File size: 11,762 Bytes
c891504 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 | """
Gemini AI Processor module for the Construction Intelligence Hub.
Handles multimodal document analysis using Google Gemini API.
Uses gemini-3-flash-preview model for best multimodal performance.
"""
import io
import os
import json
import time
import logging
from typing import Dict, Any, List, Optional
from pydantic import BaseModel, Field
logger = logging.getLogger(__name__)
# Lazy import - will be imported when needed
_client = None
class DocumentAnalysis(BaseModel):
"""Schema for Gemini's structured output."""
project_name: Optional[str] = Field(None, description="Project name, usually 'DSC Hotel' or 'Dubai Studio City'")
document_type: Optional[str] = Field(None, description="One of: RFI, WIR, MIR, NCR, Shop Drawing, IFC, Payment/IPA, General")
document_reference_number: Optional[str] = Field(None, description="Document reference number e.g. '621-DXB24-RFI-CVL-CRCC-MIM-0036'")
status: Optional[str] = Field(None, description="One of: Approved, Approved with Comments, Rejected, Resubmit, Pending/Info")
consultant_comments: Optional[str] = Field(None, description="Summary of MIMAR/Consultant comments, drawing changes, or discrepancies")
action_required: Optional[str] = Field(None, description="What the contractor needs to do next")
assigned_discipline: Optional[str] = Field(None, description="One of: Civil, MEP, Structural, Architectural")
priority: Optional[str] = Field("Normal", description="Priority level: High, Normal, Low")
key_dates: Optional[str] = Field(None, description="Any deadline or date mentioned")
summary: Optional[str] = Field(None, description="Brief 2-3 sentence summary of the entire communication")
# The system prompt for Gemini
SYSTEM_PROMPT = """You are a Senior Civil Engineer and Project Manager working on the DSC Hotel project
(G+8 Hotel) in Dubai Studio City for China Railway 18th Bureau Group (CRCC). The Consultant on this
project is MIMAR.
Your task is to analyze incoming project emails and their attachments, then extract structured data.
RULES:
1. Be precise with document reference numbers - copy them exactly as written.
2. For document_type, use ONLY one of: RFI, WIR, MIR, NCR, Shop Drawing, IFC, Payment/IPA, General
3. For status, use ONLY one of: Approved, Approved with Comments, Rejected, Resubmit, Pending/Info
4. For assigned_discipline, use ONLY one of: Civil, MEP, Structural, Architectural
5. For priority, assess based on urgency words and deadlines: High, Normal, Low
6. consultant_comments should be a concise summary (max 200 words) of the key technical points
7. action_required should be specific and actionable
8. If information is not available in the email/attachments, use null
CONTEXT:
- Project: DSC Hotel (Dubai Studio City), G+8 Hotel
- Contractor: CRCC (China Railway 18th Bureau Group)
- Consultant: MIMAR
- Common document types: RFIs, WIRs (Work Inspection Reports), NCRs (Non-Conformance Reports),
Shop Drawings, IFC (Issued for Construction), Payment Applications (IPA)
"""
# MANDATORY: Use gemini-3-flash-preview model
GEMINI_MODEL = "gemini-3-flash-preview"
def get_gemini_client():
"""Get or create the Gemini client."""
global _client
if _client is None:
from google import genai
api_key = os.getenv("GOOGLE_API_KEY")
if not api_key:
raise ValueError("GOOGLE_API_KEY not set. Please add it in Space Settings → Variables & Secrets.")
_client = genai.Client(api_key=api_key)
return _client
def process_email_with_gemini(
email_body: str,
subject: str = "",
sender: str = "",
attachments: List[Dict[str, Any]] = None,
) -> Dict[str, Any]:
"""
Process an email and its attachments using Gemini.
Args:
email_body: The email body text
subject: Email subject line
sender: Email sender
attachments: List of attachment dicts with 'file_data', 'mime_type', 'filename'
Returns:
Dict with extracted document analysis fields
"""
from google import genai
from google.genai import types
client = get_gemini_client()
model = GEMINI_MODEL
max_size_mb = int(os.getenv("MAX_ATTACHMENT_SIZE_MB", "50"))
# Build the content parts
content_parts = []
# Add the email context as text
email_context = f"""
--- EMAIL METADATA ---
Subject: {subject}
From: {sender}
--- EMAIL BODY ---
{email_body[:10000]}
--- END OF EMAIL ---
Please analyze this email and any attached documents below, then extract the structured project data.
"""
content_parts.append(types.Part.from_text(text=email_context))
# Process attachments
processed_attachments = []
if attachments:
for att in attachments:
if not att.get("file_data") or not att.get("can_process", False):
continue
file_data = att["file_data"]
file_size_mb = len(file_data) / (1024 * 1024)
mime_type = att.get("mime_type", "application/octet-stream")
filename = att.get("filename", "unknown")
# Skip files that are too large
if file_size_mb > max_size_mb:
logger.warning(
f"Skipping {filename} ({file_size_mb:.1f}MB) - exceeds {max_size_mb}MB limit"
)
processed_attachments.append({
"filename": filename,
"status": "skipped",
"reason": f"File too large ({file_size_mb:.1f}MB)"
})
continue
try:
# For files under 20MB, use inline bytes
if file_size_mb < 20:
part = types.Part.from_bytes(data=file_data, mime_type=mime_type)
content_parts.append(types.Part.from_text(
text=f"\n--- ATTACHMENT: {filename} ---"
))
content_parts.append(part)
else:
# For larger files, use the File API
uploaded_file = client.files.upload(
file=io.BytesIO(file_data),
config=types.UploadFileConfig(
mime_type=mime_type,
display_name=filename
)
)
# Wait for processing
_wait_for_file_active(client, uploaded_file)
part = types.Part.from_uri(
file_uri=uploaded_file.uri,
mime_type=uploaded_file.mime_type
)
content_parts.append(types.Part.from_text(
text=f"\n--- ATTACHMENT: {filename} ---"
))
content_parts.append(part)
processed_attachments.append({
"filename": filename,
"status": "processed",
})
logger.info(f"Added attachment to Gemini request: {filename}")
except Exception as e:
logger.error(f"Error processing attachment {filename}: {e}")
processed_attachments.append({
"filename": filename,
"status": "error",
"reason": str(e)
})
# Call Gemini API
try:
response = client.models.generate_content(
model=model,
contents=content_parts,
config=types.GenerateContentConfig(
system_instruction=SYSTEM_PROMPT,
response_mime_type="application/json",
response_schema=DocumentAnalysis,
temperature=0.1, # Low temperature for extraction accuracy
)
)
# Parse the response
if response.parsed:
result = response.parsed.model_dump()
else:
# Fallback to parsing raw text
result = json.loads(response.text)
result["_processing_metadata"] = {
"model_used": model,
"attachments_processed": processed_attachments,
"timestamp": time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime()),
}
logger.info(f"Successfully processed email: {subject}")
return result
except Exception as e:
logger.error(f"Gemini API error: {e}")
# Attempt fallback: process just the email body without attachments
if attachments:
logger.info("Retrying with email body only (no attachments)...")
return _process_body_only(client, model, email_context)
raise
def _process_body_only(client, model: str, email_context: str) -> Dict[str, Any]:
"""Fallback: Process just the email body without attachments."""
from google.genai import types
try:
response = client.models.generate_content(
model=model,
contents=[types.Part.from_text(text=email_context)],
config=types.GenerateContentConfig(
system_instruction=SYSTEM_PROMPT,
response_mime_type="application/json",
response_schema=DocumentAnalysis,
temperature=0.1,
)
)
if response.parsed:
result = response.parsed.model_dump()
else:
result = json.loads(response.text)
result["_processing_metadata"] = {
"model_used": model,
"fallback": True,
"reason": "Attachment processing failed, body-only analysis",
"timestamp": time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime()),
}
return result
except Exception as e:
logger.error(f"Fallback processing also failed: {e}")
# Return minimal result
return {
"project_name": "DSC Hotel",
"document_type": "General",
"document_reference_number": None,
"status": "Pending/Info",
"consultant_comments": f"AI processing failed: {str(e)}",
"action_required": "Manual review required",
"assigned_discipline": None,
"priority": "Normal",
"key_dates": None,
"summary": "Automated processing failed. Please review email manually.",
"_processing_metadata": {
"error": str(e),
"fallback": True,
}
}
def _wait_for_file_active(client, file_obj, timeout: int = 120):
"""Wait for a Gemini file upload to become active."""
from google.genai import types
start = time.time()
while file_obj.state == types.FileState.PROCESSING:
if time.time() - start > timeout:
raise TimeoutError(f"File processing timed out after {timeout}s")
time.sleep(3)
file_obj = client.files.get(name=file_obj.name)
if file_obj.state == types.FileState.FAILED:
raise RuntimeError(f"File processing failed: {file_obj.name}")
def test_gemini_connection() -> tuple:
"""Test the Gemini API connection."""
try:
from google import genai
from google.genai import types
client = get_gemini_client()
model = GEMINI_MODEL
response = client.models.generate_content(
model=model,
contents=[types.Part.from_text(text="Reply with exactly: CONNECTION_OK")],
config=types.GenerateContentConfig(temperature=0.0)
)
if "CONNECTION_OK" in response.text:
return True, f"Gemini API connected successfully (model: {model})"
else:
return True, f"Gemini API responded (model: {model}): {response.text[:100]}"
except Exception as e:
return False, f"Gemini API error: {str(e)}"
|