File size: 11,904 Bytes
76b068b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 |
# app.py
import os
import io
import json
import re
from datetime import datetime
import pytz
from typing import Optional
from fastapi import FastAPI, UploadFile, File, Form, Request
from fastapi.responses import JSONResponse, HTMLResponse
from fastapi.templating import Jinja2Templates
from PIL import Image
# google-genai
from google import genai
from google.genai import types
# =====================================================
# CONFIG
# =====================================================
API_KEY = os.environ.get("GENAI_API_KEY", "AIzaSyCjMsYC-mDTwOr1at1-91EkMwI2O6eOvXg")
MODEL = os.environ.get("GENAI_MODEL", "gemini-2.5-flash")
client = genai.Client(api_key=API_KEY)
# =====================================================
# MINI-AI TOOL FUNCTIONS (unchanged behavior)
# =====================================================
def time_tool(location: str = "UTC") -> dict:
if location and "india" in location.lower():
tz = pytz.timezone("Asia/Kolkata")
else:
tz = pytz.utc
now = datetime.now(tz)
return {
"date": now.strftime("%Y-%m-%d"),
"time_24": now.strftime("%H:%M:%S"),
"time_12": now.strftime("%I:%M:%S %p"),
"timezone": str(tz)
}
def date_tool(location: str = "UTC") -> dict:
if location and "india" in location.lower():
tz = pytz.timezone("Asia/Kolkata")
else:
tz = pytz.utc
now = datetime.now(tz)
return {"date": now.strftime("%A, %d-%m-%Y"), "timezone": str(tz)}
def math_tool(expression: str) -> dict:
try:
allowed_names = {}
value = eval(expression, {"__builtins__": None}, allowed_names)
return {"expression": expression, "result": str(value)}
except Exception:
return {"expression": expression, "error": "Could not evaluate expression."}
def weather_tool(location: str) -> dict:
return {"location": location, "temperature": 25, "unit": "C", "note": "dummy data; integrate a weather API for real results."}
# =====================================================
# LLM wrappers
# =====================================================
def generate_text(system_instruction: str, content: str) -> str:
cfg = types.GenerateContentConfig(system_instruction=system_instruction)
resp = client.models.generate_content(model=MODEL, config=cfg, contents=content)
return getattr(resp, "text", "").strip()
def grounded_search(query: str) -> str:
grounding_tool = types.Tool(google_search=types.GoogleSearch())
cfg = types.GenerateContentConfig(tools=[grounding_tool])
resp = client.models.generate_content(model=MODEL, config=cfg, contents=query)
return getattr(resp, "text", "").strip()
# =====================================================
# Router logic (kept same as your code)
# =====================================================
import re as _re
FACTUAL_KEYWORDS = _re.compile(
r"\b(time|date|today|now|what's the time|what is the time|weather|forecast|temperature|convert|calculate|solve|sum|add|subtract|multiply|divide|what is)\b",
flags=_re.I
)
MATH_PATTERN = _re.compile(r"^[0-9\.\s\+\-\*\/\(\)]+$")
MATH_KEYWORDS = _re.compile(r"\b(calculate|solve|what is|evaluate|sum|add|subtract|multiply|divide)\b", flags=_re.I)
def decide_tool(user_query: str) -> dict:
q = user_query.strip().lower()
if _re.search(r"\bhello\b|\bhi\b|\bhey\b|\bgood morning\b|\bgood evening\b", q):
return {"function_to_use": "chat", "reason": "Greeting detected by rule."}
if "weather" in q or "forecast" in q or "temperature" in q:
return {"function_to_use": "weather", "reason": "Weather-related keyword matched."}
if _re.search(r"\bthermostat\b|\bset thermostat\b|\bset temperature\b", q):
return {"function_to_use": "thermostat", "reason": "Thermostat control intent matched."}
if "india" in q and _re.search(r"\btime\b|\bdate\b|\bnow\b|\bcurrent\b", q):
if "time" in q:
return {"function_to_use": "time", "reason": "Explicit 'time' + 'India' matched."}
if "date" in q:
return {"function_to_use": "date", "reason": "Explicit 'date' + 'India' matched."}
if MATH_PATTERN.match(user_query) or (_re.search(MATH_KEYWORDS, q) and any(ch.isdigit() for ch in q)):
return {"function_to_use": "math", "reason": "Math expression or math keywords detected."}
if _re.search(FACTUAL_KEYWORDS, q):
if "time" in q and "india" not in q:
return {"function_to_use": "time", "reason": "Time query detected; using deterministic time tool."}
return {"function_to_use": "search", "reason": "Factual query matched; using grounded search."}
system_instruction = """
You are a strict router assistant. Decide exactly one tool for this query and return only valid JSON with keys:
{"function_to_use": "<one of: chat, search, time, date, math, weather, thermostat, science>", "reason": "short explanation"}
Do not return anything else.
"""
try:
resp = client.models.generate_content(model=MODEL, config=types.GenerateContentConfig(system_instruction=system_instruction), contents=user_query)
text = getattr(resp, "text", "").strip()
parsed = json.loads(text)
if "function_to_use" in parsed:
return parsed
except Exception:
pass
return {"function_to_use": "chat", "reason": "Default fallback to chat."}
def teacher_polish(user_query: str, tool_name: str, tool_output) -> str:
system_instruction = (
"You are ICIS AI teacher. Produce a concise (1-3 sentence) explanation in teacher tone.\n"
"IF the tool_output contains numeric facts (dates, times, numbers), DO NOT change them; only rephrase and add a short real-life example.\n"
"If the tool_output is an action confirmation (like thermostat status), confirm the action succinctly.\n"
"Return only the final user-facing text."
)
content = f"User query: {user_query}\nTool used: {tool_name}\nTool output: {json.dumps(tool_output, ensure_ascii=False)}"
return generate_text(system_instruction=system_instruction, content=content)
def hub_handle(user_query: str):
decision = decide_tool(user_query)
tool_name = decision.get("function_to_use", "chat")
tool_output = None
if tool_name == "time":
loc = "India" if "india" in user_query.lower() else "UTC"
tool_output = time_tool(location=loc)
elif tool_name == "date":
loc = "India" if "india" in user_query.lower() else "UTC"
tool_output = date_tool(location=loc)
elif tool_name == "math":
expr = _re.sub(r"[^0-9\.\+\-\*\/\(\)\s]", "", user_query).strip() or user_query
tool_output = math_tool(expr)
elif tool_name == "weather":
m = _re.search(r"in ([A-Za-z\s]+)$", user_query, flags=_re.I)
loc = m.group(1).strip() if m else "London"
tool_output = weather_tool(loc)
elif tool_name == "thermostat":
m = _re.search(r"(\d+)", user_query)
temp = int(m.group(1)) if m else 20
tool_output = {"status": "success", "set_to": temp}
elif tool_name == "search":
tool_output_text = grounded_search(user_query)
tool_output = {"search_text": tool_output_text}
elif tool_name == "science":
system_inst = "You are an ICIS science teacher; explain succinctly in 2-3 sentences with a simple example."
expl = generate_text(system_inst, user_query)
tool_output = {"explanation": expl}
else:
system_inst = "You are a friendly ICIS AI teacher, reply casually and briefly."
reply = generate_text(system_inst, user_query)
tool_output = {"reply": reply}
final = teacher_polish(user_query=user_query, tool_name=tool_name, tool_output=tool_output)
return {
"user_query": user_query,
"decision": decision,
"tool_output": tool_output,
"final_response": final
}
# =====================================================
# Helpers: strip markdown -> plain text, concise
# =====================================================
def strip_markdown(md: Optional[str]) -> str:
if not md:
return ""
text = str(md)
# remove code fences
text = re.sub(r"```.*?```", "", text, flags=re.S)
# images 
text = re.sub(r"!\[.*?\]\(.*?\)", "", text)
# links [text](url) -> text
text = re.sub(r"\[([^\]]+)\]\([^\)]+\)", r"\1", text)
# inline codes `x`
text = re.sub(r"`([^`]*)`", r"\1", text)
# remove remaining markdown symbols like # * > -
text = re.sub(r"(^|\s)[#>*\-]+\s*", r"\1", text)
# collapse whitespace
text = re.sub(r"\s+\n", "\n", text)
text = re.sub(r"\n{2,}", "\n\n", text)
text = text.strip()
return text
def concise_text(plain: str, max_sentences: int = 2) -> str:
if not plain:
return ""
# naive sentence split
parts = re.split(r'(?<=[\.\?\!])\s+', plain.strip())
if len(parts) <= max_sentences:
return " ".join([p.strip() for p in parts]).strip()
return " ".join(p.strip() for p in parts[:max_sentences]).strip()
# =====================================================
# FastAPI app + endpoints
# =====================================================
app = FastAPI(title="ICIS Mini-Hub")
templates = Jinja2Templates(directory="templates")
@app.get("/", response_class=HTMLResponse)
async def index(request: Request):
return templates.TemplateResponse("index.html", {"request": request})
@app.post("/chat")
async def chat_endpoint(payload: dict):
q = payload.get("query") if isinstance(payload, dict) else None
if not q:
return JSONResponse(status_code=400, content={"error": "Missing 'query' in JSON payload."})
out = hub_handle(q)
# extract final_response and function used
final_md = out.get("final_response", "")
plain = strip_markdown(final_md)
concise = concise_text(plain, max_sentences=2)
function_used = out.get("decision", {}).get("function_to_use", "chat")
return JSONResponse(content={
"function_used": function_used,
"response": concise
})
@app.post("/analyze_image")
async def analyze_image(file: UploadFile = File(...), prompt: str = Form(...)):
# read and ensure it's an image
content_type = file.content_type or ""
if not content_type.startswith("image/"):
return JSONResponse(status_code=400, content={"error": "Uploaded file is not an image."})
image_bytes = await file.read()
try:
image = Image.open(io.BytesIO(image_bytes))
except Exception:
return JSONResponse(status_code=400, content={"error": "Could not open image."})
# call genai with image + prompt
try:
response = client.models.generate_content(model=MODEL, contents=[image, prompt])
text_md = getattr(response, "text", "")
except Exception as e:
return JSONResponse(status_code=500, content={"error": f"GenAI image analysis failed: {str(e)}"})
plain = strip_markdown(text_md)
concise = concise_text(plain, max_sentences=2)
return JSONResponse(content={"mode": "image", "response": concise})
@app.post("/summarize_pdf")
async def summarize_pdf(file: UploadFile = File(...), prompt: str = Form(...)):
ct = file.content_type or ""
if ct != "application/pdf":
return JSONResponse(status_code=400, content={"error": "Uploaded file is not a PDF."})
data = await file.read()
try:
part = types.Part.from_bytes(data=data, mime_type='application/pdf')
response = client.models.generate_content(model=MODEL, contents=[part, prompt])
text_md = getattr(response, "text", "")
except Exception as e:
return JSONResponse(status_code=500, content={"error": f"GenAI PDF summarization failed: {str(e)}"})
plain = strip_markdown(text_md)
concise = concise_text(plain, max_sentences=2)
return JSONResponse(content={"mode": "pdf", "response": concise})
|