The flow begins by collecting job requirements through multiple text input fields that define the position criteria. These inputs include job title, seniority level, main activities, required skills, desired skills, education requirements, work model, and job challenges. All of these criteria feed into a comprehensive prompt template that instructs the AI model on how to evaluate candidates against the specific job requirements.

Resume intake happens through the Read File component, which handles PDF, DOCX, ZIP, and CSV formats. For batch processing, the Directory component can load multiple files at once. Job descriptions can arrive through Chat Input, Read File, or webhook triggers that receive JSON from external systems.

The system normalizes resume data using the Structured Output component to extract standardized fields like skills, years of experience, and education into a consistent data format. This component makes it easy to emit structured JSON fields including compatibility scores, rationale, and potential red flags for each candidate.

For similarity-based matching, the workflow generates embeddings using embedding models and stores resume data in a vector database. This enables retrieval of the most relevant candidates based on semantic similarity to job requirements. The common pattern uses a "Load Data" subflow for initial ingestion and a "Retriever" subflow for query-time similarity search.

The core evaluation happens through a Prompt Template that defines the scoring rubric, connected to a Language Model that evaluates each candidate against the job description. For processing multiple resumes, the system uses Batch Run or Loop components to iterate through the candidate pool efficiently.

Final results get processed through DataFrame Operations for sorting and filtering by compatibility scores. The system can output results as structured data files, send them to external systems via API calls, or display them through the chat interface.

The workflow can be extended significantly using other Langflow components. You can integrate quality controls with the Cleanlab Evaluator to flag uncertain AI assessments, route edge cases to human reviewers using Smart Router logic, or connect to external systems through API Request components for ATS integration. Custom Components allow you to implement company-specific scoring algorithms or compliance checks that reflect your organization's unique hiring criteria.