# InboxOps
InboxOps is a rule based Tech VA inbox automation tool that helps reduce inbox overload by safely assisting with email triage, lead tracking, and scheduling support.
It is designed to behave like a careful human assistant rather than an autonomous AI system.
## Overview
Many founders, solo operators, and small teams already delegate inbox work to virtual assistants. InboxOps demonstrates how repetitive and low risk inbox tasks can be handled by software while preserving human control over sensitive or ambiguous situations.
The system is intentionally conservative. It prioritizes transparency, predictability, and safety over automation depth.
InboxOps is built as a realistic internal tool, not a speculative AI assistant.
## What InboxOps Does
InboxOps processes incoming emails and performs the following actions:
- Classifies emails into admin, customer support, lead, ignore, or ambiguous categories
- Detects risk signals such as angry language or legal keywords and blocks automation
- Logs potential leads into a structured CSV file
- Suggests draft calendar actions for meeting requests without auto booking
- Records every decision and action in an audit log
All behavior is explicit, rule driven, and configurable via simple YAML and JSON files.
## What InboxOps Intentionally Does Not Do
InboxOps avoids tasks that introduce operational or legal risk.
It does not:
- Negotiate pricing or contracts
- Reply to legal or emotional emails
- Auto book meetings or make commitments
- Make autonomous business decisions
When uncertainty is detected, the system escalates to a human instead of guessing.
## Why This Project Exists
Inbox management is one of the first operational bottlenecks faced by growing teams. Missed leads, delayed replies, and context switching all reduce focus and productivity.
InboxOps exists to demonstrate:
- How Tech VA workflows can be automated responsibly
- How rule based systems can outperform black box automation in business contexts
- How to design software that fails safely and predictably
This project is suitable for portfolio review, interviews, and internal tooling demonstrations.
## Project Structure
inboxops/
├── configs/ # Classification rules and reply templates
├── sample_data/ # Sample inbox emails for demo runs
├── src/ # Core inbox processing logic
├── tests/ # Automated test suite
├── README.md
├── requirements.txt
└── TEST_RESULTS.md
## How Classification Works
InboxOps uses explicit keyword rules defined in configuration files to classify emails.
Categories include:
- admin
- customer_support
- lead
- ignore
- ambiguous
Guardrails are applied before any action is taken. If an email contains legal terms, aggressive language, or unclear intent, automation is blocked and the message is escalated for human review.
This approach favors reliability and explainability over probabilistic interpretation.
## Running the Demo Locally
### Requirements
- Python 3.9 or newer
### Setup
pip install -r requirements.txt
\###Run the demo
python -m src.cli demo --in sample\_data/inbox\_emails.json --out out
The out/ directory will contain:
\- classification decisions
\- lead logs in CSV format
\- calendar draft suggestions
\- audit logs explaining each action
\###Testing
InboxOps includes automated tests to verify classification logic and guardrail behavior.
Run tests with:
pytest
Test results and validation notes are documented in TEST\_RESULTS.md.
\###Design Principles
InboxOps follows a few strict design principles:
* Rule based before AI
* Human in the loop by default
* Transparency over cleverness
* Safe failure over silent mistakes
These principles mirror how real internal operations tools are built and reviewed.
Intended Use Cases
InboxOps is suitable for:
* Tech VA workflow demonstrations
* Internal operations tooling
* Inbox triage experiments
* Portfolio and interview evaluation
It is not intended for unsupervised production use without additional review and safeguards.
\###License
This project is released under the MIT License.