Skip to content

royisme/eve

BranchesTags

Repository files navigation

TypeScript Bun React SQLite

Eve & Wall-E

A Local-First AI Personal Agent Platform
Your Jarvis for the Modern World

Features • Architecture • Tech Stack • Current Focus • Roadmap

🎯 What is Eve?

Eve is an AI-native personal assistant platform that aggregates your digital signals, analyzes them with LLMs, and helps you take action—all while keeping your data local and private.

Think of it as building your own Jarvis: a modular, extensible system where AI capabilities are first-class citizens.

Component Role Description
Eve The Mind Backend intelligence - reasoning, memory, and orchestration
Wall-E The Body Chrome extension - eyes on the web, hands on the keyboard

✨ Features

Core Platform

  • 🧠 AI-Native Architecture: Built on a capability-based agent framework where every feature is an LLM-invokable tool
  • 🏠 Local-First: All data stored locally in SQLite - your information never leaves your machine
  • 🔌 Modular Capabilities: Add new domains (Jobs, Email, Calendar) as pluggable modules
  • 🌐 Multi-Provider LLM: Seamlessly switch between Anthropic, OpenAI, or Google models
  • ⚡ Real-time Sync: SSE-based streaming for live updates

Current Capability: Job Hunting

  • 📥 Intelligent Inbox: Auto-aggregates job alerts from LinkedIn, Indeed via Gmail
  • 🔍 Deep Analysis: LLM-powered fit scoring against your resume
  • 📝 Smart Tailoring: One-click resume customization for each application
  • 📊 Analytics Dashboard: Funnel visualization, skill gap analysis
  • 🎯 Pre-scoring: Quick compatibility check before full analysis

🏗 Architecture

┌─────────────────────────────────────────────────────────────────┐
│                    Wall-E (Chrome Extension)                    │
│                 React • Tailwind • Milkdown Editor              │
└───────────────────────────────┬─────────────────────────────────┘
                                │ HTTP / SSE
                                ▼
┌─────────────────────────────────────────────────────────────────┐
│                         Eve Backend                             │
│  ┌──────────────────────────────────────────────────────────┐  │
│  │                    Hono HTTP Server                      │  │
│  └────────────────────────┬─────────────────────────────────┘  │
│                           ▼                                     │
│  ┌──────────────────────────────────────────────────────────┐  │
│  │              Eve Agent (pi-agent-core)                   │  │
│  │                                                          │  │
│  │   ┌─────────┐  ┌─────────┐  ┌─────────┐  ┌───────────┐  │  │
│  │   │  Jobs   │  │ Resume  │  │  Email  │  │ Analytics │  │  │
│  │   │ 8 tools │  │ 6 tools │  │ 3 tools │  │ services  │  │  │
│  │   └─────────┘  └─────────┘  └─────────┘  └───────────┘  │  │
│  └──────────────────────────────────────────────────────────┘  │
│                           │                                     │
│  ┌────────────────────────▼─────────────────────────────────┐  │
│  │           Services: LLM • Firecrawl • Gmail              │  │
│  └──────────────────────────────────────────────────────────┘  │
│                           │                                     │
│  ┌────────────────────────▼─────────────────────────────────┐  │
│  │                 SQLite + Drizzle ORM                     │  │
│  └──────────────────────────────────────────────────────────┘  │
└─────────────────────────────────────────────────────────────────┘

Key Design Decisions

Decision Rationale
Capability-based Tools Each feature exposed as AgentTool - LLM decides when to invoke
TypeBox Validation Runtime type safety for all tool parameters
Local SQLite Zero-config, portable, privacy-preserving
Hono + Bun Blazing fast HTTP with modern runtime
SSE for Sync Real-time progress without WebSocket complexity

🛠 Tech Stack

Backend (Eve)

Layer Technology Why
Runtime Bun 3x faster than Node, native TypeScript
HTTP Hono Ultra-lightweight, edge-ready framework
Database SQLite + Drizzle Type-safe ORM, zero-config persistence
AI Runtime pi-agent-core Production-grade agent orchestration
LLM Anthropic Claude Best reasoning for complex analysis
Scraping Firecrawl Reliable web content extraction

Frontend (Wall-E)

Layer Technology Why
Framework React 18 Component-driven, huge ecosystem
Bundler Vite Instant HMR, optimized builds
Styling Tailwind CSS Utility-first, consistent design
Components Base UI / Radix Accessible primitives
Editor Milkdown Extensible Markdown WYSIWYG
Extension Chrome MV3 Modern extension architecture

🎯 Current Focus: Job Hunting Copilot

The first domain implementation - turning job hunting from a chore into a data-driven process.

Workflow

📧 Gmail Alert → 🔍 Parse & Store → 🌐 Enrich JD → 🤖 LLM Analysis → 📝 Tailor Resume → 📊 Track Progress

Implemented Tools (17 AgentTools)

Jobs Capability (8)

  • jobs_search - Query job database
  • jobs_list - List with filters
  • jobs_enrich - Scrape full JD
  • jobs_analyze - Batch LLM analysis
  • jobs_analyze_single - Single job deep analysis
  • jobs_prescore - Quick compatibility check
  • jobs_tailor - Generate tailored resume
  • jobs_get_tailored_versions - Version history

Resume Capability (6)

  • resume_list, resume_import, resume_get, resume_update, resume_delete, resume_set_default

Email Capability (3)

  • email_status, email_setup, email_sync

🗺 Roadmap

Phase Status Focus
1. Foundation ✅ Complete Core framework, capability system
2. Job Copilot 🔄 Current Full job hunting workflow
3. Auto-Apply 📋 Next Semi-automated applications
4. Multi-Domain 🔮 Future Calendar, Contacts, Finance

Next Up

  • PDF generation backend (Playwright)
  • Universal Application Protocol (UAP)
  • LinkedIn Easy Apply adapter
  • Voice interface integration

🚀 Quick Start

# Prerequisites: Bun v1.0+

# Install dependencies
bun install

# Start the server (dev)
bun run src/index.ts serve

# Or launch TUI dashboard (dev)
bun run src/index.ts

# Build production dist
bun run build

# Run compiled output
bun dist/index.js serve

Data Directory

Eve stores data in a user-scoped directory by default:

  • Default: ~/.config/eve/eve.db
  • Override: EVE_DATA_DIR=/custom/path or --data-dir=/custom/path

You should not need to interact with this directory directly.

Migrations

  • Drizzle migrations live in drizzle/
  • Build copies them to dist/drizzle/
  • App startup runs migrations automatically
  • When schema changes: npx drizzle-kit generate --name <tag> and commit drizzle/

Configuration

# Set up LLM provider
eve config:set services.llm.provider "anthropic"

# Connect Gmail accounts
eve config:set services.google.accounts '["your@gmail.com"]'

# Add Firecrawl for web scraping
eve config:set services.firecrawl.api_key "fc-..."

🚢 Release Flow

Versioning

We use semantic versioning for releases.

  • Patch: bug fixes, no API changes
  • Minor: new capabilities or non-breaking behavior
  • Major: breaking changes or data migrations with incompatible behavior

Recommended bump (manual):

  • Update version in package.json
  • Run npx drizzle-kit generate --name <tag> if schema changed
  • Commit changes and tag: git tag vX.Y.Z

CI Build

CI should run:

bun install
bun run build
bun dist/index.js --help

Release Artifacts

  • dist/ folder (compiled backend + dist/drizzle)
  • Homebrew formula updated to point at the new release artifact

Upgrade Behavior

  • Users upgrade via Homebrew
  • On first run after upgrade, Eve auto-applies new migrations
  • Existing data in ~/.config/eve is preserved

📖 Documentation

Document Description
ROADMAP.md Future plans and milestones
STATUS.md Current implementation status
TECH_SPEC.md Technical architecture details
AGENTS.md AI agent development guide

🤝 Philosophy

Local-first, AI-native, privacy-respecting.

Eve is built on the belief that personal AI assistants should:

  1. Keep your data local - No cloud dependency for core functionality
  2. Be transparent - You control what the AI sees and does
  3. Be extensible - Add new capabilities without touching core code
  4. Be practical - Solve real problems, not demo toys

📫 Connect

Built by a software engineer who believes AI should augment human capability, not replace human agency.

Currently exploring: AI agent architectures, local-first software, and the future of personal computing.

Eve & Wall-E — Building Jarvis, one capability at a time.

About

No description, website, or topics provided.

Resources

Readme

Contributing

Contributing
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •  

Languages