Skip to content
BranchesTags

Repository files navigation

FollowFlow — Autonomous Commitment Network

Keep your promises. Let AI handle the follow-through.
The autonomous AI employee that remembers what you promised, monitors deadlines, verifies evidence across industry apps, and drives follow-through without human nagging.

AWS Strands Bedrock AWS Builder Next.js 15 FastAPI PostgreSQL License: MIT

🌐 Live Production Deployments

🏗️ System Architecture

FollowFlow Architecture

🎯 The Core Problem

People make commitments every day:

  • "I'll publish the revised API spec tomorrow."
  • "I'll send the SOC-2 compliance checklist by Friday."
  • "I'll complete the AWS infrastructure migration by Monday morning."

Most commitments disappear after the conversation. The bottleneck isn't making promises — it is remembering them, structuring milestones, tracking deadlines across disconnected tools, verifying completion with immutable proof, and chasing deliverables.

FollowFlow is the autonomous AI agent that detects promises, turns them into structured multi-team workflows, monitors progress in the background, verifies deliverables directly against industry tools (GitHub, Slack, Notion, Jira, AWS), and interrupts humans only when a decision is required.

🌟 What Makes FollowFlow Industrial-Grade

1. Unique Username Culture (@username)

Every collaborator has a unique, verifiable username handle stored in the PostgreSQL database:

  • Zero Dummy Data: No hardcoded mock users in UI components. Every persona (@rahulk, @sarahc, @alexr, @priyap, @mittai) is resolved natively through /api/users.
  • Dynamic Identity: Profiles, team memberships, SLA compliance, and cryptographic badges are bound to real DB records.
  • Persona Switcher: Switch active handles seamlessly across the app to test multi-user workflows.

2. AWS Strands Agents SDK (v1.55.1) Core

Built strictly on the AWS Strands Agents SDK:

  • Tool Calling: Registered @tool decorators for lookup_industry_integration, calculate_commitment_horizon, and verify_sla_compliance.
  • Synchronous-to-Async Bridge: Non-blocking execution wrapped in run_in_executor to seamlessly interface with FastAPI's asynchronous event loop.
  • Inline Strands AI Copilot UX: Rather than bouncing users to a separate screen, an inline toggle [✨ Strands AI Copilot: ON/OFF] sits right inside the commitment dialog, autocompleting fields in real-time.

3. Industry Apps Integration Matrix

Commitments connect directly to industry platforms for evidence collection and automated verification:

Industry App Provider Key Evidence Collected & Verified
GitHub github Commits, PR merge state, git tree hash, repo accessibility
Slack / Stack slack Message permalinks, emoji sign-offs, channel confirmation
Notion notion Page status property, edit timestamps, content hash
LinkedIn linkedin Post publication URL, author URN verification
Jira / Linear jira_linear Ticket status (Done/Closed), resolution timestamp
AWS Cloud aws CloudFormation stack state, S3 receipts, CloudTrail events
Google Docs google_docs Document revision ID, sharing status, comment approvals

4. Multi-Organization & Multi-Team Governance

Enterprise-grade multi-tenancy:

  • Hierarchical Scoping: Organizations (FollowFlow Labs, Acme Systems, Starlight AI) host scoped teams (Core Platform, AI Research & Agents, Infrastructure & Cloud SRE).
  • Configurable SLA Policies: Custom target fulfillment rates (e.g. 95%) and automated grace periods (24h–72h).
  • Interactive Management: Create and manage organizations and teams on the fly with live API persistence.

5. AWS Builder Center Badges & Cryptographic Trust Layer

Inspired by AWS Builder Center Badges:

  • Cryptographic Evidence Modals: Every earned badge and fulfilled commitment contains an immutable audit payload (SHA-256 evidence digest, issuer verification, validation ID FF-AWS-2026-XXXXX).
  • Earned Badges: Cloud Practitioner, Autonomous Agent Specialist, Generative AI Builder, Solutions Architect, and Zero-Failure Finisher.

6. Enterprise Production Suite

A full suite of production-ready pages designed with a clean, quiet Linear/Vercel aesthetic:

  • 404 (/not-found.tsx): Enterprise recovery page with incident status link.
  • 500 (/error.tsx): Error boundary capturing exceptions with unique trace IDs.
  • Terms of Service (/terms): SLA commitments, acceptable use, subscription plans.
  • Privacy Policy (/privacy): SOC-2 Type II posture, GDPR compliance, zero-retention token handling.
  • AI Transparency Disclaimer (/disclaimer): Ambiguity Stop Rule and human override guarantees.
  • Security Whitepaper (/security): AWS Bedrock Guardrails, TLS 1.3, encrypted vault.
  • Real-Time System Status (/status): Component uptime (99.99%), latency telemetry, 45-day history.
  • Enterprise Support Desk (/support): Dedicated Slack Connect, P1-P3 SLA escalation matrix.

🏗️ System Architecture

                                  USER / CLIENT
                                        │
                                        ▼
                         ┌─────────────────────────────┐
                         │   Next.js 15 Web App        │
                         │   (Standalone Container)    │
                         └──────────────┬──────────────┘
                                        │  HTTP / REST
                                        ▼
                         ┌─────────────────────────────┐
                         │   FastAPI Backend API       │
                         │   (Python 3.12 + Uvicorn)   │
                         └──────────────┬──────────────┘
                                        │
                 ┌──────────────────────┼──────────────────────┐
                 ▼                      ▼                      ▼
    ┌─────────────────────────┐  ┌──────────────┐  ┌─────────────────────────┐
    │  AWS Strands Agent      │  │  Evidence    │  │  Amazon Bedrock         │
    │  (v1.55.1 SDK + Tools)  │  │  Engine      │  │  AgentCore Runtime      │
    └────────────┬────────────┘  └──────┬───────┘  │  (/ping + /invocations) │
                 │                      │          └─────────────────────────┘
                 ▼                      ▼
    ┌─────────────────────────┐  ┌───────────────────────────────────────────┐
    │  Supabase PostgreSQL    │  │  Industry Connectors                      │
    │  - users (@username)    │  │  - GitHub API (Commits, Tree Hash)        │
    │  - organizations & teams│  │  - Slack / Notion / Jira / AWS / Docs     │
    │  - commitments (RLS)    │  │  - Mailpit / Amazon SES (SMTP Follow-up)  │
    └─────────────────────────┘  └───────────────────────────────────────────┘

🤖 The Autonomous Agent Execution Loop

1. Commitment Detected / Captured
       ↓
2. Strands Agent Architect Structures Workflow
       (Who [@username], What, When, Evidence Provider, Scope)
       ↓
3. Organization SLA Compliance Checked
       (Target fulfillment rate, grace period window)
       ↓
4. Autonomous Background Monitoring
       (Time-to-deadline triggers, risk tier calculation)
       ↓
5. Multi-Platform Evidence Inspection
       (GitHub SHA check, Notion block check, Jira ticket state)
       ↓
6. Contextual Follow-Up Dispatched
       (Non-nagging, actionable email via Amazon SES / Mailpit)
       ↓
7. Ambiguity Stop Rule (Human-in-the-Loop)
       (If conflicting evidence or low confidence: pause, recommend, await decision)
       ↓
8. Cryptographic Proof Generated
       (SHA-256 evidence digest + AWS Builder Center Badge update)
       ↓
9. Reliability Score & Org Metrics Updated

📡 Core API Reference

Method Endpoint Description
GET /api/commitments Query commitments filtered by username, organization_name, team_name
POST /api/commitments Create structured commitment bound to @username and industry provider
POST /api/commitments/ai/guide Consult AWS Strands Agent Architect with prompt
GET /api/organizations List all tenant organizations and SLA policies
POST /api/organizations Create tenant organization with default platform team
GET /api/teams Query teams with optional ?organization_id= filter
POST /api/teams Create new team with designated lead handle
GET /api/users Autocomplete user handles across the company
GET /api/users/{username} Fetch user profile, bio, title, and reliability score
GET /api/integrations Returns industry app integration catalog & capabilities
POST /api/integrations/connect Live-validate and connect user accounts for industry apps
GET /api/profile/{username} Returns AWS Builder Center profile with cryptographic badge verification
GET /ping Amazon Bedrock AgentCore health check endpoint
POST /invocations Amazon Bedrock AgentCore invocation contract

🚀 Getting Started (Local Development)

1. Prerequisites

  • Node.js 20+
  • Python 3.12+
  • Docker (for Mailpit email simulator)
  • Ollama with qwen2.5:1.5b (or Amazon Bedrock credentials)

2. Clone & Environment Setup

git clone https://github.com/mittai17/followflow.git
cd followflow
cp .env.example .env

3. Start Local Services

# Start Mailpit (Local SMTP simulator)
docker run -d --name mailpit -p 8025:8025 -p 1025:1025 axllent/mailpit:latest

# Ensure Ollama has the base model
ollama run qwen2.5:1.5b

4. Start Backend (Port 8000)

cd apps/agent
pip install -r requirements.txt
python3 -m uvicorn main:app --host 0.0.0.0 --port 8000 --reload

5. Start Frontend (Port 3000)

cd apps/web
npm install
npm run dev

Open http://localhost:3000 to explore FollowFlow.

☁️ Live Production Deployment on AWS

FollowFlow is hosted live on Amazon ECS Fargate behind an AWS Application Load Balancer with Amazon ECR and Amazon Bedrock:

Automated Deployment Script

Deploy updates to production with a single command:

./scripts/deploy_aws.sh

📄 License

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

About

FollowFlow — Autonomous AI Commitment Network powered by AWS Bedrock & Strands AgentCore

followflow-alb-2061937775.us-east-1.elb.amazonaws.com

Topics

autonomous-agentawsaws-bedrockfastapihackathonis-a-devnextjsstrands-agents-sdk

Resources

Readme
MIT license
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

Packages

Contributors

Languages