feat: add Claude Vision form filler + browser reliability improvements#1157
Open
Sangsiva wants to merge 4 commits intofeder-cr:mainfrom
Open
feat: add Claude Vision form filler + browser reliability improvements#1157Sangsiva wants to merge 4 commits intofeder-cr:mainfrom
Sangsiva wants to merge 4 commits intofeder-cr:mainfrom
Conversation
Adds two improvements to the AIHawk browser automation layer:
1. src/utils/visual_form_filler.py (new)
A Claude Vision-powered utility that fills any web-based job
application form using screenshots + Selenium. Useful for
non-standard forms that can't be handled by DOM inspection alone.
Flow:
- Takes a screenshot of the current page
- Sends it to Claude Sonnet with the candidate profile
- Claude identifies visible fields and values to fill
- Selenium fills each field using label/placeholder/aria strategies
- Handles Next/Submit navigation, cookie banners, and CAPTCHA pauses
- Pauses before final submit for human review
Profile data (name, contact, salary, notice period, education,
skills, work authorization) is read entirely from plain_text_resume.yaml
— nothing is hardcoded.
2. src/utils/chrome_utils.py (improved)
- Add user-agent rotation across 5 real Chrome UA strings
- Add undetected-chromedriver support with graceful fallback to
standard Selenium when the package is not installed
- Patch navigator.webdriver via CDP to reduce automation detection
- Suppress automation extension flags (excludeSwitches, useAutomationExtension)
- Add _clear_profile_locks() to kill orphaned Chrome processes and
remove stale SingletonLock files before init — fixes crashes on
restart after unclean shutdown
- Make persistent profile directory configurable via BROWSER_PROFILE_DIR
env var (defaults to ~/.aihawk_browser_profile)
- Remove --incognito flag (incompatible with persistent profiles)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds two new modules to the AIHawk library layer: ### src/libs/jd_resume_matcher/jd_matcher.py A Claude-based alternative to the existing OpenAI resume tailoring flow. Instead of regenerating sections from YAML, it works from pre-written HTML resume variants and selects + tailors the best one for each JD. Flow: 1. Haiku reads stripped text of all resume variants + JD, picks the best base and does a gap analysis (cheap — ~1k tokens) 2. Sonnet receives full HTML of the selected resume + stripped text of others as bullet reference, outputs a complete tailored HTML resume 3. HTML is converted to PDF via headless Chrome 4. review_resume_pdf() auto-checks formatting before saving Key design choices vs the existing approach: - 5x fewer input tokens (stripped HTML vs full HTML for selection step) - Multi-resume selection — picks best positioning from a pool of variants - Config-driven: all paths and candidate name in jd_matcher_config.yaml (gitignored), example at data_folder_example/jd_matcher_config.yaml - Exposes tailor_resume_for_jd() for programmatic use and run_jd_match() for interactive CLI use ### src/utils/resume_reviewer.py Scans a generated resume PDF for common formatting issues: - Text overflow / clipping - Blank pages - Font rendering problems Returns structured review notes and optionally triggers auto-fix. ### data_folder_example/jd_matcher_config.yaml Setup template for jd_matcher — paths, candidate name, and resume filename map. Copy to data_folder/jd_matcher_config.yaml to use. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
RESUME_DESCRIPTIONS was hardcoded in jd_matcher.py with role names specific to one user's career. Moved to jd_matcher_config.yaml under resume_descriptions so any user can define their own variant labels and descriptions. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Covers setup, config file structure, CLI and programmatic usage, cover letter generation, output structure, and tips. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
⏺ ## Summary
This PR adds three new capabilities to AIHawk's automation and resume generation layer.
1.
src/libs/jd_resume_matcher/(new module)A Claude-based alternative to the existing resume tailoring flow. Instead of
regenerating sections from YAML, it works from pre-written HTML resume variants
and selects + tailors the best one per job description.
Two-step approach:
and does a gap analysis (~1k tokens, very cheap)
match the JD — reordering bullets, pulling stronger ones from other variants,
adding missing keywords
5x fewer input tokens vs sending full HTML to the selection step.
All configuration (paths, candidate name, resume variants and their descriptions)
lives in
data_folder/jd_matcher_config.yaml(gitignored).Template at
data_folder_example/jd_matcher_config.yaml.Full setup and usage docs in
src/libs/jd_resume_matcher/README.md.Exposes two entry points:
run_jd_match(api_key)— interactive CLI, prompts for a job URLtailor_resume_for_jd(api_key, jd_text, company_name)— programmatic, no promptsgenerate_cover_letter_for_jd(...)— role-specific cover letter via Haiku2.
src/utils/visual_form_filler.py(new)Claude Vision-powered utility that fills any web-based job application form
using screenshots + Selenium — useful for non-standard forms that DOM inspection
can't handle.
Flow:
All candidate data (name, contact, salary, notice period, education, skills,
work authorization) is read from
plain_text_resume.yaml— nothing hardcoded.3.
src/utils/chrome_utils.py(improved)undetected-chromedriversupport with graceful fallback to standard Seleniumnavigator.webdriverpatching via CDP to reduce automation fingerprintingexcludeSwitches,useAutomationExtension)_clear_profile_locks()— kills orphaned Chrome processes and removes staleSingletonLockfiles before init, fixing crashes on restart after unclean shutdownBROWSER_PROFILE_DIRenv var--incognitoflag (incompatible with persistent profiles)Also included
src/utils/resume_reviewer.py— scans generated resume PDFs for formattingissues (overflow, blank pages, font rendering) and returns structured notes
data_folder_example/jd_matcher_config.yaml— setup template for the JD matcher