Summary

Pepper is a personal AI assistant designed to proactively manage your email and delegate tasks. It targets users needing to stay on top of critical communications and offload background work, offering autonomous email summarization and task delegation to free up user time.

How It Works

Pepper connects to Gmail to summarize emails and identify critical updates, delegating tasks to a "swarm of workers" for background processing. It uses an "Episodic Context Store" for state and external services like Composio for orchestration and LLMs (OpenAI/Anthropic) for NLP.

Quick Start & Requirements

1. Clone: git clone --recurse-submodules https://github.com/agentica-org/pepper
2. Environment Setup:
   • Create/activate Python 3.12 Conda env: conda create -n pepper python=3.12 pip -y && conda activate pepper.
   • Install episodic-sdk: cd episodic-sdk && pip install -e .[semantic].
   • Install Pepper requirements: cd ../pepper && pip install -r requirements.txt.
   • Start Context Store: episodic serve --port 8000 (must remain running).
3. Configuration:
   • Copy env template: cp env_var.example.sh env_var.sh.
   • Set COMPOSIO_API_KEY (from composio.dev) and LLM API key (OPENAI_API_KEY or ANTHROPIC_API_KEY).
   • Optionally configure LLM provider (export LLM_PROVIDER="anthropic") and update model names in Python files.
   • Load env vars: source env_var.sh.
4. Gmail Auth: Run python -m pepper.services.email_service and authorize via the provided URL.
5. Launch: Execute python -m pepper.launch_pepper. UI at http://localhost:5050/pepper/ui.html.
   • Prerequisites: Python 3.12, Conda, Git, Gmail account, Composio & LLM API keys.
   • Resource Footprint: Requires running multiple services; setup is moderately complex.

Highlighted Details

• Proactive email summarization and critical update surfacing.
• Task delegation to a "swarm of workers" for background processing.
• Integration with external services like Composio and LLM providers.

Maintenance & Community

Developed by the Agentica Team (Berkeley Sky Computing Lab). No explicit community channels or roadmap links are provided.

Licensing & Compatibility

The README does not specify a software license, preventing assessment of commercial use or closed-source linking compatibility.

Limitations & Caveats

• Requires direct Gmail access, potentially raising privacy concerns.
• Setup is complex, involving multiple external dependencies and API key configurations.
• "Swarm of workers" concept is high-level; implementation details and robustness are not detailed.
• Lack of license information hinders adoption decisions.