Eng-Musa commited on
Commit
e6f6cbd
·
1 Parent(s): ee25266

update to use gemini api to precess cv

Browse files
__pycache__/config.cpython-312.pyc ADDED
Binary file (706 Bytes). View file
 
__pycache__/main.cpython-312.pyc CHANGED
Binary files a/__pycache__/main.cpython-312.pyc and b/__pycache__/main.cpython-312.pyc differ
 
config.py ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """
2
+ config.py
3
+ =========
4
+ Global configuration for Job Processor (Gemini V2 pipeline).
5
+ Lives at processor root alongside .env.
6
+ """
7
+
8
+ import os
9
+ from pathlib import Path
10
+ from dotenv import load_dotenv
11
+
12
+ # Load .env file from root
13
+ env_path = Path(__file__).parent / ".env"
14
+ load_dotenv(dotenv_path=env_path)
15
+
16
+ # ---------------------------------------------------------------------------
17
+ # Google Gemini Settings
18
+ # ---------------------------------------------------------------------------
19
+ GEMINI_API_KEY = os.getenv("GEMINI_API_KEY", "")
20
+
21
+ # Default model: Gemini 3.5 Flash-Lite
22
+ GEMINI_MODEL = "gemini-3.5-flash-lite"
23
+
24
+ MAX_OUTPUT_TOKENS = 4096
25
+ TEMPERATURE = 0.1
main.py CHANGED
@@ -417,44 +417,82 @@ def batch_match_cv(payload: BatchMatchCVRequest):
417
 
418
 
419
  # ---------------------------------------------------------------------------
420
- # POST /v2/ping-llm — quick diagnostic endpoint to test LLM & latency
421
  # ---------------------------------------------------------------------------
422
 
423
 
424
  class PingLLMRequest(BaseModel):
425
- prompt: str = Field(default="Say hello in one short sentence.", description="Prompt to send to the LLM")
426
- max_tokens: Optional[int] = Field(default=64, description="Max tokens to generate")
427
 
428
 
429
  @app.post("/v2/ping-llm", response_model=APIResponse)
430
  def ping_llm(payload: PingLLMRequest = PingLLMRequest()):
431
- """Diagnostic endpoint to ping the local LLM and measure cold start / response time."""
432
  t_start = datetime.now(timezone.utc)
433
  try:
434
- resp = cv_processor_v2.extractor.llm.create_chat_completion(
435
- messages=[{"role": "user", "content": payload.prompt}],
436
- max_tokens=payload.max_tokens or 64,
437
- temperature=0.1,
438
- )
439
- reply = resp["choices"][0]["message"]["content"].strip()
440
  t_end = datetime.now(timezone.utc)
441
  duration_ms = int((t_end - t_start).total_seconds() * 1000)
442
 
443
- active_cfg = cv_processor_v2.extractor._active_cfg or {}
444
  return APIResponse(
445
- message="LLM ping successful",
446
  statusCode=200,
447
  payload={
448
- "model": active_cfg.get("label", "unknown"),
449
  "prompt": payload.prompt,
450
- "reply": reply,
451
  "duration_ms": duration_ms,
452
  "duration_seconds": round(duration_ms / 1000, 2),
 
453
  },
454
  )
455
  except Exception as exc:
456
  return APIResponse(
457
- message=f"LLM ping error: {exc}",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
458
  statusCode=500,
459
  )
460
 
 
417
 
418
 
419
  # ---------------------------------------------------------------------------
420
+ # POST /v2/ping-llm — quick diagnostic endpoint to test Gemini LLM & latency
421
  # ---------------------------------------------------------------------------
422
 
423
 
424
  class PingLLMRequest(BaseModel):
425
+ prompt: str = Field(default="Say hello in one short sentence.", description="Prompt to send to Gemini")
 
426
 
427
 
428
  @app.post("/v2/ping-llm", response_model=APIResponse)
429
  def ping_llm(payload: PingLLMRequest = PingLLMRequest()):
430
+ """Diagnostic endpoint to ping Gemini API directly and measure latency."""
431
  t_start = datetime.now(timezone.utc)
432
  try:
433
+ res = cv_processor_v2.gemini_extractor.ping(prompt=payload.prompt)
 
 
 
 
 
434
  t_end = datetime.now(timezone.utc)
435
  duration_ms = int((t_end - t_start).total_seconds() * 1000)
436
 
 
437
  return APIResponse(
438
+ message="Gemini ping successful",
439
  statusCode=200,
440
  payload={
441
+ "model": res.get("model", ""),
442
  "prompt": payload.prompt,
443
+ "reply": res.get("reply", ""),
444
  "duration_ms": duration_ms,
445
  "duration_seconds": round(duration_ms / 1000, 2),
446
+ "usageMetadata": res.get("usageMetadata", {}),
447
  },
448
  )
449
  except Exception as exc:
450
  return APIResponse(
451
+ message=f"Gemini ping error: {exc}",
452
+ statusCode=500,
453
+ )
454
+
455
+
456
+ # ---------------------------------------------------------------------------
457
+ # GET /v2/models — List models & inspect limits (context window, rate limits)
458
+ # ---------------------------------------------------------------------------
459
+
460
+
461
+ @app.get("/v2/models", response_model=APIResponse)
462
+ def list_models():
463
+ """Lists available Gemini models and their input/output token limits."""
464
+ try:
465
+ data = cv_processor_v2.gemini_extractor.list_models()
466
+ return APIResponse(
467
+ message="Models fetched successfully",
468
+ statusCode=200,
469
+ payload=data,
470
+ )
471
+ except Exception as exc:
472
+ return APIResponse(
473
+ message=f"Failed to fetch models: {exc}",
474
+ statusCode=500,
475
+ )
476
+
477
+
478
+ # ---------------------------------------------------------------------------
479
+ # GET /v2/models/{model_name} — Inspect specific model token limits
480
+ # ---------------------------------------------------------------------------
481
+
482
+
483
+ @app.get("/v2/models/{model_name:path}", response_model=APIResponse)
484
+ def get_model_details(model_name: str):
485
+ """Inspects a specific model's token limits and parameters."""
486
+ try:
487
+ data = cv_processor_v2.gemini_extractor.get_model(model_name)
488
+ return APIResponse(
489
+ message=f"Model {model_name} details fetched successfully",
490
+ statusCode=200,
491
+ payload=data,
492
+ )
493
+ except Exception as exc:
494
+ return APIResponse(
495
+ message=f"Failed to fetch model {model_name}: {exc}",
496
  statusCode=500,
497
  )
498
 
requirements.txt CHANGED
@@ -37,4 +37,5 @@ optimum[onnxruntime]
37
  onnxruntime
38
 
39
  # In-process CV embedding cache (no external service required)
40
- cachetools
 
 
37
  onnxruntime
38
 
39
  # In-process CV embedding cache (no external service required)
40
+ cachetools
41
+ python-dotenv
services/new/__pycache__/cv_processor_v2.cpython-312.pyc CHANGED
Binary files a/services/new/__pycache__/cv_processor_v2.cpython-312.pyc and b/services/new/__pycache__/cv_processor_v2.cpython-312.pyc differ
 
services/new/config.py DELETED
@@ -1,68 +0,0 @@
1
- """
2
- config.py
3
- =========
4
- Model configuration for the V2 LLM-powered CV extraction pipeline.
5
-
6
- The processor will attempt to use the model at index 0.
7
- If its download fails, it falls back to index 1.
8
- Put your preferred model first.
9
- """
10
-
11
- from pathlib import Path
12
-
13
- # ---------------------------------------------------------------------------
14
- # Model priority list
15
- # ---------------------------------------------------------------------------
16
- # Each entry:
17
- # repo_id — HuggingFace repo (must be ungated / publicly accessible)
18
- # filename — exact GGUF filename inside that repo
19
- # n_ctx — context window in tokens (input + output combined)
20
- # n_threads — CPU threads to use during inference (match your vCPU count)
21
- # max_tokens — max tokens the LLM may generate in one response
22
- # ---------------------------------------------------------------------------
23
-
24
- MODELS = [
25
- {
26
- "label": "Qwen2.5-1.5B-Instruct Q4_K_M",
27
- "repo_id": "Qwen/Qwen2.5-1.5B-Instruct-GGUF",
28
- "filename": "qwen2.5-1.5b-instruct-q4_k_m.gguf",
29
- "n_ctx": 8192,
30
- "n_threads": 2,
31
- "max_tokens": 2048,
32
- },
33
- {
34
- "label": "Phi-3-mini-4k Q4_K_M",
35
- "repo_id": "bartowski/Phi-3-mini-4k-instruct-GGUF",
36
- "filename": "Phi-3-mini-4k-instruct-Q4_K_M.gguf",
37
- "n_ctx": 4096,
38
- "n_threads": 2,
39
- "max_tokens": 2048,
40
- },
41
- {
42
- "label": "TinyLlama-1.1B Q4_K_M",
43
- "repo_id": "TheBloke/TinyLlama-1.1B-Chat-v1.0-GGUF",
44
- "filename": "tinyllama-1.1b-chat-v1.0.Q4_K_M.gguf",
45
- "n_ctx": 2048,
46
- "n_threads": 2,
47
- "max_tokens": 1024,
48
- },
49
- ]
50
-
51
- # ---------------------------------------------------------------------------
52
- # Cache directory — GGUF files are downloaded here once and reused
53
- # ---------------------------------------------------------------------------
54
- # On HuggingFace Spaces free tier the filesystem is ephemeral, so the model
55
- # re-downloads on cold start. Point this to a mounted persistent volume
56
- # (e.g. /data/model_cache) if you have one.
57
- # ---------------------------------------------------------------------------
58
-
59
- MODEL_CACHE_DIR = str(Path(__file__).parent.parent.parent / "model_cache")
60
-
61
- # ---------------------------------------------------------------------------
62
- # CV truncation limit
63
- # ---------------------------------------------------------------------------
64
- # Maximum characters of CV markdown sent to the LLM.
65
- # 12,000 chars easily fits 2-4 page CVs comfortably into an 8k context window.
66
- # ---------------------------------------------------------------------------
67
-
68
- MAX_CV_CHARS = 12000
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
services/new/cv_processor_v2.py CHANGED
@@ -1,4 +1,4 @@
1
- """Orchestrates V2 CV processing: file → markdown → LLM extraction → payload."""
2
 
3
  from __future__ import annotations
4
 
@@ -8,23 +8,22 @@ from pathlib import Path
8
  from typing import Any, Dict
9
 
10
  from services.new.cv_converter_v2 import CvConverterV2
11
- from services.new.cv_llm_extractor import CvLlmExtractor
12
 
13
  logger = logging.getLogger(__name__)
14
 
15
 
16
  class CvProcessorV2:
17
- """Single entry point for V2 processing. Instantiate once, reuse across requests."""
18
 
19
  def __init__(self) -> None:
20
  self.converter = CvConverterV2()
21
- self.extractor = CvLlmExtractor()
22
 
23
  def process(self, file_bytes: bytes, filename: str) -> Dict[str, Any]:
24
  """
25
- Convert file → markdown, then run LLM extraction.
26
- Returns a dict with all fields to be saved to CvProfile.
27
- Raises ValueError on conversion failure.
28
  """
29
  suffix = Path(filename).suffix.lower() if filename else ".pdf"
30
  tmp_path = None
@@ -39,8 +38,8 @@ class CvProcessorV2:
39
  if not conversion.success:
40
  raise ValueError(conversion.error or "Conversion failed")
41
 
42
- # Step 2 — markdown → structured fields via LLM
43
- extracted = self.extractor.extract(conversion.markdown)
44
 
45
  return {
46
  # ── LLM-extracted fields ──
@@ -59,6 +58,7 @@ class CvProcessorV2:
59
  "projects": extracted.get("projects", []),
60
  "awards": extracted.get("awards", []),
61
  "sections": extracted.get("sections", []),
 
62
  # ── Conversion metadata ──
63
  "file_type": conversion.file_type,
64
  "is_scanned": conversion.is_scanned,
 
1
+ """Orchestrates V2 CV processing: file → markdown → Gemini extraction → payload."""
2
 
3
  from __future__ import annotations
4
 
 
8
  from typing import Any, Dict
9
 
10
  from services.new.cv_converter_v2 import CvConverterV2
11
+ from services.new.google.gemini_extractor import GeminiCvExtractor
12
 
13
  logger = logging.getLogger(__name__)
14
 
15
 
16
  class CvProcessorV2:
17
+ """Direct Gemini CV processor without local fallbacks."""
18
 
19
  def __init__(self) -> None:
20
  self.converter = CvConverterV2()
21
+ self.gemini_extractor = GeminiCvExtractor()
22
 
23
  def process(self, file_bytes: bytes, filename: str) -> Dict[str, Any]:
24
  """
25
+ Convert file → markdown, then directly run Gemini extraction.
26
+ Raises ValueError on conversion failure or RuntimeError on API failure.
 
27
  """
28
  suffix = Path(filename).suffix.lower() if filename else ".pdf"
29
  tmp_path = None
 
38
  if not conversion.success:
39
  raise ValueError(conversion.error or "Conversion failed")
40
 
41
+ # Step 2 — direct Gemini extraction
42
+ extracted = self.gemini_extractor.extract(conversion.markdown)
43
 
44
  return {
45
  # ── LLM-extracted fields ──
 
58
  "projects": extracted.get("projects", []),
59
  "awards": extracted.get("awards", []),
60
  "sections": extracted.get("sections", []),
61
+ "usage_metadata": extracted.get("usage_metadata", {}),
62
  # ── Conversion metadata ──
63
  "file_type": conversion.file_type,
64
  "is_scanned": conversion.is_scanned,
services/new/google/__init__.py ADDED
@@ -0,0 +1 @@
 
 
1
+ # Google Gemini Client & Extractor
services/new/google/__pycache__/__init__.cpython-312.pyc ADDED
Binary file (145 Bytes). View file
 
services/new/google/__pycache__/config.cpython-312.pyc ADDED
Binary file (846 Bytes). View file
 
services/new/google/__pycache__/gemini_extractor.cpython-312.pyc ADDED
Binary file (11.6 kB). View file
 
services/new/google/gemini_extractor.py ADDED
@@ -0,0 +1,219 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """
2
+ gemini_extractor.py
3
+ ===================
4
+ CV field extraction using Google Gemini Interactions API.
5
+ Direct call with zero fallbacks.
6
+ """
7
+
8
+ from __future__ import annotations
9
+
10
+ import json
11
+ import logging
12
+ import re
13
+ from typing import Any, Dict, List, Optional
14
+ import httpx
15
+
16
+ from config import (
17
+ GEMINI_API_KEY,
18
+ GEMINI_MODEL,
19
+ MAX_OUTPUT_TOKENS,
20
+ TEMPERATURE,
21
+ )
22
+
23
+ logger = logging.getLogger(__name__)
24
+
25
+ _SYSTEM_INSTRUCTION = (
26
+ "You are an expert CV/Resume parser. "
27
+ "Extract all information from the provided CV markdown and return a single valid JSON object. "
28
+ "Extract ONLY what is explicitly present in the CV — never invent or assume details."
29
+ )
30
+
31
+ _USER_PROMPT = """\
32
+ Parse the CV below and return a JSON object with EXACTLY this structure.
33
+ Use "" for missing strings and [] for missing arrays. Do not omit fields.
34
+
35
+ {{
36
+ "cv_title": "candidate professional title or most recent job title",
37
+ "seniority": "one of: intern | entry | mid | senior | lead",
38
+ "years_experience": 0,
39
+ "category": "one of: engineering | design | finance | marketing | sales | operations | healthcare | legal | education | hr | other",
40
+ "summary": "professional summary text or empty string",
41
+ "contact": {{"email": "", "phone": "", "location": ""}},
42
+ "links": [{{"label": "GitHub | LinkedIn | Portfolio | Website | Twitter", "url": "full URL"}}],
43
+ "skills": {{
44
+ "items": ["every skill as listed in CV"],
45
+ "technical": ["technical/hard skills only"],
46
+ "soft": ["soft/interpersonal skills only"]
47
+ }},
48
+ "experience": [{{"role": "", "company": "", "period": "", "responsibilities": []}}],
49
+ "education": [{{"institution": "", "degree": "", "period": ""}}],
50
+ "projects": [{{"name": "", "description": "", "tech_stack": [], "url": ""}}],
51
+ "awards": [{{"title": "", "year": "", "issuer": ""}}],
52
+ "sections": [{{"title": "EXACT heading as written in CV", "type": "summary | experience | education | skills | projects | awards | unknown", "content": "full section text"}}]
53
+ }}
54
+
55
+ Rules:
56
+ - sections: preserve the EXACT title of each section header from the CV so the UI keeps original naming.
57
+ - seniority: intern=0-1yr, entry=1-2yr, mid=2-5yr, senior=5-8yr, lead=8+yr or people manager.
58
+ - years_experience: integer from earliest start year to current year.
59
+
60
+ CV MARKDOWN:
61
+ {cv_markdown}"""
62
+
63
+
64
+ class GeminiCvExtractor:
65
+ """Directly calls Google Gemini Interactions API without any fallback."""
66
+
67
+ def __init__(self, api_key: Optional[str] = None, model: Optional[str] = None):
68
+ self.api_key = api_key or GEMINI_API_KEY
69
+ self.model = model or GEMINI_MODEL
70
+
71
+ def _call_interactions(self, prompt: str, system_instruction: str = "") -> dict:
72
+ """Call /v1beta/interactions directly."""
73
+ if not self.api_key or self.api_key == "YOUR_GEMINI_API_KEY_HERE":
74
+ raise ValueError("GEMINI_API_KEY is not set in processor/.env")
75
+
76
+ url = "https://generativelanguage.googleapis.com/v1beta/interactions"
77
+ headers = {
78
+ "x-goog-api-key": self.api_key,
79
+ "Content-Type": "application/json",
80
+ }
81
+ body: Dict[str, Any] = {
82
+ "model": self.model,
83
+ "input": prompt,
84
+ }
85
+ if system_instruction:
86
+ body["system_instruction"] = system_instruction
87
+
88
+ with httpx.Client(timeout=60.0) as client:
89
+ resp = client.post(url, headers=headers, json=body)
90
+ if resp.status_code != 200:
91
+ raise RuntimeError(f"Gemini Interactions API error ({resp.status_code}): {resp.text}")
92
+
93
+ data = resp.json()
94
+ reply = ""
95
+ for step in data.get("steps", []):
96
+ for item in step.get("content", []):
97
+ if item.get("type") == "text":
98
+ reply += item.get("text", "")
99
+
100
+ return {
101
+ "text": reply.strip(),
102
+ "usage": data.get("usage", {}),
103
+ }
104
+
105
+ def ping(self, prompt: str = "Say hello in one short sentence.") -> Dict[str, Any]:
106
+ """Test API key and latency directly."""
107
+ res = self._call_interactions(prompt=prompt)
108
+ return {
109
+ "model": self.model,
110
+ "reply": res["text"],
111
+ "usageMetadata": res["usage"],
112
+ }
113
+
114
+ def list_models(self) -> Dict[str, Any]:
115
+ """Call GET /v1beta/models directly."""
116
+ if not self.api_key or self.api_key == "YOUR_GEMINI_API_KEY_HERE":
117
+ raise ValueError("GEMINI_API_KEY is not set in processor/.env")
118
+
119
+ url = f"https://generativelanguage.googleapis.com/v1beta/models?key={self.api_key}"
120
+ with httpx.Client(timeout=30.0) as client:
121
+ resp = client.get(url)
122
+ if resp.status_code != 200:
123
+ raise RuntimeError(f"Failed to list models ({resp.status_code}): {resp.text}")
124
+ return resp.json()
125
+
126
+ def get_model(self, model_id: str) -> Dict[str, Any]:
127
+ """Call GET /v1beta/models/{model_id} directly."""
128
+ if not self.api_key or self.api_key == "YOUR_GEMINI_API_KEY_HERE":
129
+ raise ValueError("GEMINI_API_KEY is not set in processor/.env")
130
+
131
+ clean_id = model_id.replace("models/", "")
132
+ url = f"https://generativelanguage.googleapis.com/v1beta/models/{clean_id}?key={self.api_key}"
133
+ with httpx.Client(timeout=30.0) as client:
134
+ resp = client.get(url)
135
+ if resp.status_code != 200:
136
+ raise RuntimeError(f"Failed to get model ({resp.status_code}): {resp.text}")
137
+ return resp.json()
138
+
139
+ def extract(self, markdown: str) -> Dict[str, Any]:
140
+ """Extract structured CV profile and dynamic sections directly."""
141
+ prompt_text = _USER_PROMPT.format(cv_markdown=markdown)
142
+ res = self._call_interactions(
143
+ prompt=prompt_text,
144
+ system_instruction=_SYSTEM_INSTRUCTION,
145
+ )
146
+
147
+ data = self._parse_json(res["text"])
148
+ data = self._normalize(data)
149
+ data["completeness_score"] = self._completeness(data)
150
+ data["markdown"] = markdown
151
+ data["usage_metadata"] = res.get("usage", {})
152
+ return data
153
+
154
+ @staticmethod
155
+ def _parse_json(raw: str) -> Dict[str, Any]:
156
+ try:
157
+ return json.loads(raw.strip())
158
+ except json.JSONDecodeError:
159
+ match = re.search(r"\{.*\}", raw, re.DOTALL)
160
+ if match:
161
+ return json.loads(match.group())
162
+ raise ValueError(f"Could not parse valid JSON from Gemini output: {raw[:300]}")
163
+
164
+ @staticmethod
165
+ def _normalize(data: Dict[str, Any]) -> Dict[str, Any]:
166
+ data.setdefault("cv_title", "")
167
+ data.setdefault("seniority", "entry")
168
+ data.setdefault("years_experience", 0)
169
+ data.setdefault("category", "other")
170
+ data.setdefault("summary", "")
171
+
172
+ try:
173
+ data["years_experience"] = int(data["years_experience"])
174
+ except (TypeError, ValueError):
175
+ data["years_experience"] = 0
176
+
177
+ contact = data.get("contact")
178
+ if not isinstance(contact, dict):
179
+ data["contact"] = {"email": "", "phone": "", "location": ""}
180
+ else:
181
+ contact.setdefault("email", "")
182
+ contact.setdefault("phone", "")
183
+ contact.setdefault("location", "")
184
+
185
+ for key in ("links", "experience", "education", "projects", "awards", "sections"):
186
+ if not isinstance(data.get(key), list):
187
+ data[key] = []
188
+
189
+ skills = data.get("skills")
190
+ if not isinstance(skills, dict):
191
+ data["skills"] = {"items": [], "technical": [], "soft": []}
192
+ else:
193
+ for sub in ("items", "technical", "soft"):
194
+ if not isinstance(skills.get(sub), list):
195
+ skills[sub] = []
196
+
197
+ return data
198
+
199
+ @staticmethod
200
+ def _completeness(data: Dict[str, Any]) -> int:
201
+ score = 0
202
+ if isinstance(data.get("summary"), str) and len(data["summary"]) > 50:
203
+ score += 20
204
+ if data.get("experience"):
205
+ score += 25
206
+ if len(data.get("skills", {}).get("items", [])) >= 3:
207
+ score += 15
208
+ if data.get("education"):
209
+ score += 15
210
+ if data.get("contact", {}).get("email"):
211
+ score += 10
212
+ if data.get("years_experience", 0) > 0:
213
+ score += 5
214
+ if data.get("links"):
215
+ score += 5
216
+ title = data.get("cv_title", "")
217
+ if title and title.strip().lower() not in ("", "unknown"):
218
+ score += 5
219
+ return min(100, score)