz/j6p8yfhx1mv_0grj5xl4650h0000gp/T/abs_dek8463j1w/croot/sentencepiece_1684523571928/work/python
+six==1.16.0
+smmap==5.0.1
+soupsieve==2.5
+streamlit==1.33.0
+sympy @ file:///home/conda/feedstock_root/build_artifacts/sympy_1684180539862/work
+tenacity==8.2.3
+threadpoolctl @ file:///home/conda/feedstock_root/build_artifacts/threadpoolctl_1710943558485/work
+tokenizers @ file:///private/var/folders/nz/j6p8yfhx1mv_0grj5xl4650h0000gp/T/abs_77bzam0w9g/croot/tokenizers_1708633828244/work
+toml==0.10.2
+toolz==0.12.1
+torch==2.2.2
+torchaudio==2.2.2
+torchvision @ file:///private/var/folders/k1/30mswbxs7r1g6zwn8y4fyt500000gp/T/abs_cfzx6ndngz/croot/torchvision_1689077985227/work
+tornado==6.4
+tqdm==4.66.2
+transformers==4.39.3
+typing_extensions==4.10.0
+tzdata==2024.1
+urllib3 @ file:///home/conda/feedstock_root/build_artifacts/urllib3_1708239446578/work
+zipp @ file:///home/conda/feedstock_root/build_artifacts/zipp_1695255097490/work