Add address/location to resume contact header for ATS checks

- Add a `location` field to the Contact model and render it on the header line, and document it in the LLM resume schema (Option A)
- Parse a City/State/Country line from the resume when present, else fall back to a configurable CONTACT_LOCATION (default: Hyderabad, Telangana, India - Open to relocate) (Option B)
- Preserve the base location through LLM tailoring so it survives across every provider; bump the parse-cache version so existing caches are rebuilt with the address

Co-authored-by: Cursor <cursoragent@cursor.com>

config.py

@@ -5,6 +5,14 @@ load_dotenv()
 
 NVIDIA_BASE_URL = "https://integrate.api.nvidia.com/v1"
 
+# ── Candidate contact ────────────────────────────────────────────────────────
+# Resume-header address line. ATS checkers (Jobalytics / Resume Worded) flag a
+# missing address; a "City, State, Country" line satisfies that check. Used as
+# the fallback when the uploaded resume has no parseable location.
+CONTACT_LOCATION = os.getenv(
+    "CONTACT_LOCATION", "Hyderabad, Telangana, India · Open to relocate"
+)
+
 # Primary key — GLM 5.1 (most reliable, used for resume parsing + customization)
 NVIDIA_API_KEY = os.getenv("NVIDIA_API_KEY")
 GLM_BASE_URL   = NVIDIA_BASE_URL

src/resume_customizer.py

@@ -1348,6 +1348,10 @@ class ResumeCustomizer:
             tailored.name = base_resume.name
         if not tailored.contact.email and not tailored.contact.phone:
             tailored.contact = base_resume.contact
+        # Contact is identity — never let the model drop the address line
+        # (keeps the ATS "address" check green across every provider).
+        if not tailored.contact.location:
+            tailored.contact.location = base_resume.contact.location
 
         # 2. Restore dropped roles. If the LLM returned fewer roles than the
         # base resume has, append the missing ones with original bullets.

src/resume_model.py

@@ -26,6 +26,7 @@ class Contact:
     email: str = ""
     linkedin: str = ""
     website: str = ""
+    location: str = ""   # City, State, Country — needed for ATS "address" checks
 
     def render_line(self) -> str:
         """Format as a single dot-separated line for the resume header."""
@@ -41,6 +42,8 @@ class Contact:
         if self.website:
             ws = self.website.replace("https://", "").replace("http://", "")
             parts.append(ws)
+        if self.location:
+            parts.append(self.location)
         return " · ".join(parts)
 
 
@@ -100,6 +103,7 @@ class Resume:
                 email=c.get("email", ""),
                 linkedin=c.get("linkedin", ""),
                 website=c.get("website", ""),
+                location=c.get("location", ""),
             ),
             summary=data.get("summary", ""),
             skills=[str(s) for s in (data.get("skills") or []) if str(s).strip()],
@@ -218,7 +222,8 @@ RESUME_JSON_SCHEMA_DESCRIPTION = """\
     "phone": "<phone>",
     "email": "<email>",
     "linkedin": "<linkedin URL>",
-    "website": "<optional personal site>"
+    "website": "<optional personal site>",
+    "location": "<city, state, country — keep the candidate's existing value verbatim>"
   },
   "summary": "<4-6 sentence summary. Opens with: 'Strong-fit candidate for [role] at [company]: ...' Weave 8+ JD keywords naturally.>",
   "roles": [

src/resume_parser_v2.py

@@ -21,6 +21,11 @@ from .resume_customizer import (
     _extract_candidate_name, _normalize_spaced_text, _read_docx_text,
 )
 
+try:
+    from config import CONTACT_LOCATION
+except Exception:  # pragma: no cover - config always present in app runtime
+    CONTACT_LOCATION = "Hyderabad, Telangana, India · Open to relocate"
+
 
 _DATE_PATTERN = re.compile(
     r"(?:Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Oct|Nov|Dec)\w*\s+\d{4}\s*[-–—to]+\s*"
@@ -70,14 +75,19 @@ def parse_resume_pdf_cached(pdf_path: str, cache_path: str = "data/resume/_parse
     if not os.path.exists(pdf_path):
         raise FileNotFoundError(pdf_path)
     stat = os.stat(pdf_path)
-    cache_key = f"{stat.st_mtime_ns}_{stat.st_size}"
+    # cache version bumped to v2 when the contact `location` field was added, so
+    # pre-existing caches (without an address) are rebuilt.
+    cache_key = f"v2_{stat.st_mtime_ns}_{stat.st_size}"
 
     if os.path.exists(cache_path):
         try:
             with open(cache_path, encoding="utf-8") as f:
                 cached = json.load(f)
             if cached.get("_cache_key") == cache_key:
-                return Resume.from_dict(cached["resume"])
+                resume = Resume.from_dict(cached["resume"])
+                if not resume.contact.location:
+                    resume.contact.location = CONTACT_LOCATION
+                return resume
         except Exception:
             pass  # Cache invalid, re-parse
 
@@ -93,6 +103,27 @@ def parse_resume_pdf_cached(pdf_path: str, cache_path: str = "data/resume/_parse
 # Section extractors
 # ─────────────────────────────────────────────────────────────────────────
 
+def _extract_location(text: str) -> str:
+    """Best-effort: find a 'City, State, Country' line near the top of the resume.
+
+    Conservative on purpose — only accepts a short, comma-bearing header line so
+    we don't mistake a sentence for an address. Returns "" when nothing matches,
+    and the caller falls back to the configured CONTACT_LOCATION.
+    """
+    for line in text.splitlines()[:10]:
+        s = line.strip().strip("|•·-").strip()
+        if not (3 <= len(s) <= 60) or "," not in s:
+            continue
+        if re.search(r"@|https?://|linkedin\.com|\d{6,}", s, re.I):
+            continue  # skip email / url / phone lines
+        # India-based or a generic "City, Region(, Country)" shape
+        if re.search(r"\bindia\b", s, re.I) or re.match(
+            r"^[A-Z][a-zA-Z.]+(?:\s[A-Z][a-zA-Z.]+)*,\s*[A-Z][a-zA-Z.]+", s
+        ):
+            return s
+    return ""
+
+
 def _extract_contact(text: str) -> Contact:
     email_m = re.search(r"[\w.+-]+@[\w-]+\.[a-zA-Z]{2,}", text)
     phone_m = re.search(r"[\+]?[0-9]{1,4}[\s.-]?[0-9]{4,5}[\s.-]?[0-9]{4,5}", text)
@@ -101,6 +132,9 @@ def _extract_contact(text: str) -> Contact:
         phone=phone_m.group() if phone_m else "",
         email=email_m.group() if email_m else "",
         linkedin=("https://" + linkedin_m.group()) if linkedin_m else "",
+        # Option B: use the resume's own location if present, else the
+        # configured fallback so the ATS "address" check always passes.
+        location=_extract_location(text) or CONTACT_LOCATION,
     )