An autonomous, privacy-first personal AI operating system for unified signal ingestion, cognitive signal understanding, multi-agent intelligence routing, and automated life management.
Jarvis Personal AI OS is a multi-agent artificial intelligence framework designed to process ambient real-world signals—ranging from financial transactions, bank SMS notifications, and GPay receipts, to action items, notes, and informational alerts.
By combining deterministic fast-path metadata extraction with local LLM cognitive capabilities (qwen2.5:1.5b), Jarvis transforms raw multi-source streams into structured, domain-specific ledgers and daily executive summaries.
Jarvis pairs natively with the mobile collection background app:
- Android App Repository: 📲 pradeepgithubrepo/jarviscollector
- Description: Runs as a lightweight Android service intercepting financial SMS messages and transaction notifications on-device, securely transmitting JSON signal envelopes to Supabase storage buckets for automated ingestion.
flowchart LR
subgraph Ingestion ["📱 1. Ingestion"]
Mobile["Android Collector\n(jarviscollector)"] --> Bucket["Supabase Storage"]
Statements["Bank PDFs / CSVs"] --> Bucket
Bucket --> Consumer["Consumer Agent\n(Deduplication & Qualification)"]
end
subgraph Intelligence ["🧠 2. Cognitive Layer"]
Consumer --> SUA["Signal Understanding Agent\n(SUA)"]
SUA --> Router["Intelligence Router\n& Dispatcher"]
end
subgraph Agents ["🤖 3. Specialist Agents"]
Router --> Fin["💳 Financial Agent"]
Router --> Todo["📝 Todo Agent"]
Router --> FYI["ℹ️ FYI Agent"]
Router --> Brief["🗞️ Daily Briefing Agent"]
end
subgraph Ledgers ["💾 4. Persistence"]
Fin --> L1[("financial_transactions")]
Todo --> L2[("todos")]
FYI --> L3[("fyi_entries")]
Brief --> L4[("daily_briefings")]
end
Explore the detailed architecture, pipelines, and agent specifications under docs/architecture/:
| Document | Description | Key Focus Areas |
|---|---|---|
| 📥 01. Consumption Pipeline | Data ingestion, file hashing, deduplication & qualification. | jarviscollector, SHA-256 dedup, qualification scoring rules. |
| 🧠 02. SUA & Signal Classification | Signal Understanding & Analysis cognitive engine. | Metadata bypass, local LLM parsing, unified contracts, routing. |
| 🤖 03. Agent Ecosystem | Visual guide & breakdown of all 8 Jarvis AI agents. | Consumer, SUA, Financial, Financial Summary, Todo, FYI, Briefing, Lifecycle. |
| 📐 04. End-to-End System Architecture | Topology, database schemas & system lifecycle. | Database ER diagrams, replay framework, backfill processing. |
The repository contains extensive technical deep-dives, migration blueprints, validation reports, and SQL schemas built during system development:
🔍 Expand Comprehensive Technical Index (20+ Specifications & Reports)
- 🎨 Phase 1a Ingestion Design: DESIGN.md
- 🗄️ Phase 1a Database Schema: SCHEMA.md
- 📕 Phase 1b PDF Parsing Spec: PDF_PARSING.md
- 📊 Phase 1c Dedup Validation: DEDUP_VALIDATION_REPORT.md
- 🔍 Consumer Pipeline Forensic Report: CONSUMER_PIPELINE_FORENSIC_REPORT.md
- 📐 Qualification Pipeline Blueprint: QUALIFICATION_PIPELINE_BLUEPRINT.md
- 🎨 Qualification Layer V2 Design: QUALIFICATION_LAYER_V2_DESIGN.md
- 🗺️ Qualification Migration Plan: QUALIFICATION_MIGRATION_PLAN.md
- 🛠️ Qualification Schema Changes: QUALIFICATION_SCHEMA_CHANGES.sql
- 📐 Understood Signals Blueprint V1: UNDERSTOOD_SIGNALS_BLUEPRINT_V1.md
- 🧠 Understanding Classification RCA: UNDERSTANDING_CLASSIFICATION_RCA.md
- 📊 LLM Benchmark Report: LLM_BENCHMARK_REPORT.md
- 📜 Contract Schema Specification V1: CONTRACT_SCHEMA_V1.md
- 🔀 Dispatch Framework: DISPATCH_FRAMEWORK.md
- 🏗️ Routing Architecture: ROUTING_ARCHITECTURE.md
- 🚦 Routing Rules Spec: ROUTING_RULES.md
- 🔄 Replay Framework: REPLAY_FRAMEWORK.md
- 📝 Todo Agent Blueprint V1: TODO_AGENT_BLUEPRINT_V1.md
- 🎯 Todo Agent Alignment Review: TODO_AGENT_ALIGNMENT_REVIEW.md
- 📱 Android Integration Spec: ANDROID_INTEGRATION_SPEC.md
- 📊 Financial Summary Phase 1 Validation: FINANCIAL_SUMMARY_PHASE1_VALIDATION.md
- 📈 Pipeline Backfill Execution Report: PIPELINE_BACKFILL_EXECUTION_REPORT.md
jarvis/
├── README.md # GitHub Repository Frontage
├── docs/
│ ├── architecture/ # Core System Architecture Suite
│ │ ├── 01_CONSUMPTION_PIPELINE.md
│ │ ├── 02_SUA_AND_SIGNAL_CLASSIFICATION.md
│ │ ├── 03_AGENT_ECOSYSTEM.md
│ │ └── 04_END_TO_END_SYSTEM_ARCHITECTURE.md
│ ├── v2/ # Historical V2 Architecture & Phase Reports
│ └── v2.1/ # Forensic Pipeline Audits
├── src/
│ ├── agents/ # Specialist & Core AI Agents
│ │ ├── consumer/ # Ingestion, Parsers & Qualification
│ │ ├── sua/ # Signal Understanding & Contract Builder
│ │ ├── financial/ # Personal Finance & Transaction Specialist
│ │ ├── financial_summary/ # Expenditure Analytics & Rollups
│ │ ├── todo/ # Action Item & Reminder Specialist
│ │ ├── fyi/ # Contextual Knowledge & Info Alerts
│ │ ├── daily_briefing/ # Morning & Evening Digest Synthesis
│ │ └── lifecycle/ # System Supervisor & Backfill Orchestrator
│ └── intelligence/ # LLM Clients, Dispatchers & Routing Engine
├── config/ # Qualification Rules & Domain Rules
└── tests/ # End-to-End Pipeline & Agent Test Suites
- Python 3.10+
- uv (Fast Python package manager)
- Ollama (with
qwen2.5:1.5bmodel pulled for local offline execution) - Supabase PostgreSQL Database
# Clone repository
git clone https://github.com/pradeepgithubrepo/jarvis.git
cd jarvis
# Install dependencies using uv
uv sync
# Configure environment variables
cp .env.example .env