Skip to content

OkeyAmy/lift

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Lift — Autonomous SDG Marketing Agent

Lift is an agentic AI system built with OpenClaw that detects viral conversations related to the UN Sustainable Development Goals (SDGs) and converts them into actionable campaigns distributed across messaging platforms.

Lift continuously monitors social signals, analyzes their relevance to global challenges, generates campaign content using open-source AI models through the FLock API, and delivers campaign proposals via a Telegram bot interface.

The system acts as a marketing intelligence agent for social impact, helping NGOs, creators, and communities amplify initiatives aligned with sustainable development.


Project Vision

The internet constantly produces viral conversations around social issues such as climate change, education access, economic development, and health awareness. However, organizations working on these issues often lack the tools to capitalize on these moments.

Lift bridges this gap by transforming trending conversations into structured campaigns that can be deployed immediately.

Instead of manually searching for opportunities, Lift autonomously:

  • Detects viral discussions
  • Classifies them according to SDGs
  • Generates campaign strategies
  • Distributes campaign content

This allows impact-driven organizations to react to global conversations in real time.


Key Features

Autonomous Trend Monitoring

Lift continuously scans social conversations to detect viral topics aligned with global development goals.

SDG Classification

Detected trends are categorized into relevant Sustainable Development Goals.

AI Campaign Generation

Campaign messaging is automatically generated using open-source AI models through the FLock API.

Multi-Channel Communication

Lift interacts with users through messaging platforms such as Telegram.

Human-in-the-Loop Approval

Campaign proposals are sent to Telegram for approval before distribution.

Impact Analytics

Lift evaluates engagement metrics and calculates campaign performance scores.


Technology Stack

Lift is built using the following technologies.

OpenClaw Framework

OpenClaw orchestrates Lift’s autonomous agents and modular skills.

FLock API

Lift uses the FLock API for open-source AI model inference, enabling transparent and decentralized AI generation for campaign content.

Telegram Bot Interface

Users interact with Lift through a Telegram bot that serves as the control interface.

Cloud Infrastructure

Lift runs on AWS EC2 for scalable deployment.

Open Source Models

All AI generation relies exclusively on open-source models.


System Architecture

Lift follows a modular agent orchestration architecture where a central orchestrator manages specialized agents.

The architecture consists of:

  • Trend detection
  • Campaign generation
  • Distribution orchestration
  • Impact analytics

All modules are coordinated by the main Lift agent.


CreatorLift Architecture

Based on the configuration and documentation files in the workspace (AGENTS.md, SOUL.md, HEARTBEAT.md, openclaw.json), the CreatorLift system operates using a structured orchestration model.

The Core Philosophy: Orchestration vs Execution

Lift follows a strict division of responsibilities.

Lift acts as the central orchestrator (the brain) while multiple specialized agents act as execution workers (the hands).

  • Lift performs decision making, planning, and coordination.
  • Sub-agents perform isolated tasks and report results back to Lift.

Sub-agents operate in separate sessions and write outputs to files.

The system limits itself to a maximum of four concurrent sub-agents to maintain stability and resource efficiency.


Agent Roster

1. Lift — The Orchestrator

Role

Evaluates social signals, aligns them with UN Sustainable Development Goals, and manages the full lifecycle of campaigns.

Operation

Lift does not run continuously performing every task. Instead it runs in periodic cycles known as the Heartbeat loop.

During each cycle Lift:

  • Reads reports from sub-agents
  • Checks queues
  • Triggers new tasks
  • Communicates with users through Telegram

Lift is the only component allowed to make decisions.


2. Profiler — The Onboarder

Role

Creates professional creator profiles.

Trigger

Activated when a new creator registers in queue/registrations.json.

Frequency

Checked every 30 minutes.

Actions

  • Browses public creator accounts
  • Collects engagement metrics
  • Determines content niche
  • Generates structured profile.json
  • Produces a professional media_kit.pdf

Supported platforms include Instagram, TikTok, and YouTube.


3. Scout — The Matchmaker

Role

Identifies brand partnership opportunities for creators.

Trigger

Executed every 6 hours for active creators.

Actions

  • Searches for brands aligned with the creator’s niche
  • Uses Brave Search and web intelligence
  • Evaluates brands based on alignment with SDGs
  • Produces ranked opportunities

Output file

brand_opportunities.json

Scout avoids contacting brands listed in the cool-down list.


4. Writer — The Pitcher

Role

Drafts personalized outreach emails.

Trigger

Activated when new brand opportunities are discovered.

Actions

  • Reads creator profile
  • Reads brand data
  • Visits brand websites
  • Finds a relevant company detail
  • Writes a 180-250 word pitch

Important rule

Writer never sends emails.

All drafts are stored in:

outreach_queue.json

Lift later processes and sends them.


5. DealMaker — The Negotiator

Role

Manages brand communication and deal negotiation.

Trigger

Runs every 2 hours when Lift checks email replies.

Actions

  • Reads brand responses

  • Categorizes replies

    • Interested
    • Offer
    • Rejected
  • Drafts counteroffers using AI

  • Uses creator pricing constraints

Important feature

Before confirming a deal, DealMaker can contact the creator using ElevenLabs voice synthesis to obtain final approval.


6. Tracker — The Analyst

Role

Measures campaign performance.

Trigger

Runs every 48 hours for active campaigns.

Actions

  • Reads public campaign URLs
  • Collects engagement data
  • Generates campaign reports
  • Updates creator media kits
  • Alerts operator through Telegram

Reports are generated as professional PDF documents.


The Heartbeat Loop

Lift runs according to a structured execution rhythm.

Every 30 minutes

Scan for new creator registrations.

Every 2 hours

Check inbox for brand replies.

Every 3 hours

Process email outreach queue.

Maximum emails per run: 15

Maximum per day: 50

Every 6 hours

Search for new brand partnership opportunities.

Every 48 hours

Generate campaign performance reports.

Daily @ 07:30 WAT

Send a morning briefing to the operator via Telegram.

Daily @ 23:30 WAT

Run the Self-Improving Check to learn from patterns and mistakes.


Memory Architecture

Lift uses two types of memory.

1. Factual Memory

Location:

workspace-main/MEMORY.md
memory/YYYY-MM-DD.md

Maintained exclusively by Lift.

Stores:

  • Active creator lists
  • Total campaign revenue
  • Brand cool-down list
  • Market rate benchmarks

2. Execution Memory

Location:

~/self-improving/

Stores operational improvements.

Examples:

  • Tone corrections
  • Campaign style preferences
  • Operational mistakes

Patterns used successfully three times become HOT rules that the system follows automatically.


Telegram Command Center

Lift uses a Telegram bot as its main interface.

Users can control the system through commands.

Start monitoring campaigns

/startcampaign

Generate campaign content

/campaign climate

Display campaign performance

/impactreport

AI Generation Using FLock API

Lift integrates the FLock API to perform AI inference using open-source models.

The FLock API is used for:

  • Campaign generation
  • Trend summarization
  • Email drafting
  • Negotiation messaging

Example generated outputs:

  • Tweet threads
  • Telegram announcements
  • WhatsApp broadcasts
  • Outreach emails

All AI inference runs through FLock’s decentralized infrastructure.


Impact Scoring

Lift evaluates campaign success using the following formula:

Impact Score = Reach × Engagement × Issue Relevance

This metric allows Lift to continuously improve campaign strategies.


Deployment

Lift can be deployed on cloud infrastructure.

Recommended environment:

AWS EC2

Deployment steps:

  1. Launch EC2 instance
  2. Install Docker and dependencies
  3. Configure OpenClaw agent
  4. Set environment variables
  5. Connect Telegram webhook
  6. Start agent services

Ethical Guidelines

Lift operates according to ethical AI principles.

The system ensures:

  • Transparency through open-source models
  • Human oversight in campaign approval
  • Avoidance of misinformation
  • Alignment with legitimate social causes

Future Roadmap

Planned improvements include:

  • Multilingual campaign generation
  • Real-time crisis detection
  • Creator-NGO collaboration networks
  • Automated campaign optimization
  • Expanded social platform integrations

Conclusion

Lift demonstrates how agentic AI systems can transform social conversations into coordinated campaigns that accelerate progress toward global development goals.

By combining:

  • OpenClaw autonomous agents
  • FLock open-source AI inference
  • Messaging platform integrations

Lift creates a scalable infrastructure for impact-driven marketing automation.

The long-term vision is to build a global AI network that helps communities respond to social challenges faster and more effectively.


About

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors