Skip to content

Decima8 IDE

Visual environment for baking neuromorphic personalities

Status: Public
Version: 1.0


🏗️ What is Decima8 IDE?

Decima8 IDE is a visual environment for baking neuromorphic personalities. Here you manually configure tiles, weights, and thresholds while observing the recognition process in real-time.


📥 Download

Binary files

IDE for Windows

SHA256: 88312735a2647986dc9de1ee9778e4cc88d2e637

IDE for Linux

SHA256: d6f49f6e7ee48d6e9a4576d1e6ecaaa41664610e


System Requirements

Requirement Value
OS Windows 10/11, Linux (Ubuntu 20.04+)
Memory 8 MB RAM minimum
Disk 2 MB free space

🏗️ IDE Interface

Decima8 IDE Interface

Visual baking environment: input pattern accordion, tile swarm, parameter and solution panels


🧩 Interface Components

🛠 Control Panel

Button Function
▶ FLASH Run EV_FLASH cycle
⏸ RESET Reset accumulators of selected domain or all
🔁 Auto-Bake Automatic baking under pattern
⚙ Swarm Params Global settings (size, domains)

🪗 Accordion (Input Patterns)

VSB tape — Visual representation of input data:

Parameter Description
8 columns 8 lanes VSB[0..7]
Values 0..15 Level16
Tape scrolls Data fed by ticks

Pattern Example:

Tick 1: [5, 10, 3, 8, 12, 0, 7, 15] → Recognition "5"
Tick 2: [0, 2, 7, 1, 9, 4, 6, 3]    → Recognition "0"

🕸 Swarm (Tile Array)

Tile array visualization:

Element Description
Each tile One neuron with local memory
Color Activity (thr_cur16)
White Locked status (fuse latched)
Arrows Children activation directions (N,E,S,W...)

Display Modes:

Mode Description
Weights 8×8 weight matrix
Activation Current thr_cur16
Routing Routing flags

🎛 Tile Parameter Panel

Click on tile to open editor:

Parameter Description Range
BUS_R Read bus (ACTIVE source) 0/1
BUS_W Write to bus (WRITE phase) 0/1
Threshold Fuse range [thr_lo..thr_hi] -32768..+32767
Decay Decay force to zero 0..32767
Pattern ID Recognized pattern ID 0..32767
Domain Reset group 0..15
Priority Winner on collision 0..255
Directions Children activation (N,E,S,W,NE,SE,SW,NW) 8 bits

🎯 Solution Panel

Recognized patterns output:

Field Description
Pattern Recognized pattern ID
Confidence Confidence (0..1)
Tile ID Which tile made decision

Output Example:

[EV_FLASH]
frame_tag,domain_id,tile_id,collision,pattern_ids,bus16
2,0,256,0,1,0|0|0|0|0|0|0|0
8,0,385,0,2,0|0|0|0|0|0|0|0
13,0,386,0,3,0|0|0|0|0|0|0|0

🔄 Workflow

1. Load Personality

1. Open personality (.d8p file)
2. Swarm shows tile topology
3. Load vsb tape from file or enable UDP listening
4. Accordion shows VSB chords - what's fed to machine input
5. If tape loaded from file, can press FLASH / BACK / RESET to run patterns
6. When working via network, swarm reacts to chords from socket

2. Configure Tiles

1. Click on tile in swarm
2. Configure thresholds, weights, decay
3. Set activation directions
4. Repeat for all tiles

3. Baking

1. Select required chord in accordion (right-click)
2. Select tile that should latch on it
3. Press 🔁 Auto-Bake
4. IDE selects weights for pattern
5. Check FLASH, adjust firing corridor
6. Save as .d8p

4. Recognition

1. Press ▶ Play
2. Tape moves through swarm
3. Solution panel shows recognized patterns
4. When working via network, swarm reacts to chords from socket

🤗 AI Agent (Coming Soon)

In development — AI agent for automatic weight and topology selection:

Stage Description
Task You set pattern bank and target metrics
Agent Runs machine, selecting weights and topology
Result Ready baked personality for store

Bake the Future. Build the Substrate. 🛠️⚡️