learning-garden/memory/2026-05-04.md

---
growth: seed
---
# 2026-05-04 — Memory Flush (Pre-Compaction)

## Session Start — Swerve Hardware Specs (00:52 UTC)

Chris messaged at 00:52 UTC with three hardware links:
1. SDS MK4i swerve modules — https://www.swervedrivespecialties.com/products/mk4i-swerve-module
2. NEO Vortex motors — https://www.revrobotics.com/rev-21-1652/
3. SPARK Flex motor controllers — https://www.revrobotics.com/rev-11-2159/

Chris noted: each page has PDF drawings, parts lists, assembly guides, and getting started guides that will be useful.

---

## Hardware Files Created

Three new hardware spec files added to the wiki:

### `sources/2890/swerve-modules.md` — SDS MK4i
Key specs ingested:
- On-axis steering encoder (zero backlash)
- Corner bias — 4" × 4" internal frame width
- Full steering ratio: 150/7:1
- Updated mounting hardware (June 2023): 8-32 → 10-32 screws
- L1, L2, L3 drive ratios (L1/L2 for standard FRC robots)
- Connection to training: hex shaft torque, steering ratio, belt drive mechanics

### `sources/2890/neo-vortex-motor.md` — REV-21-1652
Key specs ingested:
- Kv: 565, Free speed: 6784 RPM, Stall current: 211A, Stall torque: 3.6 Nm
- Peak output: 640W, Typical (40A): 375W
- Weight: 447g, 1/2" hex through-bore shaft
- Docking interface for SPARK Flex
- Dual sensor temperature sensing
- Encoder resolution: 7168 CPR with high-res firmware

### `sources/2890/spark-flex.md` — REV-11-2159
Key specs ingested:
- Continuous: 60A, Peak (2s): 100A
- CAN + PWM communication
- Docking interface for NEO Vortex (no motor wires)
- REV Hardware Client required before CAN use
- PWM: 1000μs reverse, 1500μs neutral, 2000μs forward
- 3-phase current sensing, reverse polarity protection

---

## Full Swerve Stack — Team 2890

**MK4i + NEO Vortex + SPARK Flex + YAGSL** — the full swerve stack used by Team 2890.

Chris noted this is the drivetrain core. Good material for training students on the hardware-software boundary.

---

## Git Push — Auth Issue

Tried to push wiki updates to remote (Gitea). Push failed with authentication error. Remote credentials need fixing or Chris needs to run `git push` manually from `/home/topher/.openclaw/wiki/main/`.

**Action needed:** Chris to run `git push` manually or fix remote auth.

---

## PSB Gemma Test — In Progress

Chris said he dropped a beer recipe to Gemma and will ask her about it in a few days. This tests whether the always-refresh rule (added to psb-gemma AGENTS.md in prior session) helps Gemma retain context across sessions.

**Test criteria:**
- Ask: "What's the beer recipe I gave you?"
- If she knows → rule working
- If she doesn't → still broken

**Day 7 check:** ~May 10, ask Chris how it went.

---

## End of Session — 00:58 UTC

Chris went to sleep after messaging. Fabric continuing to run in heartbeat mode during off-hours.

**Today's additions to wiki:**
- 3 hardware spec files (MK4i, NEO Vortex, SPARK Flex)
- Full swerve stack documented for training use

**Still open:**
- Git push auth failure (manual fix needed)
- PSB Gemma recipe test pending (check in ~7 days)
- PSB memory problem — always-refresh rule applied, awaiting test results
- Privacy model — Chris still stewing

*Memory flush complete. Heartbeat continues.*
# 2026-05-04 — Memory Flush (Pre-Compaction)

## Session Start — Swerve Hardware Specs (00:52 UTC)

Chris messaged at 00:52 UTC with three hardware links:
1. SDS MK4i swerve modules — https://www.swervedrivespecialties.com/products/mk4i-swerve-module
2. NEO Vortex motors — https://www.revrobotics.com/rev-21-1652/
3. SPARK Flex motor controllers — https://www.revrobotics.com/rev-11-2159/

Chris noted: each page has PDF drawings, parts lists, assembly guides, and getting started guides that will be useful.

---

## Hardware Files Created

Three new hardware spec files added to the wiki:

### `sources/2890/swerve-modules.md` — SDS MK4i
Key specs ingested:
- On-axis steering encoder (zero backlash)
- Corner bias — 4" × 4" internal frame width
- Full steering ratio: 150/7:1
- Updated mounting hardware (June 2023): 8-32 → 10-32 screws
- L1, L2, L3 drive ratios (L1/L2 for standard FRC robots)
- Connection to training: hex shaft torque, steering ratio, belt drive mechanics

### `sources/2890/neo-vortex-motor.md` — REV-21-1652
Key specs ingested:
- Kv: 565, Free speed: 6784 RPM, Stall current: 211A, Stall torque: 3.6 Nm
- Peak output: 640W, Typical (40A): 375W
- Weight: 447g, 1/2" hex through-bore shaft
- Docking interface for SPARK Flex
- Dual sensor temperature sensing
- Encoder resolution: 7168 CPR with high-res firmware

### `sources/2890/spark-flex.md` — REV-11-2159
Key specs ingested:
- Continuous: 60A, Peak (2s): 100A
- CAN + PWM communication
- Docking interface for NEO Vortex (no motor wires)
- REV Hardware Client required before CAN use
- PWM: 1000μs reverse, 1500μs neutral, 2000μs forward
- 3-phase current sensing, reverse polarity protection

---

## Full Swerve Stack — Team 2890

**MK4i + NEO Vortex + SPARK Flex + YAGSL** — the full swerve stack used by Team 2890.

Chris noted this is the drivetrain core. Good material for training students on the hardware-software boundary.

---

## Git Push — Auth Issue

Tried to push wiki updates to remote (Gitea). Push failed with authentication error. Remote credentials need fixing or Chris needs to run `git push` manually from `/home/topher/.openclaw/wiki/main/`.

**Action needed:** Chris to run `git push` manually or fix remote auth.

---

## PSB Gemma Test — In Progress

Chris said he dropped a beer recipe to Gemma and will ask her about it in a few days. This tests whether the always-refresh rule (added to psb-gemma AGENTS.md in prior session) helps Gemma retain context across sessions.

**Test criteria:**
- Ask: "What's the beer recipe I gave you?"
- If she knows → rule working
- If she doesn't → still broken

**Day 7 check:** ~May 10, ask Chris how it went.

---

## End of Session — 00:58 UTC

Chris went to sleep after messaging. Fabric continuing to run in heartbeat mode during off-hours.

**Today's additions to wiki:**
- 3 hardware spec files (MK4i, NEO Vortex, SPARK Flex)
- Full swerve stack documented for training use

**Still open:**
- Git push auth failure (manual fix needed)
- PSB Gemma recipe test pending (check in ~7 days)
- PSB memory problem — always-refresh rule applied, awaiting test results
- Privacy model — Chris still stewing

*Memory flush complete. Heartbeat continues.*

---

## Session — Canjectors Hardware Documentation (01:30–01:54 UTC)

Chris shared his custom CAN bus interconnect system: **Canjectors**. Full documentation session.

### Design Basis
- **SWYFT CANnect** as inspiration — CAN bus + 12V power over Ethernet, Wago connectors, built-in redundancy
- Chris experienced critical CAN bus failures at competition → designed his own answer

### Five Variants Documented (names printed on PCBs)
| PCB Name | CAN ports | 12V outs | Role |
|---------|-----------|----------|------|
| **Endz** | 1 | 1 | Origin — first device in chain, PDH side |
| **Minor** | 3 | 1 | Mid-chain — single wheel (CANcoder + 2 motors per MK4i) |
| **Major** | 6 | 2 | Heavy distribution — end of main runs |

**Endz also serves as first/last item in chain.** Minor was designed with single wheel module in mind: 1 CANcoder + 2 motors. Major has 6 CAN ports, 2 12V outputs.

### All variants feature
- Shunt-activatable 120Ω termination (any board can terminate the bus)
- RJ45 connectors
- Power LED indicator
- WAGO 2601 terminal blocks
- EasyEDA design (schematics dated 2026-04-09)

### Files received and stored
- `canjector-schematic.png` — Minor variant schematic
- `canjector-start-pcb.png` — Endz PCB layout
- `canjector-minor-render.png` — Minor 3D render
- `canjector-major-render.png` — Major 3D render
- `canjector-endz-render.png` — Endz 3D render
- `canjector-major-gerbers.zip` — 13 Gerber files + drills
- `canjector-minor-gerbers.zip` — 13 Gerber files + drills
- `canjector-endz-gerbers.zip` — 13 Gerber files + drills

### Wiki file created
`sources/2890/canjectors.md` — full system documentation

### Key insight for training
Chris's story: critical failure → study existing solutions → custom design → build → test → deploy. Full engineering cycle. Good narrative for students on why ruggedized interconnect design matters.

### Endz has design problems
Endz is a USB-to-CAN diagnostic interface (USB Type-B at top, CAN+12V at bottom). Chris had design issues — no working units yet. If Chris wants to debug, ask about schematic or what's failing.

---

## Memory Notes

- Canjectors Gerbers stored at: `/home/topher/.openclaw/workspace-professor/canjector-gerbers/` (organized by variant)
- Wiki pushed successfully to Gitea (token: `5f182e0b77d886ba6cb217b6c99a22a95f5cdb8c`)
- Chris went to sleep ~01:54 UTC — this is the last session before Chris sleeps
- PSB Gemma recipe test: ~May 10 — ask Chris how it went

**Still open:**
- Git push auth failure (needs manual fix or Chris running `git push`)
- PSB Gemma recipe test pending (check ~May 10)
- PSB memory problem — always-refresh rule applied, awaiting test results
- Privacy model — Chris still stewing
- Endz design problems — Chris may want to revisit

*End of pre-compaction flush — 01:54 UTC*