GitResume Agent 🚀

Autonomous Git-Driven Resume & Portfolio Agent
A zero-touch agentic pipeline that perceives git diffs and codebase ASTs upon commit, synthesizes grounded Google XYZ accomplishment bullets with adversarial critique, and autonomously compiles role-tailored resumes directly to production portfolio endpoints.

💡 Why GitResume Agent?

As software and AI engineers, our codebases evolve every day — we ship features, refactor parsers, optimize vector retrieval latencies, and increase test coverage. Yet, our resumes remain static, outdated, and ungrounded.

GitResume Agent automates the entire lifecycle of engineering credentials as an event-driven CI/CD system:

1. Perception: Interrogates active Git repositories (commits, lines of code, test suites, and AST diffs) in < 5ms.
2. Schema Discovery: Dynamically detects tech stack additions and hackathon track changes from README.md files and package manifests, updating configuration in-place.
3. Reasoned Synthesis: Synthesizes quantified achievement bullets following the Google XYZ formula ("Accomplished X, measured by Y, by doing Z").
4. Adversarial Critique: Validates every cited metric and library against the commit AST to mathematically eliminate hallucinations.
5. Multi-Format Execution & Sync: Injects fresh bullets into Word (.docx) runs preserving exact styling, exports headless PDFs via COM automation, and syncs directly to your live portfolio website.

🤖 Truly Autonomous: How It Works

GitResume Agent operates on an event-driven, zero-touch autonomy loop:

                       ┌─────────────────────────────────────┐
                       │           DEVELOPER ACTION          │
                       │          `git commit -m "..."`      │
                       └──────────────────┬──────────────────┘
                                          │
                                          ▼ (Post-Commit Hook Trigger)
 ┌──────────────────────────────────────────────────────────────────────────────────┐
 │                          AUTONOMOUS AGENT PIPELINE                               │
 │                                                                                  │
 │  1. PERCEPTION           2. SCHEMA DISCOVERY      3. REASONING & SYNTHESIS       │
 │  ┌──────────────────┐    ┌──────────────────┐    ┌───────────────────────────┐   │
 │  │ InspectorAgent   │───▶│ SchemaDiscoverer │───▶│ SynthesizerAgent          │   │
 │  │ Git Log / AST    │    │ README & Badges  │    │ Flagship Ollama Cloud LLM │   │
 │  └──────────────────┘    └──────────────────┘    └─────────────┬─────────────┘   │
 │                                                                │                 │
 │                                                                ▼                 │
 │  5. DUAL-TIER SYNC       4. COMPILATION           4. ADVERSARIAL CRITIQUE        │
 │  ┌──────────────────┐    ┌──────────────────┐    ┌───────────────────────────┐   │
 │  │ Portfolio / Web  │◀───│ Docx & PDF Engine│◀───│ GroundingVerifierAgent    │   │
 │  │ Public Mirroring │    │ Headless Word COM│    │ AST / Metric Verification │   │
 │  └──────────────────┘    └──────────────────┘    └───────────────────────────┘   │
 └──────────────────────────────────────────────────────────────────────────────────┘

• Zero-Touch Git Hooks: Once installed, you never need to manually trigger compilation. Committing to any project repository automatically initiates background resume updates.
• Scoped, Per-Repo Sync (opt-in): with output.scoped_sync: true, a commit in one tracked repo updates only that project's section of each resume — other projects' text, PDF exports, and pushes are left untouched. See Scoped Sync.
• In-Place Schema Auto-Sync: Adding a new framework to package.json or a hackathon badge to README.md automatically mutates gitresume.yaml idempotently.
• Deterministic Guardrails: Human governance controls initial role personas and target paths, while multi-agent execution handles perception, critique, and artifact compilation.

🏗️ Multi-Agent Architecture

Agent Roles:

• SchemaDiscoverer Agent: Parses project README.md files, badges, and manifests (package.json, pyproject.toml, requirements.txt). When you add a tool or win a hackathon track, it updates gitresume.yaml in-place.
• Inspector Agent: Interrogates git ls-files, git log, and test directories across 100K+ lines of code in < 5ms without traversal bottlenecks.
• Synthesizer Agent: Employs flagship coding LLMs (e.g. kimi-k2.7-code, deepseek-v4-pro, qwen2.5-coder) via Ollama Cloud or local Ollama to craft high-impact accomplishment bullets.
• Grounding Verifier Agent: Acts as an adversarial critic, verifying that cited frameworks and performance metrics are mathematically grounded in the git history.
• Persona Dispatcher & Compilers: Modifies specific paragraph runs in Word .docx documents preserving exact typography, triggers MS Word COM automation for pixel-perfect PDF export, and syncs directly to public portfolio directories.

📂 Where Generated Resumes Are Stored & Synced

GitResume Agent manages a dual-tier storage system defined in your gitresume.yaml:

1. Primary Master Directory (output.resume_dir)

• The primary folder where your master .docx templates and compiled .pdf files are stored.
• Keeps all role-specific persona documents organized (e.g. Forward_Deployed_Engineer.docx, GenAI_Engineer.docx, AI_Engineer.docx, Master_Resume.docx).

2. Public Web / Portfolio Sync (output.sync_paths)

• Mirrors compiled .docx and .pdf files directly to your live portfolio website (e.g. ./portfolio_site/public/resume), static site generator public folders, or CDN buckets.
• Every time you run git-resume sync (or make a git commit), your live portfolio immediately serves the freshest resumes.

⚡ Quickstart & Installation

1. Install via PyPI (Recommended)

pip install git-resume-agent

git clone https://github.com/Naman6019/git-resume-agent.git
cd git-resume-agent

# Create virtual environment
python -m venv .venv
.venv\Scripts\activate      # Windows (or `source .venv/bin/activate` on Linux/macOS)

# Install editable package
pip install -e .

2. Configure Environment (Optional LLM Keys)

Copy .env.example to .env:

# Option A: Ollama Cloud (Uses flagship cloud models without local download)
OLLAMA_BASE_URL=https://ollama.com
OLLAMA_API_KEY=your_ollama_api_key

# Option B: Local Ollama (100% Free, Offline)
OLLAMA_BASE_URL=http://127.0.0.1:11434

# Option C: Google Gemini Free Endpoint
GEMINI_API_KEY=your_gemini_api_key

🕹️ CLI Usage & Core Workflows

1. Live Codebase Intelligence Dashboard

Displays real-time commit counts, file statistics, total lines of code, test suites, and deployment status across all tracked repositories:

git-resume scan

Output:

                     GitResume Live Codebase Intelligence                      
┌───────────┬───────────────────────────┬──────────────────────────────────┬─────────┬───────┬───────────┬─────────────┐
│ Repos     │ Tag / Track               │ Status & Links                   │ Commits │ Files │ Total LOC │ Test Suites │
├───────────┼───────────────────────────┼──────────────────────────────────┼─────────┼───────┼───────────┼─────────────┤
│ FundersAI │ OpenAI Build Week         │ ● Deployed (https://www.fund...  │     410 │   752 │   132,449 │         142 │
│           │                           │ https://github.com/...           │         │       │           │             │
│ TalentOS  │ All Things Agentic        │ ● Deployed (https://all-thin...  │      56 │   130 │    22,348 │          26 │
│           │                           │ https://github.com/...           │         │       │           │             │
│ CareFlow  │ Clinical Research         │ ○ In-Repo                        │       3 │    77 │    10,410 │           7 │
│           │                           │ https://github.com/...           │         │       │           │             │
│ GitResume │ Autonomous Multi-Agent... │ ○ In-Repo                        │      17 │    22 │     2,554 │           1 │
│           │                           │ https://github.com/...           │         │       │           │             │
└───────────┴───────────────────────────┴──────────────────────────────────┴─────────┴───────┴───────────┴─────────────┘

2. Auto-Discover Schema & Stack from READMEs

Inspects configured repositories, sanitizes code blocks, discovers git origin remotes, and syncs newly added tech tools or hackathon tags into gitresume.yaml:

git-resume auto-config

3. Full End-to-End Multi-Agent Sync (All Projects at Once)

Runs the entire multi-agent pipeline across all projects simultaneously (Inspect $\rightarrow$ Synthesize $\rightarrow$ Verify $\rightarrow$ Compile .docx with clickable hyperlinks $\rightarrow$ Export .pdf via Word COM $\rightarrow$ Sync to Portfolio):

git-resume sync

Pass --repo to scope this to a single project — only that repo's section of each resume is rewritten, and only the resumes that actually changed are exported and pushed (see Scoped Sync):

git-resume sync --repo FundersAI

4. Synthesize Grounded Achievement Bullets

Uses multi-agent reasoning over recent git commits to generate role-specific Google XYZ bullets:

# Generate bullet for Forward-Deployed Engineer role
git-resume generate --repo FundersAI --persona fde

# Generate bullet for Generative AI Engineer role
git-resume generate --repo TalentOS --persona genai

# Generate bullet for AI/ML Engineer role
git-resume generate --repo CareFlow --persona ai_engineer

5. Install Zero-Touch Git Post-Commit Hooks

Automatically wires up post-commit hooks across all repositories listed in gitresume.yaml:

git-resume install-hooks

6. Check & Propose Portfolio Description Updates

Checks tracked repos for README changes and opens a review PR against the portfolio site if a description looks stale (see Keeping Portfolio Project Descriptions in Sync):

git-resume sync-descriptions

📝 Keeping Portfolio Project Descriptions in Sync

Resumes aren't the only thing that goes stale. The prose describing each project on the portfolio site (src/content/portfolio.ts) drifts too, every time a README gains a new capability, drops a limitation, or changes its stack. git-resume sync-descriptions closes that loop, safely:

1. Watches README.md, not every commit. For each repository with a portfolio_slug set in gitresume.yaml, it checks whether README.md has a newer commit than the last time it looked (tracked in the local .gitresume_state.json, never committed).
2. First sighting of a repo is a baseline, not a proposal. The very first time a repository is checked there's nothing to diff against, so it just records the current README commit and moves on -- it never rewrites a description out of the blue on day one.
3. Diffs, then judges. On a real change, it feeds the README diff (plus the current site copy, for tone) to an LLM and asks a yes/no question first: is this change material enough to make the site copy stale? Typo fixes and badge shuffling get NO_CHANGE. A new capability, a changed boundary, or a stack change gets specific field-level rewrites.
4. Grounds every claim before touching anything. Any proposed sentence citing a number or stack item that doesn't actually appear in the README is dropped from the proposal. If everything gets dropped, nothing happens.
5. Edits portfolio.ts field-by-field, never a blind overwrite. The editor locates the exact object literal for the matching slug by brace-matching and replaces only the specific fields that changed, preserving formatting everywhere else.
6. npx tsc --noEmit is the final gate. The edited file is type-checked before anything is committed. A failure reverts the file on disk -- nothing broken ever reaches git.
7. Opens a PR, never pushes to main. A branch is cut from the portfolio's base_branch, the single file is committed and pushed, and a PR is opened via gh pr create with the README diff's context in the body. Netlify will build a deploy preview on it. Nothing reaches production without a human merging it.

Wire it into the same zero-touch loop as resume sync by adding portfolio_slug (matching the slug in portfolioProjects) to a repo entry and a portfolio: block pointing at the site:

repositories:
  - name: MyNewProject
    portfolio_slug: my-new-project
    # ...

portfolio:
  repo_path: C:/Users/yourusername/Desktop/portfolio_site
  content_file: src/content/portfolio.ts
  base_branch: main

git-resume install-hooks wires this into the same post-commit hook as resume sync, so a commit that changes a tracked repo's README triggers a check automatically -- no separate hook to install.

➕ How to Add a New Repository to GitResume

Adding a new project to your resume pipeline takes less than 30 seconds:

Step 1: Add the Repository to gitresume.yaml

Add an entry under the repositories: section:

repositories:
  - name: MyNewProject
    path: C:/Users/yourusername/Desktop/MyNewProject
    repo_url: https://github.com/yourusername/MyNewProject   # Or leave blank for auto-discovery
    live_url: https://mynewproject.com                      # Optional (if deployed)
    deployed: true                                          # Set to true if live, false if in-repo

Step 2: Auto-Discover Schema & Tools

Run auto-config to inspect the project's README.md, package.json, or pyproject.toml and populate tags and tools:

git-resume auto-config

Step 3: Install Zero-Touch Git Hooks

Ensure background auto-sync is triggered every time you make a commit in the new repo:

git-resume install-hooks

Step 4: Re-Sync Resumes

git-resume sync

📄 Managing Resume Length: 1-Page vs 2-Page Persona Strategy

When you have 4+ active repositories, fitting everything onto a single page without overcrowding is a common challenge. GitResume supports Persona-Driven Project Budgeting (Strategy A):

                                  MASTER PIPELINE
                                         │
                 ┌───────────────────────┴───────────────────────┐
                 ▼                                               ▼
     🎯 1-PAGE TARGETED RESUMES                      📑 2-PAGE MASTER RESUME
    (fde, genai, ai_engineer)                               (master)
 ┌───────────────────────────────┐               ┌───────────────────────────────┐
 │ • 2 Flagship Projects (Full)  │               │ • All 4+ Projects (Full)      │
 │   - FundersAI (3 bullets)     │               │   - FundersAI (4 bullets)     │
 │   - TalentOS (3 bullets)      │               │   - TalentOS (3 bullets)      │
 │ • Compact Secondary Projects  │               │   - CareFlow (3 bullets)      │
 │   - CareFlow & GitResume      │               │   - GitResume (3 bullets)     │
 │     (1-line in-repo links)    │               │                               │
 └───────────────────────────────┘               └───────────────────────────────┘

1. Targeted 1-Page Resumes (fde, genai, ai_engineer)

• Flagship Focus: Dedicate full multi-line accomplishment bullets to your top 2 relevant flagships (e.g. FundersAI and TalentOS).
• Secondary / In-Progress Aggregation: Mention non-flagship projects (e.g. CareFlow, GitResume) as a single compact, high-impact bullet with clickable in-repo links (Repo: github.com/...).
• Bullet Count: Maintain 3 bullets for primary work experience and 3 bullets per flagship project.

2. Comprehensive 2-Page Master Resume (master)

• Full Coverage: Dedicate full sections and 3–4 quantified bullets to every project in your portfolio (FundersAI, TalentOS, CareFlow, and GitResume).
• Deep Architecture: Include full infrastructure breakdowns, CI/CD pipeline counts, and test suite metrics.

🔗 Clickable Native Word & PDF Hyperlinks

GitResume Agent automatically embeds native Word OpenPackaging XML hyperlinks (w:hyperlink elements) into .docx documents:

• Deployed Projects: Subheaders feature clickable Live and Repo links (e.g. Live: www.fundersai.co.in | Repo: github.com/Naman6019/FundersAI).
• Non-Deployed Projects: Subheaders feature clickable In-Repo links (e.g. Repo: github.com/Naman6019/CareFlow-Intelligence).
• PDF Preservation: When Word COM exports the .pdf during git-resume sync, all hyperlinks remain interactive across web browsers, PDF viewers, and ATS portals.

⚙️ Configuration (gitresume.yaml)

version: "1.0"

developer:
  name: "Naman Manocha"
  email: "namanmanocha6019@gmail.com"
  github: "https://github.com/Naman6019"
  linkedin: "https://linkedin.com/in/naman-d-manocha"
  location: "Mumbai, India"

# Repositories to inspect & ground
repositories:
  - name: FundersAI
    path: C:/Users/naman/OneDrive/Desktop/FundersAI
    tag: OpenAI Build Week Submission
    repo_url: https://github.com/Naman6019/FundersAI
    live_url: https://www.fundersai.co.in
    deployed: true
    primary_stack:
      - Python
      - FastAPI
      - React
      - TypeScript
      - Qdrant
      - PostgreSQL
      - Next.js
      - LangGraph

  - name: TalentOS
    path: C:/Users/naman/OneDrive/Desktop/ALLThingsAgentic
    tag: All Things Agentic Hackathon (Taskmaster Track)
    repo_url: https://github.com/Naman6019/all-things-agentic
    live_url: https://all-things-agentic--allthingsagentic-505213.asia-southeast1.hosted.app
    deployed: true
    primary_stack:
      - Python
      - Google ADK
      - LangGraph
      - Next.js
      - Firestore
      - Google Cloud Run

  - name: CareFlow
    path: C:/Users/naman/OneDrive/Desktop/CareFlow Intelligence
    tag: Clinical Research Intelligence
    repo_url: https://github.com/Naman6019/CareFlow-Intelligence
    deployed: false
    primary_stack:
      - Python
      - FastAPI
      - Next.js
      - TypeScript
      - ChromaDB

  - name: GitResume
    path: C:/Users/naman/OneDrive/Desktop/resume_automation
    tag: Autonomous Multi-Agent Resume & Portfolio Intelligence Engine
    repo_url: https://github.com/Naman6019/git-resume-agent
    deployed: false
    primary_stack:
      - Python
      - Typer
      - Rich
      - Ollama
      - Pytest
      - python-docx

# Resume Personas & Target Documents
personas:
  - id: fde
    title: Agentic AI / Forward-Deployed Engineer (1 Page)
    resume_file: Naman_Manocha_Agentic_AI_FDE_Resume.docx
    emphasis:
      - evals
      - hybrid-search
      - strict-abstention
      - production-agents
      - forward-deployed

  - id: genai
    title: Generative AI Engineer (1 Page)
    resume_file: Naman_Manocha_GenAI_Engineer_Resume.docx
    emphasis:
      - rag
      - vector-embeddings
      - langgraph
      - llm-pipelines
      - tool-calling

  - id: ai_engineer
    title: AI / ML Engineer (1 Page)
    resume_file: Naman_Manocha_AI_Engineer_Resume.docx
    emphasis:
      - mlops
      - benchmarks
      - latency-optimization
      - fine-tuning
      - rag-systems

  - id: master_1page
    title: Master Resume (1 Page)
    resume_file: Naman_Manocha_Master_1Page_Resume.docx
    emphasis:
      - full-stack
      - ai-agents
      - rag
      - applied-ml
      - production-systems

  - id: master
    title: Master Comprehensive Resume (2 Page)
    resume_file: Naman_Manocha_Master_Resume.docx
    emphasis:
      - full-stack
      - distributed-systems
      - ai-agents
      - production-architecture

# Dual-Tier Storage & Compilation Targets
output:
  formats:
    - docx
    - pdf
  resume_dir: C:/Users/naman/OneDrive/Desktop/Personal/Resume
  sync_paths:
    - C:/Users/naman/OneDrive/Desktop/Personal/portfolio_site/public/resume
  # Opt-in: makes `install-hooks` write per-repo hooks so a commit in one repo only
  # updates/exports/pushes that project's section. See "Scoped Sync" below. Default: false.
  scoped_sync: false

# LLM Intelligence Engine
llm:
  provider: ollama
  model: kimi-k2.7-code
  fallback_model: gpt-4o-mini

🪝 Zero-Touch Git Hook Integration

How does the automation work on every commit? When you run git-resume install-hooks (or python scripts/install_hooks.py):

1. It reads your configured repository paths in gitresume.yaml.
2. Locates each repository's .git/hooks/ folder.
3. Installs an executable post-commit script that triggers git-resume sync in the background upon every commit (plus sync-descriptions, which checks that repo's README.md for portfolio-worthy changes).

git-resume install-hooks

Output:

🔧 Installing Git post-commit hooks...
  * ✓ Installed hook for [EnterpriseAI]: ./projects/EnterpriseAI
  * ✓ Installed hook for [AgentOS]: ./projects/AgentOS
✅ Successfully installed hooks across 2 repositories!

🎯 Scoped Sync: Update Only the Project That Changed

By default, a commit in any tracked repo runs a full sync: every project's section is re-inspected and rewritten in every persona resume, every PDF is re-exported via Word COM, and every synced file is re-pushed — even if only one repo actually changed. That's fine for a handful of projects, but it means a commit in a repo nobody's resume-reads about still pays for a full Word export and push cycle.

Set output.scoped_sync: true in gitresume.yaml and re-run install-hooks to make each repo's hook self-scoped instead:

output:
  scoped_sync: true

git-resume install-hooks   # regenerates hooks to pass --repo <name>

Each repo's post-commit hook now runs git-resume sync --repo "<RepoName>" instead of a bare sync. A commit in FundersAI:

1. Still inspects every repo's live git stats (cheap — no LLM calls, no Word) so personas that blend two projects into one sentence (e.g. genai/ai_engineer, which mention TalentOS and GitResume in the same paragraph) stay correct.
2. Rewrites only the paragraph(s) that mention FundersAI in each persona resume — TalentOS, CareFlow, and GitResume's text are left byte-for-byte as they were.
3. PDF-exports only the resumes that were actually rewritten (skipped entirely if a persona never mentions FundersAI), instead of every .docx in resume_dir.
4. Copies and pushes only those changed docx/pdf files to output.sync_paths, instead of the whole directory.

scoped_sync is off by default — existing setups keep today's full-sync-on-every-commit behavior until you opt in. You can also invoke it manually without touching the config flag at all:

git-resume sync --repo TalentOS

🧪 Testing

Run the automated unit test suite:

pytest tests/ -v

collected 8 items
tests/test_git_resume.py::test_load_config PASSED                         [ 12%]
tests/test_git_resume.py::test_inspector_agent PASSED                     [ 25%]
tests/test_git_resume.py::test_grounding_verifier PASSED                  [ 37%]
tests/test_git_resume.py::test_synthesizer_agent PASSED                   [ 50%]
tests/test_git_resume.py::test_schema_discoverer PASSED                   [ 62%]
tests/test_git_resume.py::test_find_config_path_and_git_utils PASSED      [ 75%]
tests/test_git_resume.py::test_master_personas_config_and_compiler PASSED [ 87%]
tests/test_git_resume.py::test_scoped_update_resume PASSED                [100%]
====================== 8 passed in 1.22s ======================

📄 License

Distributed under the MIT License. See LICENSE for more information.

👤 Author

Naman Manocha — @Naman6019