William Peidong Gao

William Peidong Gao

Ph.D. candidate in Genetics at Case Western Reserve University. I work on bioinformatics, multi-omics, human islet biology, and diabetes.

WillGaoLab

MorningBrief

Published:

MorningBrief transforms GPT-5.6 and Codex into your personal Chief of Staff. By combining your knowledge, priorities, and operating principles, it generates consistent, decision-ready morning briefings tailored to how you think and work.

MorningBrief is a local application for running repeatable Codex-based briefing workflows and browsing saved reports through a read-only Streamlit interface. mb run executes your configured prompt, saves Codex’s exact final response as Markdown, and records the run in an append-only history; mb serve opens a local reader for successful and failed runs.

Version 2.1 expands the public sample into a complete, privacy-safe demo. You can validate the bundled workflow, generate a real briefing through the Codex CLI, and browse the result without providing private prompts, notes, memory, or configuration. Version 2.0 established the reliable application foundation: portable configuration, setup diagnostics, private state initialization, atomic writes, rollback on failed runs, and non-destructive same-day reruns.

Watch the video walkthrough

Install and Try the Demo

The repository includes fictional sample data, so you can explore the reader without Codex, private files, or local configuration. Python 3.9 or newer is required.

git clone https://github.com/WillGaoLab/MorningBrief.git
cd MorningBrief
python -m venv .venv
source .venv/bin/activate
pip install -e .
mb serve --demo

On Windows, activate the environment with .venv\Scripts\activate before installing. The command opens the read-only MorningBrief app with a pre-generated fictional report, run metadata, and scan statistics. Use Close MorningBrief in the sidebar when you are finished.

Run the Complete v2.1 Demo

With the Codex CLI installed and available on PATH, you can exercise the entire workflow using only the repository’s public fictional fixtures:

mb doctor --demo
mb run --demo
mb serve --demo

The full demo uses a bundled prompt, a 12-file Markdown vault, and seed operational memory. It ignores .env and writes mutable memory, generated reports, and history only under the Git-ignored MorningBriefDataDemo/runtime/ directory. After a successful demo run, mb serve --demo automatically selects the generated runtime history instead of the pre-generated sample.

Configure Your MorningBrief

To generate your own reports, you also need the Codex CLI, a prompt, and an input directory that Codex may access.

First, initialize private operational memory. This command will not overwrite an existing memory.md file:

mb init --state-dir /path/to/your/morningbrief-state

Then copy .env.example to .env and replace the placeholder paths:

MB_CODEX_REPO=/path/to/your/codex-workspace
MB_PROMPT_FILE=/path/to/your/prompt.md
MB_VAULT_DIR=/path/to/your/input-notes
MB_STATE_DIR=/path/to/your/morningbrief-state
MB_DATA_ROOT=/path/to/your/output-data
MB_CODEX_BINARY=codex

MB_AUTOMATION_MEMORY_DIR remains available as the legacy name for MB_STATE_DIR. MorningBrief no longer searches global Codex session directories, so MB_SESSIONS_ROOT is not needed.

Validate the setup before generating your first briefing:

mb doctor
mb run
mb serve

You can also provide a configuration file explicitly from any working directory:

mb --config /path/to/config.env doctor
mb --config /path/to/config.env run

Environment variables override file values. MorningBrief also supports a platform user-config location, making the installed CLI portable beyond the repository directory.

Commands

mb run           Generate and save a new briefing
mb run --demo    Generate from bundled fictional inputs
mb serve         Browse reports from the configured output directory
mb serve --demo  Browse sample or generated fictional demo runs
mb doctor        Validate configuration without generating
mb doctor --demo Validate the full fictional demo workflow
mb init          Create private operational memory safely

The reader displays reports, scan statistics, run status, metadata, and file locations. It remains read-only: it does not launch the workflow or modify report data.

Data and Reliability

Each run receives a unique ID. Reports and history.csv are written atomically, and multiple runs on the same day are preserved instead of replacing one another. Existing date-named reports and older history rows remain readable.

MorningBrief captures the exact final Codex response through codex exec --output-last-message; it does not discover or parse global session directories. Before generation, it validates the configuration and snapshots operational memory. If generation or extraction fails, it restores the pre-run memory and records the failed run for diagnosis.

The demo follows the same separation of source inputs, operational memory, and generated data as a private workflow. Its public fixtures remain unchanged; all locally generated demo state lives in the ignored runtime/ directory.

Privacy and Scope

MorningBrief is local-first. Private prompts, notes, operational memory, Codex sessions, generated reports, and local configuration stay outside the public repository. The included demo contains fictional public content only.

Users decide which local files the configured workflow may read and are responsible for reviewing generated reports before using, publishing, or redistributing them.

Usage Disclaimer

This tool is for informational and personal workflow purposes only. It does not provide legal, medical, financial, investment, tax, security, or other professional advice. Generated briefs may be incomplete, outdated, inaccurate, or inappropriate for a particular decision.

Users are responsible for complying with the terms of any model provider, data provider, API, website, document source, or other third-party service involved in their configured workflow.

Attribution and Affiliation

This is a WillGaoLab project created and maintained by William Peidong Gao.

William Peidong Gao
        |
    WillGaoLab
        |
Open-source Projects
        |
   MorningBrief

This project is not affiliated with, endorsed by, sponsored by, or officially associated with OpenAI, Codex, Streamlit, GitHub, any data provider, any model provider, or any other entity referenced here.

All trademarks, product names, company names, and service names are the property of their respective owners and are used solely for identification and descriptive purposes.

License

Original code and documentation are released under the MIT License © 2026 William Peidong Gao. The MIT grant covers this project’s own source and documentation only; it does not extend to third-party content, private documents, API outputs, model outputs, datasets, trademarks, logos, product names, or other materials referenced by a configured workflow or generated report.