Functional Non-Axiomatic Reasoning System
A pure functional Clojure port of OpenNARS for Applications (ONA). The entire reasoner state is a single immutable map — no globals, no mutation. Supports sensorimotor reasoning (NAL 6-8) and declarative inference (NAL 1-6).
# Interactive shell
bb repl
# Run a .nal file
bb nal OpenNARS-for-Applications/examples/nal/asthma2.nal
# Run all tests
bb test:allfNARS runs on six targets from the same .cljc source:
| Target | Command | Use case |
|---|---|---|
| nbb (bun) | bb repl |
Development |
| JVM Clojure | bb repl:jvm |
Max speed |
| GraalVM native | ./target/fnars |
Fast startup binary |
| Babashka | bb test:bb |
Scripting |
| shadow-cljs | bb build:shadow |
Node.js bundle |
| Browser | bb build:browser |
Web UI |
bb build:shadow # target/fnars.js (Node)
bb build:browser # public/js/fnars-browser.js
bb build:uber # target/fnars.jar (uberjar)
bb build:native # target/fnars (GraalVM native binary)The interactive shell accepts Narsese statements, cycle counts, and *commands:
>> <cat --> animal>.
Input: <cat --> animal>. Truth: {1.000000 0.900000}
>> <animal --> living>.
Input: <animal --> living>. Truth: {1.000000 0.900000}
>> *nallevel=6
Semantic inference NAL level set to 6
>> 100
>> <cat --> living>?
Answer: <cat --> living>. Truth: {1.000000 0.810000}
Sensorimotor reasoning (NAL 6-8) is always on. For declarative inference (NAL 1-6), set the level with *nallevel=N.
You can also pipe input:
echo '*nallevel=6' | cat - file.nal | ./target/fnarsbb test:nbb # Unit + snapshot tests (nbb)
bb test:jvm # Unit + snapshot tests (JVM)
bb test:bb # Unit + snapshot tests (babashka)
bb test:shadow # Unit + snapshot tests (shadow-cljs)
bb test:browser # Browser smoke tests (Playwright)
bb test:ch6 # Chapter 6 operant conditioning (4 tests)
bb test:ch7 # Chapter 7 identity matching
bb test:nal # NAL 1-6 comparison tests (15 tests, ONA-verified)
bb test:all # All of the above- Immutable state: The NAR is a single Clojure map threaded through every function
- Binary heap terms: Terms are flat vectors of 64 integer atom IDs (root at index 0, left child at 2i+1, right child at 2i+2)
- Integer atom registry: Atoms are integer IDs for fast comparison (
==), mapped to keywords via a shared registry - Temporal mining: Two-phase cycle — Phase 1 mines
<(precondition &/ ^op) =/> postcondition>triples, Phase 2 derives declarative implications and sequences - Decision making: Goals trigger operation execution by matching implication preconditions against current beliefs
src/fNARS/
truth.cljc Truth functions (revision, deduction, induction, ...)
term.cljc Binary heap term encoding
stamp.cljc Evidential base (zip-merge, overlap detection)
event.cljc Belief/goal event structures
implication.cljc Implication structures
nar_config.cljc Tunable parameters
atom_registry.cljc Keyword <-> integer atom ID mapping
priority_queue.cljc Bounded sorted priority queue
table.cljc Implication tables (sorted by expectation)
concept.cljc Concept (belief, goal, spike, tables, usage)
memory.cljc Concept store, inverted atom index, time index
inference.cljc Temporal/procedural inference rules
rule_table.cljc NAL 1-6 declarative rule table
narsese.cljc Term construction utilities
variable.cljc Unification, substitution, variable introduction
decision.cljc Decision making, motor babbling
cycle.cljc Main inference cycle
nar.cljc Top-level NAR API
parser.cljc Narsese parser (instaparse grammar)
shell.cljc Interactive shell
nal_runner.cljc .nal file runner
mprf.cljc Machine Psychology Research Framework
test/fNARS/
nar_test.cljc Unit tests
snapshot_test.cljc Snapshot tests
ch6_test.cljc Operant conditioning tests
ch7_test.cljc Identity matching test
nal_comparison_test.cljc NAL 1-6 ONA-verified comparison tests
Replicates experiments from the machine psychology framework:
bb mprf op1 # Simple operant conditioning
bb mprf op2 # Discriminative operant conditioning
bb mprf op3 # Transitive inference- Babashka — task runner
- bun + nbb — ClojureScript runtime
- instaparse — Narsese parser (nbb-compatible fork, included as git submodule)
- GraalVM — native binary builds (optional)
- Hammer, P. (2022). Reasoning-learning systems based on non-axiomatic reasoning system theory. In International Workshop on Self-Supervised Learning (pp. 89-107). PMLR.
- Hammer, P., & Lofthouse, T. (2020). 'OpenNARS for Applications': Architecture and Control. AGI 2020.
- Wang, P. (2013). Non-Axiomatic Logic: A Model of Intelligent Reasoning. World Scientific.
MIT