Update app.py
Browse files
app.py
CHANGED
|
@@ -1,13 +1,14 @@
|
|
| 1 |
import streamlit as st
|
| 2 |
import json
|
| 3 |
-
import
|
| 4 |
-
import
|
| 5 |
|
| 6 |
def normalize(s):
|
| 7 |
return ' '.join(str(s).lower().replace("_", " ").replace("-", " ").replace(".", " ").split())
|
| 8 |
|
| 9 |
def is_fuzzy_match(a, b, threshold=0.7):
|
| 10 |
-
|
|
|
|
| 11 |
return ratio >= threshold or a in b or b in a
|
| 12 |
|
| 13 |
def recursive_fuzzy_value_search(target_value):
|
|
@@ -17,26 +18,13 @@ def recursive_fuzzy_value_search(target_value):
|
|
| 17 |
def _search(obj, path):
|
| 18 |
if isinstance(obj, dict):
|
| 19 |
for k, v in obj.items():
|
| 20 |
-
# Match ANY primitive value (str, int, float, bool)
|
| 21 |
if isinstance(v, (str, int, float, bool)) and is_fuzzy_match(norm_target, normalize(v)):
|
| 22 |
matches.append({
|
| 23 |
-
"
|
| 24 |
-
"matched_value": v,
|
| 25 |
"key": k,
|
| 26 |
-
"
|
| 27 |
-
"
|
| 28 |
})
|
| 29 |
-
# Check inside nested dicts
|
| 30 |
-
if isinstance(v, dict):
|
| 31 |
-
for nk, nv in v.items():
|
| 32 |
-
if isinstance(nv, (str, int, float, bool)) and is_fuzzy_match(norm_target, normalize(nv)):
|
| 33 |
-
matches.append({
|
| 34 |
-
"match_path": path + [k, nk],
|
| 35 |
-
"matched_value": nv,
|
| 36 |
-
"key": nk,
|
| 37 |
-
"record": v,
|
| 38 |
-
"file": file_name
|
| 39 |
-
})
|
| 40 |
_search(v, path + [k])
|
| 41 |
elif isinstance(obj, list):
|
| 42 |
for idx, item in enumerate(obj):
|
|
@@ -44,105 +32,72 @@ def recursive_fuzzy_value_search(target_value):
|
|
| 44 |
_search(data, [])
|
| 45 |
return matches
|
| 46 |
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
for
|
| 50 |
-
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
if isinstance(nv, (str, int, float, bool)):
|
| 58 |
-
found.append(f"{file_name} | {'.'.join(path + [k, nk])} = {nv}")
|
| 59 |
-
recursive(v, path + [k])
|
| 60 |
-
elif isinstance(obj, list):
|
| 61 |
-
for idx, item in enumerate(obj):
|
| 62 |
-
recursive(item, path + [f"[{idx}]"])
|
| 63 |
-
recursive(data, [])
|
| 64 |
-
return found
|
| 65 |
-
|
| 66 |
-
def handle_user_query(query):
|
| 67 |
-
patterns = [
|
| 68 |
-
r"(?:last\s*login.*?for|when\s+did)\s+([a-zA-Z0-9 _\-\.@]+)",
|
| 69 |
-
r"when\s+was\s+([a-zA-Z0-9 _\-\.@]+)\s+last\s+(?:login|logged\s*in)",
|
| 70 |
-
r"last\s*login\s*of\s+([a-zA-Z0-9 _\-\.@]+)",
|
| 71 |
-
r"(?:info|details|record) for\s+([a-zA-Z0-9 _\-\.@]+)"
|
| 72 |
-
]
|
| 73 |
-
found_value = None
|
| 74 |
-
for pat in patterns:
|
| 75 |
-
m = re.search(pat, query, re.IGNORECASE)
|
| 76 |
-
if m:
|
| 77 |
-
found_value = m.group(1).strip()
|
| 78 |
-
break
|
| 79 |
-
if not found_value:
|
| 80 |
-
# Fallback: any word/phrase of 1+ char (letters, digits, spaces, dashes, underscores, dots)
|
| 81 |
-
m = re.search(r"([A-Za-z0-9][A-Za-z0-9 _\-\.@]*)", query)
|
| 82 |
-
if m:
|
| 83 |
-
found_value = m.group(1).strip()
|
| 84 |
-
if found_value:
|
| 85 |
-
results = recursive_fuzzy_value_search(found_value)
|
| 86 |
-
if not results:
|
| 87 |
-
return f"No records found for '{found_value}' in any file."
|
| 88 |
-
answers = []
|
| 89 |
-
for res in results:
|
| 90 |
-
answers.append(
|
| 91 |
-
f"**{res['matched_value']}** (in file `{res['file']}` | key: `{res['key']}` | path: `{'.'.join(res['match_path'])}`)"
|
| 92 |
-
)
|
| 93 |
-
return "\n\n".join(answers)
|
| 94 |
-
else:
|
| 95 |
-
return "No valid search value detected. Try a person's name, number, product, device, etc."
|
| 96 |
|
| 97 |
-
#
|
| 98 |
if "json_data" not in st.session_state:
|
| 99 |
st.session_state.json_data = {}
|
| 100 |
-
if "
|
| 101 |
-
st.session_state.
|
| 102 |
-
if "temp_input" not in st.session_state:
|
| 103 |
-
st.session_state.temp_input = ""
|
| 104 |
-
if "files_loaded" not in st.session_state:
|
| 105 |
-
st.session_state.files_loaded = False
|
| 106 |
|
| 107 |
-
st.set_page_config(page_title="
|
| 108 |
-
st.title("
|
| 109 |
|
| 110 |
uploaded_files = st.sidebar.file_uploader(
|
| 111 |
"Choose one or more JSON files", type="json", accept_multiple_files=True
|
| 112 |
)
|
| 113 |
-
if uploaded_files
|
| 114 |
st.session_state.json_data.clear()
|
| 115 |
for f in uploaded_files:
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
st.sidebar.success(f"Loaded: {f.name}")
|
| 120 |
-
except Exception as e:
|
| 121 |
-
st.sidebar.error(f"Error reading {f.name}: {e}")
|
| 122 |
-
st.session_state.messages = []
|
| 123 |
-
st.session_state.files_loaded = True
|
| 124 |
-
elif not uploaded_files:
|
| 125 |
-
st.session_state.json_data.clear()
|
| 126 |
-
st.session_state.files_loaded = False
|
| 127 |
-
|
| 128 |
-
st.markdown("### Ask about ANY value (name, product, number, device, etc) — partials/typos/substring OK!")
|
| 129 |
-
for msg in st.session_state.messages:
|
| 130 |
-
if msg["role"] == "user":
|
| 131 |
-
st.markdown(f"<div style='color: #4F8BF9;'><b>User:</b> {msg['content']}</div>", unsafe_allow_html=True)
|
| 132 |
-
else:
|
| 133 |
-
st.markdown(f"<div style='color: #1C6E4C;'><b>Agent:</b> {msg['content']}</div>", unsafe_allow_html=True)
|
| 134 |
|
| 135 |
-
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
answer = handle_user_query(user_input)
|
| 140 |
-
st.session_state.messages.append({"role": "assistant", "content": answer})
|
| 141 |
-
st.session_state.temp_input = ""
|
| 142 |
|
| 143 |
-
if st.session_state.json_data:
|
| 144 |
-
st.text_input("Your message:", key="temp_input", on_change=send_message)
|
| 145 |
-
if st.button("Show all values in uploaded JSONs"):
|
| 146 |
-
st.write(show_all_strings())
|
| 147 |
else:
|
| 148 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
import streamlit as st
|
| 2 |
import json
|
| 3 |
+
from langchain.chat_models import ChatOpenAI
|
| 4 |
+
from langchain.agents import initialize_agent, Tool
|
| 5 |
|
| 6 |
def normalize(s):
|
| 7 |
return ' '.join(str(s).lower().replace("_", " ").replace("-", " ").replace(".", " ").split())
|
| 8 |
|
| 9 |
def is_fuzzy_match(a, b, threshold=0.7):
|
| 10 |
+
from difflib import SequenceMatcher
|
| 11 |
+
ratio = SequenceMatcher(None, a, b).ratio()
|
| 12 |
return ratio >= threshold or a in b or b in a
|
| 13 |
|
| 14 |
def recursive_fuzzy_value_search(target_value):
|
|
|
|
| 18 |
def _search(obj, path):
|
| 19 |
if isinstance(obj, dict):
|
| 20 |
for k, v in obj.items():
|
|
|
|
| 21 |
if isinstance(v, (str, int, float, bool)) and is_fuzzy_match(norm_target, normalize(v)):
|
| 22 |
matches.append({
|
| 23 |
+
"file": file_name,
|
|
|
|
| 24 |
"key": k,
|
| 25 |
+
"path": ".".join(path + [k]),
|
| 26 |
+
"value": v
|
| 27 |
})
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
_search(v, path + [k])
|
| 29 |
elif isinstance(obj, list):
|
| 30 |
for idx, item in enumerate(obj):
|
|
|
|
| 32 |
_search(data, [])
|
| 33 |
return matches
|
| 34 |
|
| 35 |
+
# LangChain Tool for LLM
|
| 36 |
+
def json_search_tool(query: str) -> str:
|
| 37 |
+
"""Search all uploaded JSON files for any value (fuzzy match); returns matching fields and values."""
|
| 38 |
+
results = recursive_fuzzy_value_search(query)
|
| 39 |
+
if not results:
|
| 40 |
+
return f"No match for '{query}'."
|
| 41 |
+
answer = []
|
| 42 |
+
for res in results:
|
| 43 |
+
answer.append(f"{res['file']} | {res['key']} ({res['path']}): {res['value']}")
|
| 44 |
+
return "\n".join(answer)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 45 |
|
| 46 |
+
# Streamlit UI
|
| 47 |
if "json_data" not in st.session_state:
|
| 48 |
st.session_state.json_data = {}
|
| 49 |
+
if "chat_history" not in st.session_state:
|
| 50 |
+
st.session_state.chat_history = []
|
|
|
|
|
|
|
|
|
|
|
|
|
| 51 |
|
| 52 |
+
st.set_page_config(page_title="Chat with Your JSONs!", layout="wide")
|
| 53 |
+
st.title("Chat with Your JSON Files (powered by GPT + instant JSON search)")
|
| 54 |
|
| 55 |
uploaded_files = st.sidebar.file_uploader(
|
| 56 |
"Choose one or more JSON files", type="json", accept_multiple_files=True
|
| 57 |
)
|
| 58 |
+
if uploaded_files:
|
| 59 |
st.session_state.json_data.clear()
|
| 60 |
for f in uploaded_files:
|
| 61 |
+
content = json.load(f)
|
| 62 |
+
st.session_state.json_data[f.name] = content
|
| 63 |
+
st.sidebar.success("All JSON files loaded.")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 64 |
|
| 65 |
+
import os
|
| 66 |
+
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
|
| 67 |
+
if not OPENAI_API_KEY:
|
| 68 |
+
st.warning("You must set your OPENAI_API_KEY for chat.")
|
|
|
|
|
|
|
|
|
|
| 69 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 70 |
else:
|
| 71 |
+
llm = ChatOpenAI(model_name="gpt-4.1", openai_api_key=OPENAI_API_KEY)
|
| 72 |
+
tools = [
|
| 73 |
+
Tool(
|
| 74 |
+
name="json_search",
|
| 75 |
+
func=json_search_tool,
|
| 76 |
+
description="Find any value (name, product, number, etc) across all loaded JSON files. Input is what the user wants to find (e.g. 'iphone', 'apps installed', or 'alice')."
|
| 77 |
+
)
|
| 78 |
+
]
|
| 79 |
+
agent = initialize_agent(
|
| 80 |
+
tools=tools,
|
| 81 |
+
llm=llm,
|
| 82 |
+
agent="chat-conversational-react-description",
|
| 83 |
+
verbose=False
|
| 84 |
+
)
|
| 85 |
+
|
| 86 |
+
for msg in st.session_state.chat_history:
|
| 87 |
+
if msg["role"] == "user":
|
| 88 |
+
st.markdown(f"<div style='color: #4F8BF9;'><b>User:</b> {msg['content']}</div>", unsafe_allow_html=True)
|
| 89 |
+
else:
|
| 90 |
+
st.markdown(f"<div style='color: #1C6E4C;'><b>Agent:</b> {msg['content']}</div>", unsafe_allow_html=True)
|
| 91 |
+
|
| 92 |
+
def send_chat():
|
| 93 |
+
user_input = st.session_state.temp_input
|
| 94 |
+
if user_input.strip():
|
| 95 |
+
st.session_state.chat_history.append({"role": "user", "content": user_input})
|
| 96 |
+
agent_reply = agent.run(user_input)
|
| 97 |
+
st.session_state.chat_history.append({"role": "assistant", "content": agent_reply})
|
| 98 |
+
st.session_state.temp_input = ""
|
| 99 |
+
|
| 100 |
+
if st.session_state.json_data:
|
| 101 |
+
st.text_input("Your message:", key="temp_input", on_change=send_chat)
|
| 102 |
+
else:
|
| 103 |
+
st.info("Please upload at least one JSON file to start chatting.")
|