NopeRi: Autonomous Naukri Job Application Pipeline

Fork Notice: This repository is an enhanced production fork of Traverser25/NopeRi, transforming the raw Selenium-free API client into a fully automated, quota-aware job application pipeline with multi-dimensional search sweeps, AI evaluation, and two-tier prescreening form autofill.

---

What This Fork Adds

The upstream repository provides foundational low-level REST client capabilities (login, token generation, search, apply). This fork introduces complete production infrastructure for hands-off daily execution:

1. Multi-Dimensional Freshness Sweep:

   • Sweeps queries in priority order: Freshness (Age 3 -> 4 -> 5 -> 6 -> 7 days) x Target Titles x Cities x Experience Levels x Adaptive Pages.
   • Prioritizes the freshest job listings first, expanding to older listings only when session targets require it.

2. Search Variation State Machine (search_variation_state.json):

   • Persists state for every (title, city, experience, posting_age, page) permutation across runs.
   • Automatically skips already-queried slices on subsequent runs, eliminating duplicate requests and saving network quota.

3. Adaptive Pagination & Anti-Bot Protection:

   • Dynamically requests next pages only when a full 20-job page returns; stops early on short pages, all-duplicate pages (STOP_ON_ZERO_NEW=1), or pagination boundaries.
   • Includes randomized request jitter and micro-batch cooling pauses to prevent rate limits and IP blocks.

4. Pre-AI External ATS Filtering:

   • Filters out third-party redirect links (Workday, Greenhouse, Taleo) before AI scoring using jobTypeFlags and external_jobs.jsonl.
   • Preserves 70%+ of inference tokens and logs external jobs for manual review.

5. Two-Tier Questionnaire Engine:

   • Tier 1 (Deterministic Static Profile): application-profile.yaml handles standard questions (current CTC, expected CTC, notice period, relocation, YOE) deterministically with zero hallucinations.
   • Tier 2 (AI Fallback with Confidence Gating): Novel questions are evaluated by LLM. If confidence is below 0.85, the submission aborts safely and logs the question for review.

6. Prescreening Review Log (questionnaire_review.jsonl):

   • Captures skipped questions with confidence scores and context.
   • Enables human-in-the-loop review: add answers to questionnaire_answers.yaml and re-run.

7. Role-Agnostic Salary Floor Gating (config/ba_salary.py):

   • Enforces a configurable minimum salary threshold (MIN_SALARY_LPA, default 20 LPA) across all roles, preventing applications to low-comp listings.
   • Supports ALLOW_UNDISCLOSED_SALARY=1 to permit listings where compensation is confidential or unstated.

8. 24-Hour Rolling Quota Enforcement:

   • Automatically tracks applications against Naukri's official rolling 24-hour limit of 50 Easy Applies, stopping gracefully when reached.

9. Cursor Agent Skill (.cursor/skills/noperi/SKILL.md):

   • Built-in Cursor slash command (/noperi or "Run NopeRi") that triggers the application workflow, reports skipped questions in chat, and handles interactive auto-retries.

10. Autonomous Quota Scheduler (Configurable Time):

    • Scheduler daemon (scripts/run_scheduled_once.py) that waits until your chosen time then fires under caffeinate (default: 00:10 IST).
    • Pass --time HH:MM (24-hour IST) or set SCHEDULED_TIME_IST env var to pick any run time.
    • Laptop must remain powered on and awake; caffeinate -ims prevents sleep on AC power.

---

System Architecture

[ Trigger: ./run.sh or Cursor /noperi ]
             │
             ▼
   [ Session & Auth Engine ] ── Bearer Token & Dynamic nkparam
             │
             ├─► Freshness Sweep (Age: 3 -> 4 -> 5 -> 6 -> 7 days)
             ├─► State Resume (search_variation_state.json)
             └─► Adaptive Pagination (stop on short or duplicate page)
             │
             ▼
   [ Pre-AI Flag Filter ] ──► External Redirect? ──► Log to external_jobs.jsonl
             │ (Easy Apply Only)
             ▼
     [ Model Hub ] ◄── OpenRouter (google/gemini-2.5-flash-lite or similar)
   (Score 0-100; Threshold >= 70)
             │
             ▼
  [ Questionnaire Engine ]
       ├── Tier 1: Static YAML (application-profile.yaml)
       └── Tier 2: AI Fallback (Confidence < 0.85 skips & logs to questionnaire_review.jsonl)
             │
             ▼
   [ Naukri Apply API ] (Submit Profile Resume)
             │
             ▼
  [ applied_jobs.csv ] ── Logged and deduped

---

Quickstart

1. Clone the Fork

git clone https://github.com/rawshn97/NopeRi.git
cd NopeRi

2. Set Up Environment

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

3. Configure Credentials (.env)

Copy .env.example to .env:

cp .env.example .env

Fill in your details:

USERNAME=your_naukri_email@example.com
PASSWORD=your_naukri_password

# OpenRouter (Low cost / Free tier)
OPENROUTER_API_KEY=sk-or-v1-your-key-here
OPENAI_API_BASE=https://openrouter.ai/api/v1/chat/completions
OPENAI_MODEL=google/gemini-2.5-flash-lite

# Pipeline Controls
USE_ADVANCED_CONFIG=1
MIN_APPLY_SCORE=70
NAUKRI_DAILY_QUOTA=50

4. Configure Your Profile (application-profile.yaml)

Copy the example profile template:

cp application-profile.example.yaml application-profile.yaml

Customize with your details:

• Contact information and location preferences
• Current and expected compensation (LPA)
• Notice period
• Total and product years of experience
• Target role titles and domain tags

5. Verify Naukri Profile Resume

Make sure your master PDF resume is uploaded to your Naukri Profile. Easy Apply automatically submits your uploaded profile resume.

6. Run

chmod +x run.sh
./run.sh

Or if you are using Cursor, simply type in chat:

/noperi

---

Configuration Reference

All settings can be customized via .env or passed inline before ./run.sh:

Environment Variable	Default	Description
USE_ADVANCED_CONFIG	1	Enables the multi-dimensional search sweep and AI pipeline
MIN_APPLY_SCORE	70	Minimum AI fit score (0-100) required to submit an application
NAUKRI_DAILY_QUOTA	50	Maximum Easy Applies allowed in rolling 24-hour window
APPLY_TARGET	None	Cumulative CSV total to reach (e.g. APPLY_TARGET=100)
MIN_APPLY_COUNT	None	Minimum new applications required in this session
EXPAND_KEYWORDS	0	Set to 1 to expand search across 12+ specialized role titles
TARGET_KEYWORDS	Config	Comma-separated list of target job titles
CITY_ORDER	Config	Comma-separated list of cities to sweep
EXPERIENCE_LEVELS	4,2	Comma-separated experience levels to query
JOB_AGE_LEVELS	7	Comma-separated freshness levels in days (e.g. 3,4,5,6,7)
MAX_PAGES	2	Hard cap on pages per search variation
STOP_ON_ZERO_NEW	1	Stop paging when a page returns zero new jobs
RESET_SEARCH_VARIATIONS	0	Set to 1 to clear variation history and re-scan from scratch
MIN_SALARY_LPA	20	Minimum salary threshold in LPA across all roles
ALLOW_UNDISCLOSED_SALARY	0	Set to 1 to permit listings with undisclosed or confidential salary

---

Screening Form Workflow & Auto-Retry

When an employer form includes questions that cannot be answered with high confidence (confidence < 0.85):

1. The listing is safely skipped to avoid false claims.
2. The question and options are logged to questionnaire_review.jsonl.
3. Add the question keyword and your preferred answer to questionnaire_answers.yaml:

   by_question_contains:
     "visa sponsorship": "No"
     "willing to relocate": "Yes"

4. Re-run ./run.sh to automatically apply to previously skipped jobs.

---

Production Safety & Best Practices

• Home Broadband Recommended: Run on home broadband or local connections. Naukri aggressively fingerprints datacenter ASNs (AWS, GCP, Azure, GitHub Actions).
• Residential Proxy Support: For cloud deployment, set HTTP_PROXY and HTTPS_PROXY pointing to a residential proxy provider.
• 24-Hour Rolling Quota: Respect the 50 daily apply limit. Naukri caps applications at 50 per day across their platform.

---

License

MIT License. See LICENSE for details.