Sley Examples

Thirty small Sley programs you can run, test and change. Seventeen primitives each exercise one to three native operations, such as checked addition or a Boolean combinator. Thirteen composed programs solve everyday problems: retry decisions, invoice line totals, page windows, ledger posting, quorum decisions, days in a month. Every example ships its authoring input, a portable executable artifact, independently expected cases and a script that rebuilds or reimports it and checks the results.

Counts, kept as separate categories: 233 external assertions across the corpus (78 primitive + 155 composed) and 155 advisory workbench checks in the composed collection.

Run the composed examples

With the Sley 2.1.6 release unpacked next to this checkout:

git clone https://github.com/sley-lang/examples.git && cd examples
./composed/reproduce.sh --agent ../sley-2.1.6-linux-x86_64/bin/sley-agent

The script imports each example's checked-in pack into a fresh workspace, runs its workbench tests and its external checker, and refreshes views/fn.txt, which comes out identical.

New programs are easiest to write in Sley's typed Python dialect (sley-agent try --familiar); the Sley Quickstart starts there. The sley-lang/tools toolkit (the sley-tools CLI, Python/TypeScript/MCP SDKs, Playground, Repair Lab, Data Pipe, importers and Review) built the primitives below. Its own N01–N10 catalog of ready-to-run programs is not duplicated here.

Primitive examples (17 total, 78 external assertions)

ID	Program	What runs natively	Cases
S01	programs/bool-and	bool_and on two Bool	4
S02	programs/bool-nor	bool_or + bool_not	4
S03	programs/uint32-minimum	greater_equal vs minimum 5	5
S04	programs/is-negative	less_than vs 0	5
S05	programs/equals-answer	equal vs 42	5
S06	programs/safe-add	int_add_checked → {sum: CheckedSInt64}	5
S07	programs/bool-status	bool_and/bool_or/bool_not → {both, either, neither}	4
S08	programs/is-positive	greater_than vs 0	5
S09	programs/safe-sub	int_sub_checked → {difference: CheckedSInt64}	5
S10	programs/bool-or	bool_or on two Bool	4
S11	programs/safe-mul	int_mul_checked → {product: CheckedSInt64}	5
S12	programs/safe-neg	int_neg_checked → {negated: CheckedSInt64}	5
S13	programs/not-equal	not_equal vs 42 → Bool	5
S14	programs/at-or-below	less_equal vs 0 → Bool	5
S15	programs/at-or-above	greater_equal vs 0 → Bool	5
S16	programs/bool-not	bool_not → Bool	2
S17	programs/wrap-ok	result_ok → {wrapped: CheckedSInt64}	5

Each package holds build/graph.json (authoring recipe), tests/cases.json (independently expected), program.json (manifest bound to the artifact by sley-tools artifact export --bind-manifest), and artifact/native-graph.json (portable native graph). The accepted SSMC1 graph is the program; the recipe is only the authoring input. Sley has no canonical source text. Later cores also accept optional text authoring (the familiar frontend from 2.0.5, the Python dialect from 2.1.1), which compiles to the same graph; these examples do not use it.

Toolchain pin

• Core: Sley 2.0.1 (sley-2.0.1-linux-x86_64.tar.gz, commit c748dda), binary sha 90f48df4….
• Builder: sley-lang/tools sley-tools backend, graph_authoring.v0.
• External execution assertions: 78 total across the seventeen programs, all PASS at build time; every artifact reimports into a fresh workspace and re-passes.
• Native test admission: PENDING_CORE_EXECUTOR on the selected 2.0.1 core (NOT_RUN: production native executor unavailable). This is reported separately and is not covered by the assertion counts above.

Rebuild everything

./scripts/rebuild.sh \
  --tools /path/to/sley-tools-v0.0.1-rc.5/bin/sley-tools \
  --core /path/to/sley-2.0.1-linux-x86_64/bin/sley

Use sley-tools v0.0.1-rc.5 for the 2.0.1 pin: each sley-tools release targets one core, and rc.5 is the release built for 2.0.1 (sley-tools --version prints the tested core commit, c748dda…). --tools=PATH and --core=PATH are accepted too.

The script re-runs init + test + bound artifact export + fresh-workspace artifact import + re-test for all seventeen programs and refreshes program.json and artifact/native-graph.json in place. It refuses to run with uncommitted changes unless --allow-dirty is passed. Entity and object ids are assigned per workspace, so every rebuild rewrites the artifacts and manifests even when the recipes are unchanged; review a rebuild diff for changes to the core pin, interface or cases rather than ids.

Running with newer Sley releases

The checked-in examples also run on current releases without changes:

• Composed (C01–C13): ./composed/reproduce.sh --agent with the Sley 2.1.6 sley-agent reproduces all thirteen examples: 155/155 workbench checks, 155/155 external assertions, and identical views/fn.txt (checked 2026-10-26; 2.1.5 gave the same on 2026-10-24). Every per-example reproduce block in composed/*/README.md passed the same way with 2.1.5. --dist verifies only the pinned 2.0.3 archive, so pass --agent for a newer release.
• Primitives (S01–S17): sley-tools v0.0.1-rc.6 targets the Sley 2.1.1 core and imports the checked-in 2.0.1 artifacts directly: sley-tools artifact import programs/<id>/artifact/native-graph.json PROJECT --core /path/to/sley-2.1.1-linux-x86_64/bin/sley, then sley-tools test --project PROJECT, re-passes all 78 assertions, and every input/output example in programs/*/README.md returns the documented value under sley-tools run. Running rebuild.sh with rc.6 and the 2.1.1 core also passes, and rebinds the artifacts to 2.1.1.

Composed examples (C01–C13, sley-agent 2.0.3)

Thirteen composed programs authored with the official 2.0.3 sley-agent workbench, kept in composed/ strictly separate from the seventeen 2.0.1 primitives above. The 2.0.1 pin, the seventeen packages, and the rebuild script are unchanged by them.

ID	Example	Function	Workbench checks / External
C01	composed/retry-decision	retry_decision → Result<RetryState,RetryError>	11 / 11
C02	composed/invoice-line-total	line_total → Result<i64,InvoiceError>	12 / 12
C03	composed/duration-breakdown	split_duration → Result<(d,h,m,s),DurationError>	9 / 9
C04	composed/page-window	page_window → Result<(start,count,has_more),PageError>	16 / 16
C05	composed/ledger-post	post_ledger → Result<i64,LedgerError>	15 / 15
C06	composed/window-overlap	window_overlap → Result<(overlap,gap),WindowError>	15 / 15
C07	composed/quorum-decision	quorum_decision → Result<QuorumOutcome,QuorumError>	14 / 14
C08	composed/cooldown-gate	cooldown_gate → Result<(GateState,i64),GateError>	15 / 15
C09	composed/days-in-month	days_in_month → Result<i64,DateError>	15 / 15
C10	composed/clamp-value	clamp_value → Result<i64,ClampError>	10 / 10
C11	composed/safe-quotient	safe_quotient → Result<i64,QuotientError>	8 / 8
C12	composed/safe-absolute	abs_value → Result<i64,AbsError>	8 / 8
C13	composed/sign-class	sign_class → Sign	7 / 7

Each example holds exact authoring frames as run, a contract (parameter semantics, error precedence), independently expected cases, an external checker, a readable workbench view, the exported pack plus author name map, and a before/after record of one genuine program modification. composed/corrections.md retains one repaired authoring error and two independently justified expectation corrections. composed/admission-probe.md records the fresh-import edit/commit probe.

Reproduce (needs the 2.0.3 agent binary):

./composed/reproduce.sh --agent /path/to/sley-2.0.3-linux-x86_64/bin/sley-agent

The workbench and scripts need writable scratch space. If the default temporary directory is small or quota-limited, point TMPDIR at a writable directory before running (the scripts and the agent honor it); no path is hardcoded and no cleanup infrastructure is provided.

Authoring workflow: composed/authoring.md is the default — pinned guide plus the frozen serializer/parse/recovery appendix first, with targeted consultation of a small relevant example only when a specific problem warrants it.

Persistence limitation: a combined code+tests commit is refused (TXN_TEST_EVIDENCE_UNSUPPORTED), so each example was committed split — code-only, then tests-after (transactions in catalog.json). Code-only plus tests-after is two transactions, not a test-admitted atomic commit. Native test admission is not run — unavailable in 2.0 — and no count in composed/ is test admission.

More Sley

• sleylang.org: documentation, the Python dialect and the walkthrough.
• Sley Arena: write a bot in Sley and climb a public ladder.
• sley-lang/tools: the toolkit that built the primitives.

License

This repository is licensed under the Apache License, Version 2.0. That covers the original examples, documentation, supporting scripts, and project-owned artifacts, except where another license or notice is expressly identified. See NOTICE for attribution.

---

Found a useful pattern? Star Sley Examples on GitHub · Follow @SleyLanguage · sleylang.org

_{Built by Greyforge Labs}