How Hiring Agent Extracts GitHub Usernames from Resume Profiles

Hiring Agent extracts GitHub usernames by scanning the profile.links array in structured resume JSON, parsing URLs with domain-specific regex in transform.py, and normalizing the results through validation logic in github.py.

The interviewstreet/hiring-agent repository processes candidate resumes to identify GitHub profiles for technical evaluation. Understanding how this open-source tool extracts GitHub usernames from resume profiles reveals a two-stage pipeline that combines flexible URL parsing with strict validation rules before storing data for downstream scoring.

The Extraction Pipeline

Detecting GitHub URLs in Resume Data

The extraction begins in transform.py, where the pipeline iterates over the profile.links field of each resume. For every URL encountered, the system calls extract_username_from_url(url, domain) at line 154, passing a domain mapping that recognizes "github.com": "GitHub". This function builds a domain-specific regular expression to isolate the path component representing the username.

Parsing and Validating Usernames

Once a potential username is extracted, the raw value flows to extract_github_username(github_url) in github.py (line 116). This helper performs three critical operations:

• Strips trailing slashes, query strings, and fragments (?…, #…)
• Validates the string against the regex pattern ^[A-z0-9_-]+$
• Returns the cleaned username or None if validation fails

Storing Results for Downstream Processing

After validation, the username is stored in the transformed candidate record under the key github_username. This field feeds directly into score.py and the GitHubProfile Pydantic model defined in models.py, enabling subsequent API calls to https://api.github.com/users/<username>.

Implementation Example

The following code demonstrates the complete extraction flow according to the hiring-agent source code:

# Resume structure with profile links

resume = {
    "profile": {
        "links": [
            "https://github.com/jane-doe",
            "https://linkedin.com/in/jane-doe"
        ]
    }
}

# Extraction logic from transform.py (line 154)

for url in resume["profile"]["links"]:
    domain = url.split("://")[1].split("/")[0]
    if domain == "github.com":
        username = extract_username_from_url(url, domain)
        if username:
            # Validation via github.py (line 116)

            username = extract_github_username(url)
            resume["github_username"] = username

Running this pipeline produces the validated username:

>>> resume["github_username"]
'jane-doe'

Summary

• URL Detection: transform.py scans profile.links and identifies GitHub URLs using domain matching at line 154
• Username Extraction: extract_username_from_url() parses the path component using regex patterns tailored to the domain
• Validation: extract_github_username() in github.py sanitizes input and enforces the ^[A-z0-9_-]+$ pattern at line 116
• Storage: Valid usernames are stored under github_username for API consumption by score.py and the GitHubProfile model

Frequently Asked Questions

What file handles the initial GitHub URL detection?

The transform.py module handles initial detection by iterating over profile.links and matching domains against a mapping that includes "github.com": "GitHub", specifically through the extract_username_from_url function at line 154.

How does Hiring Agent validate extracted GitHub usernames?

Validation occurs in github.py through the extract_github_username() function at line 116, which strips URL artifacts and validates against the regex ^[A-z0-9_-]+$, returning None for invalid inputs.

What happens if a resume contains an invalid GitHub URL?

When extract_github_username() cannot parse a valid username or validation fails, the function logs a warning and returns None, preventing corrupted data from entering the candidate record under github_username.

Where is the extracted username stored in the candidate record?

The validated username is stored in the transformed candidate object under the key github_username, which is later used by score.py to fetch repository data via the GitHub API and populate the GitHubProfile model defined in models.py.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}