Projects | Antoine Weill--Duflos

Pointeuse: a native Odoo time tracker for desktop and Android

Sat, 13 Jun 2026 00:00:00 +0000

Pointeuse is a fast, native time tracker for Odoo. Pointeuse is French for a punch clock, which is exactly what it is: a small app that lives in your system tray and lets you start and stop a timer against any Odoo task without opening a browser.

Odoo is a powerful ERP, but its full interface can be overwhelming for someone who just wants to record what they worked on. Pointeuse is the opposite: the simplest possible surface in front of it. Your whole job is to say which task you are on, get a gentle reminder, and switch with a click; the time funnels into Odoo as proper timesheets behind the scenes, so the company keeps all of the ERP’s power. It is also a demonstration that integrating with Odoo is easy, on the desktop or on mobile, with no server-side module.

It does the things an Odoo timesheet workflow needs, but as a native app instead of a web tab:

Track time against Odoo tasks. Start or stop a timer on any project.task; the timesheets land in account.analytic.line where Odoo expects them.
Attendance integration. Check in and check out (hr.attendance) from the app or the tray. The timer auto-stops when you check out from anywhere else, including the Odoo web client or mobile.
Idle reminders. A configurable popup asks what you are working on, with quick-switch suggestions. On Android these become notifications with action buttons.
Offline-first. All Odoo data is cached in SQLite. Timesheets queue locally when you are offline and sync when the connection returns.
A task dashboard. A kanban board grouped by stage and project, a task detail panel, and a time log with day, week, and month views.

It talks to any Odoo 14+ instance over standard XML-RPC, with no server-side module to install. Credentials, or an API key for Odoo Online, are stored in the system keyring, never on disk.

Under the hood it is built with Tauri 2 and a Rust backend, with a deliberately zero-framework vanilla JS frontend. One codebase ships for Windows, Linux, macOS, and Android. Commits to main follow Conventional Commits, semantic-release cuts the versions, GitHub Actions builds all four platforms, and the desktop apps self-update from GitHub Releases.

Pointeuse is MIT licensed. The full story of why I built it and how it works is in the blog post Pointeuse: a native, offline-first Odoo time tracker.

cortexmd: a long-term memory and code-navigation brain for AI agents

Wed, 03 Jun 2026 00:00:00 +0000

cortexmd is a long-term memory and code-navigation brain for AI agents, exposed over the Model Context Protocol. It started as a private project on my homelab called obsidian-mcp, a server that let Claude read, search, and write notes in my Obsidian vault. I built it for myself, then cleaned it up to share.

It does two things.

The first is memory. Agents forget everything between sessions. cortexmd gives them somewhere to put what they learn: memories auto-categorised into kinds like observation, decision, insight, and plan, with a heat lifecycle where reading a memory warms it and inactivity cools it down. Recall is hybrid, fusing full-text and semantic search, boosted by temperature and links. At the start of a session the agent does a wakeup that surfaces the hottest, most relevant memories, so it picks up where it left off.

The second is code navigation. A Rust indexer walks a repo, parses it with tree-sitter, and builds a SQLite symbol database recording each symbol’s name, kind, signature, docstring, file range, and call graph. That index is exposed as cheap MCP tools: symbol search, file outline, callers and callees, change-impact, call-chain, dead code, import cycles, and copy-paste duplicates. The design goal is that an agent navigates code by querying the index, at roughly 60 tokens per result, instead of reading whole files. There is an opt-in shell hook that rewrites things like grep and cat on an indexed repo into the equivalent code-nav call.

The piece that made it shippable is the brain-vault model. cortexmd owns a separate brain vault that is the only thing it ever writes to. Your own vaults are attached as read-only sources, indexed for search and code-nav, never modified, with a default-deny allowlist so private subtrees stay out. Data flows one way, so there is no shared mutable file and no merge race.

  SOURCE_VAULTS[]  (read-only, opt-in, allowlisted)
  ┌───────────┐  ┌───────────┐  ┌───────────┐
  │  notes/   │  │  code/    │  │  docs/    │
  └─────┬─────┘  └─────┬─────┘  └─────┬─────┘
        │  index (one-way, read)      │
        └──────────────┼──────────────┘
                       ▼
              ┌──────────────────┐
              │     cortexmd     │   <- sole writer
              │   (MCP server)   │
              └────────┬─────────┘
                       │ writes
                       ▼
              ┌──────────────────┐
              │   BRAIN_VAULT    │   memories · journal · diaries
              │ (own dir, not    │   tasks · KG notes · code-repos.json
              │  your vault)     │
              └──────────────────┘

It runs in two modes: a local-stdio mode with no Docker, no auth, and no network, recommended for one person; and a self-hosted HTTP mode with auth for multi-client setups. The repo is a polyglot monorepo, a TypeScript MCP server and a single Rust binary, kept honest by a shared contract and a CI parity check.

cortexmd is pre-alpha and MIT licensed. APIs and config names are still in flux.

The full story is in a four-part blog series. Start with Giving an AI Agent a Second Brain.

OpenNeato Home Assistant Integration

Sun, 31 May 2026 00:00:00 +0000

When Neato’s parent company Vorwerk shut down the cloud, perfectly good Botvac robot vacuums lost their app, their scheduling, and even parts of their error handling. OpenNeato (the upstream project) brings them back with a small ESP32 wired to the robot’s debug serial port, serving a local web UI over WiFi with no cloud, no app, and no account.

My contribution is the Home Assistant side. I run a fork whose reason for existing is a HACS-installable custom integration: it discovers the bridge on the LAN and exposes the robot as a first-class Home Assistant device, with no YAML.

The integration polls the bridge’s local HTTP API every 5 seconds and gives you:

A vacuum entity (start/stop/pause/dock/locate/spot, battery, status, fan-speed presets, errors).
A camera entity that renders the LIDAR cleaning map.
Sensors for battery diagnostics (cycle count, voltage, temperature), brush and vacuum motor speeds, WiFi, and cleaning stats.
Switches and buttons for motors and manual controls.

The ESP32 bridge firmware, web UI, and flash tool are the upstream project’s work; the firmware changes in my fork are minor endpoints to feed the integration. The result: my own Neato D6 now works at least as well as it did on the cloud, with more telemetry than the original app ever exposed.

Full write-up: Bringing a bricked Neato back to life. Both projects are open source under the MIT license.

Pocket Rolling Stone

Sat, 30 May 2026 00:00:00 +0000

The Pocket Rolling Stone is a small handheld device that lets you feel a virtual ball rolling and sliding inside a tube that does not physically exist. You tilt it, and a synthesized marble runs from one end to the other, hits the wall, and rolls back. Nothing moves inside. The sensation is generated entirely in firmware on an ESP32-C6, from tilt sensed by a BNO085 IMU.

It is a reproduction of a haptic illusion described by Hsin-Yun Yao and Vincent Hayward in An Experiment on Length Perception with a Virtual Rolling Stone (Eurohaptics 2006), where people could estimate the length of a virtual tube purely from the rolling sensation, apparently using an internal model of gravity. The project was sparked at the workshop held in memoriam of Vincent Hayward, where it became clear that essentially one working set of the original demo was left, worth rebuilding on modern, minimal hardware. It sits in the same family as the lab’s 3D printed haptic illusions and the broader question of why touch has no museum of illusions.

A 1 kHz physics loop simulates three modes (rolling like a solid sphere, sliding with Coulomb friction, and a 3D box of three marbles of different mass), and the ball’s motion is rendered as a real-time vibration. The same signed haptic signal drives two completely different output stages, selected by a single build flag:

I2S audio path: an Adafruit ESP32-C6 Feather plus a MAX98357A amplifier driving a TITAN Haptics actuator, streaming a position-indexed wavetable at 22050 Hz.
H-bridge path: a Microbots CodeCell C6 Drive board (onboard IMU and H-bridge) driving a motor directly through PWM, with no amplifier at all. This is the build that was actually assembled into a small handheld unit; the I2S path was validated on the bench.

A Python companion app connects over USB serial or BLE to visualize and tune the simulation live.

A direction I want to explore next: rendering a localizable impact along the tube, so you feel where the virtual ball strikes, drawing on Miller et al., Sensing with tools extends somatosensory processing beyond the body, Nature 2018.

This is an ongoing personal side project, not a product. Since the original builds, a dedicated A7 board has been manufactured by JLCPCB and successfully tested in October 2026, with the physics loop running at approximately 2 kHz. The story is documented in three build posts:

The firmware and companion app are open source at github.com/Leicas/esp32-ball.

KataMarrant

Mon, 04 May 2026 00:00:00 +0000

KataMarrant is a small judo training app I build on the side. The goal is narrow and concrete: drill the 40 Gokyo no Waza, the classical Kodokan set of throws every judoka is expected to know by name, in short daily sessions. The app runs on Windows, macOS, Linux, Android, and iOS from a single Tauri 2 codebase, with vanilla JS on the frontend and Rust on the backend.

There are four practice modes. One question is the lowest-friction button on the home screen, for the moments when I have less than thirty seconds. Rafale is a ten-question burst, the warm-up I do on the metro. Drill (chrono) is a timed mode with a per-question countdown, including an audio variant where the technique name is read aloud and you pick the matching description. Browse the 40 techniques is the reference catalogue, grouped by kyo, with romaji, kanji, the French translation, and a curated demo video for each throw.

The quiz logic borrows the basics from spaced repetition: techniques you miss surface more often, a six-question cooldown prevents back-to-back repeats, and per-technique mastery scores follow you across devices when sync is enabled. Sync is opt-in, magic-link only, with no passwords stored anywhere. A daily goal, a streak counter, an XP curve, and a small grid of unlockable achievements sit on the home screen, borrowing from a decade of language-learning and fitness apps.

The technique illustrations are extracted from the Gokyo no Waza plate on Wikimedia Commons by user Mtwist (CC BY-SA). The reference videos are curated by judo.how. The project is open source, not affiliated with the Kodokan or any judo federation, and is documented on this site in a series of build posts.

Haply Inverse3

Mon, 01 Jan 2024 00:00:00 +0000

Haply 2diy

Sun, 31 Dec 2023 14:55:19 -0400

Broadcasting Haptic

Thu, 11 Jun 2020 13:32:56 -0400

3D printed haptic illusions

Sat, 12 Oct 2019 14:55:19 -0400

Haptic for OR/ICU

Sat, 12 Oct 2019 14:55:11 -0400

At present, the operating room (OR) and intensive care unit (ICU) are noisy environments, exacerbated by frequent alarms. Regardless of whether the alarms are valid or false, all command attention, raise stress, and are often irrelevant to the responsibilities of individual clinicians. To cope with these problems, we are investigating the possibility of using audio only for those alarms that should be announced to the entire team, but delivering other alarm cues individually, through haptics vibrations.

Haptic device for sensory reeducation

Sat, 12 Oct 2019 14:55:03 -0400

Nerve damage, frequently caused by injury, can result in the loss of sensorimotor functions in certain parts of the hand. After suturing the nerve, unpleasant sensations on contact, including tingling and electric shocks are often felt. Following nerve regrowth, it is necessary to re-train the brain to interpret the signals from these nerves correctly. This project involves the design of haptic devices to help in this process of sensory reeducation, which can involve two phases, depending on the severity of the loss of sensitivity: relearning how to localize sensations, and differentiation of shapes and textures in the identification of objects.

Wearable Haptic devices

Sat, 12 Oct 2019 14:54:52 -0400

HapticPhonemes

Sat, 12 Oct 2019 14:54:40 -0400

The tactile sense can be used as a channel for general communication, especially in contexts where the visual and auditory modalities are occupied with other tasks or compromised. We propose a new method for communicating generic words through the sense of touch that relies on delivering vibration patterns, representing the phonemes composing the words, to the user’s skin through two vibrotactile transducers worn on the forearm. The novelty of this technique is that vibration patterns are created from the audio of the corresponding English phoneme, resulting in vibration patterns that resemble physical characteristics when uttering the phoneme during normal speech.

BarryWhaptics

Sat, 12 Oct 2019 14:46:35 -0400

Haptic Micro Teleoperation

Mon, 16 Oct 2017 13:05:29 -0400

Force Sensors

Mon, 16 Oct 2017 13:00:27 -0400

High Fidelity Haptic Devices

Mon, 16 Oct 2017 13:00:22 -0400