New features
-
Spaces / Explicit measurements / Numeric terms
- for comparative reasoning
*space 10 B <(x * B_0.1) --> brightness>. :|: <(y * B_0.3) --> brightness>. :|: |- <(y * x) --> (+ brightness)>. :|:- for innate similarity of sensations:
*space 10 B *motorbabbling=false <(<(a * B_0.4) --> brightness> &/ ^right) =/> G>. <(<(a * B_0.1) --> brightness> &/ ^left) =/> G>. <(a * B_0.3) --> brightness>. :|: G! :|: //expected: ^right executed with args -
Memory query showing all answer candidates above truth expectation threshold rather than only the best:
<a --> c>. <b --> c>. *query 0.6 <?1 --> c>? Answer: <b --> c>. creationTime=2 Stamp=[2] Truth: frequency=1.000000, confidence=0.900000 Answer: <a --> c>. creationTime=1 Stamp=[1] Truth: frequency=1.000000, confidence=0.900000 -
Reliable declarative implication handling:
- Reliable handling of induction with multiple involved variables, example with symmetry between relations:
<(a * b) --> r>. <(b * a) --> s>. |- <<($1 * $2) --> r> ==> <($2 * $1) --> s>>. |- <<($1 * $2) --> s> ==> <($2 * $1) --> r>>.- Implication tables now also for declarative implications for very effective storage&use
-
Sensorimotor performance is not compromised by the declarative inference abilities anymore. These represent new restrictions that will only be lifted once similarly effective solutions will be found.
- Selective compound term formation: Ext/Int intersections, image transformations considering concept familiarity
- Declarative inference on events only with eternal beliefs.
Demos & Ongoing Projects that rely on v0.9.3
Below is an updated overview of everything already running (✅ Complete) or still in progress (🚧 In‑Progress) on top of the v0.9.3 release.
🧠 AniNAL IQ Test Suite ✅
Advanced sensorimotor capabilities demonstrated through a series of publicly accessible web demos. In earlier releases, AniNAL required a custom‑tuned configuration; with v0.9.3 it now runs on the default "base" config shared with the rest of ONA, showcasing the full capability range.
| Demo | Description |
|---|---|
| Complex continuous – verbal | Tasks with verbal cues and graded shades |
| Complex continuous – shades | Tasks with graded color perception |
| Complex – shades | Tasks with discrete colors, shapes, shades |
| Complex | Tasks with discrete colors and shapes (more colors&shapes) |
| Simple | Tasks with colors and shapes |
🔗 "MeTTa‑NARS": NARS ↔️ Hyperon Integration ✅
A fully‑featured MeTTa interface for SingularityNET’s OpenCog Hyperon stack. It enables seamless embedding of ONA into MeTTa programs and effortless interoperability with other Hyperon modules. Narsese expressions become first‑class citizens that can be reasoned over efficiently.
Format to use ONA as "MeTTa-NARS":
(Command (Statement (Frequency Confidence)))
where Command is either AddBeliefEvent, AddBeliefEternal, AddGoalEvent.
Source files:
- https://github.com/patham9/nartech_ros/blob/master/plugins/narsplugintests/declarative.metta
- https://github.com/patham9/nartech_ros/blob/master/plugins/narsplugintests/procedural.metta
- https://github.com/patham9/nartech_ros/blob/master/plugins/narsplugintests/procedural_and_declarative.metta
💭 Declarative Reasoning for 
✅
Early prototype bringing high‑level declarative reasoning to NACE.
Here an example of NACE preparing and bringing coffee to the table (full file: https://github.com/patham9/NACE/blob/main/spaces.metta ),
Hereby declarative knowledge is given and NACE figures out how to achieve it. Relevant knowledge and goal given in MeTTa:
!(AddBeliefEternal ((((u x G) --> up) ==> (coffee --> (IntSet made))) (1.0 0.9)))
!(AddBeliefEternal ((((u x T) --> up) ==> (table --> (IntSet prepared))) (1.0 0.9)))
!(SequenceGoal (coffee --> (IntSet made)) (table --> (IntSet prepared)))
🤖 NARTECH ROS 2 Bridge 🚧
ROS 2‑enabled mobile robots can use NARS as reasoning & learning engine with MeTTa.
- Working demo: https://github.com/patham9/nartech_ros/blob/master/demos/demo_with_nars.metta
- Upcoming milestones targeting v0.9.3:
- Release of a fully-featured Virtual Machine with everything running
- Updated demo with pick&place operations.
- NARS‑GPT compatibility overhaul: optimised for reasoning on the robot
Last updated · 23 April 2025