"""
Dingo ATS Resume Optimizer Example
This example demonstrates how to use LLMResumeOptimizer for ATS-focused resume optimization.
Two optimization modes:
1. General Mode: Polish resume with STAR method (no context provided)
2. Targeted Mode: Inject missing keywords from KeywordMatcher report (context provided)
Input Requirements:
- input_data.content: Resume text (required)
- input_data.prompt: Target position (optional)
- input_data.context: KeywordMatcher match report JSON (optional, triggers Targeted Mode)
Output:
- Optimized resume sections with before/after comparison
- Keywords added/de-emphasized summary
- Section-by-section changes
"""
import os
from dingo.config.input_args import EvaluatorLLMArgs
from dingo.io.input import Data
from dingo.model.llm.llm_resume_optimizer import LLMResumeOptimizer
# Configure LLM (从环境变量读取)
LLMResumeOptimizer.dynamic_config = EvaluatorLLMArgs(
key=os.getenv("OPENAI_API_KEY", ""),
api_url=os.getenv("OPENAI_BASE_URL", "https://api.deepseek.com"),
model=os.getenv("OPENAI_MODEL", "deepseek-chat"),
)
def example_1_general_polish():
"""Example 1: General resume polish using STAR method"""
print("=== Example 1: General Polish Mode ===")
resume = """
张三
Python开发工程师
工作经历：
2020.01-至今 | ABC公司 | 开发工程师
- 做了很多开发工作
- 参与了系统优化
- 负责API开发
"""
data = Data(
data_id='optimize_test_1',
content=resume,
prompt='高级Python开发工程师'
# No context = General Mode
)
result = LLMResumeOptimizer.eval(data)
print(f"Error Status: {result.error_status}")
print(f"Reason:\n{result.reason[0]}")
# Access full optimization result
if hasattr(result, 'optimized_content'):
opt = result.optimized_content
print(f"\nSection Changes:")
for change in opt.get('section_changes', []):
print(f" - {change.get('section_name')}: {change.get('changes')}")
print()
def example_2_targeted_optimization():
"""Example 2: Targeted optimization with missing keywords injection"""
print("=== Example 2: Targeted Optimization Mode ===")
resume = """
张三
Python开发工程师 | 5年经验
专业技能：
Python, Django, MySQL, Git
工作经历：
2020.01-至今 | ABC公司 | 后端开发
- 负责公司后端系统开发
- 优化数据库性能
"""
# Simulating KeywordMatcher output (missing skills)
match_report = {
"match_details": {
"missing": [
{"skill": "Kubernetes", "importance": "Required"},
{"skill": "Docker", "importance": "Required"},
{"skill": "Redis", "importance": "Nice-to-have"}
],
"negative_warnings": []
}
}
import json
data = Data(
data_id='optimize_test_2',
content=resume,
prompt='高级Python开发工程师',
context=json.dumps(match_report) # Triggers Targeted Mode
)
result = LLMResumeOptimizer.eval(data)
print(f"Error Status: {result.error_status}")
print(f"Reason:\n{result.reason[0]}")
if hasattr(result, 'optimized_content'):
opt = result.optimized_content
summary = opt.get('optimization_summary', {})
print(f"\nKeywords Added: {summary.get('keywords_added', [])}")
print(f"Keywords Associative: {summary.get('keywords_associative', [])}")
print(f"Keywords Unused: {summary.get('keywords_unused', [])}")
print()
def example_3_full_pipeline():
"""Example 3: Full pipeline - Match then Optimize"""
print("=== Example 3: Full Pipeline (Match -> Optimize) ===")
from dingo.model.llm.llm_keyword_matcher import LLMKeywordMatcher
# Configure KeywordMatcher
LLMKeywordMatcher.dynamic_config = LLMResumeOptimizer.dynamic_config
resume = """
王五
软件工程师
技能：Java, Spring Boot, MySQL
经验：3年后端开发
"""
jd = """
后端工程师
要求：Python（必需）、Docker（必需）、AWS（加分）
"""
# Step 1: Match
match_data = Data(data_id='pipeline_1', content=resume, prompt=jd)
match_result = LLMKeywordMatcher.eval(match_data)
print(f"Match Score: {getattr(match_result, 'score', 'N/A')}")
# Step 2: Optimize (use match result as context)
# In real usage, extract missing keywords from match_result
optimize_data = Data(
data_id='pipeline_2',
content=resume,
prompt='后端工程师',
context='{"match_details": {"missing": [{"skill": "Python", "importance": "Required"}]}}'
)
opt_result = LLMResumeOptimizer.eval(optimize_data)
print(f"Optimization:\n{opt_result.reason[0]}")
print()
if __name__ == "__main__":
print("📝 Dingo ATS Resume Optimizer Examples")
print("=" * 50)
print()
print("⚠️ Please set your API key before running!")
print()
example_1_general_polish()
# example_2_targeted_optimization()
# example_3_full_pipeline()
print("✅ Examples completed!")
