KAI OS v0.3.1 First-Run Tour

KAI OS v0.3.1 First-Run Tour

KAI OS v0.3.1 is the first public release shaped around a clear product loop:

Git change -> kaios review -> review.md + trace.json + capsule.json -> review/audit/baseline gate

KAI OS is not a chatbot framework or Kotlin LangChain clone. It is a local-first Evidence OS for AI agents in Kotlin: agent runs become process traces, replayable capsules, syscall ledgers, and CI-ready runtime evidence.

Agent    = Process
Workflow = Scheduler
Tool     = Syscall
Memory   = Process state

Try It First

Local install:

curl -fsSL https://morning-verlu.github.io/KAI/install.sh | sh
export PATH="$HOME/.kaios/bin:$PATH"
kaios tour

No local Java or Gradle setup:

https://codespaces.new/morning-verlu/KAI?quickstart=1

Then run:

./scripts/codespaces-smoke.sh

Docker path:

docker build -t kaios:local .
docker run --rm kaios:local tour

No install at all:

• Evidence Viewer: https://morning-verlu.github.io/KAI/evidence-viewer.html
• Start here: https://github.com/morning-verlu/KAI/blob/main/START_HERE.md
• Evidence Sample: https://github.com/morning-verlu/KAI/tree/main/examples/evidence-sample
• Evidence Artifact Map: https://morning-verlu.github.io/KAI/assets/kaios-evidence-map.png
• JVM Service Review: https://github.com/morning-verlu/KAI/tree/main/examples/jvm-service-review
• Baseline Gate: https://github.com/morning-verlu/KAI/tree/main/examples/baseline-gate

Kotlin/JVM runtime library path:

./gradlew -p examples/kotlin-runtime-api run

• Kotlin/JVM evaluation path: https://github.com/morning-verlu/KAI/blob/main/docs/KOTLIN_JVM_EVALUATION.md
• Kotlin Runtime API: https://github.com/morning-verlu/KAI/blob/main/docs/KOTLIN_API.md
• Runnable example: https://github.com/morning-verlu/KAI/tree/main/examples/kotlin-runtime-api

Help In 5 Minutes

If the Evidence OS direction feels useful, the fastest way to help is to make
one public signal or one small verification concrete:

• Star the repo if you want local-first agent evidence to exist in the
  Kotlin/JVM ecosystem: https://github.com/morning-verlu/KAI
• Share the no-install Evidence Viewer with one Kotlin/JVM developer:
  https://morning-verlu.github.io/KAI/evidence-viewer.html
• Pick a small launch task from the Contributor Board:
  https://github.com/morning-verlu/KAI/blob/main/docs/CONTRIBUTOR_BOARD.md
• Add a Kotlin API capability recipe: #14
• Add a denied-syscall evidence walkthrough: #15
• Verify full Docker smoke on a normal network: #16

The shortest rationale is here:

• Why Star KAI OS: https://github.com/morning-verlu/KAI/blob/main/docs/WHY_STAR.md

What The Tour Proves

kaios tour creates a disposable Git workspace, runs the no-key onboarding path, makes a small code change, runs kaios review, and prints pointers to:

• Markdown review artifact.
• process table with PID, state, tokens, memory, syscalls, tool time, and cost.
• kaios.process-trace/v1 JSON.
• replayable kaios.run-capsule/v1.
• kaios.review/v1 output.
• PR-friendly evidence summary.
• recovery dry-run report.

The point is that agent work leaves runtime evidence that can be inspected, copied, replayed offline, and compared in CI without requiring an external model provider.

Highlights

• Added kaios tour, a no-key walkthrough for the full Evidence OS loop.
• Added kaios.tour/v1 JSON output for docs, launch posts, and smoke checks.
• Added kaios try as a friendly alias for the tour.
• Added checked-in evidence examples so visitors can inspect traces, capsules, replay, and baseline drift before installing.
• Added a no-install Evidence Viewer for inspecting a process table, syscall ledger, replayable capsule, and baseline gate from the browser.
• Added Codespaces and Docker trial paths for no-local-Java evaluation.
• Added a Kotlin Runtime API example for teams evaluating KAI OS as an embeddable JVM library.
• Added visual evidence artifacts for launch posts and first-run evaluation.
• Updated README, install docs, launch site, launch kit, and quickstart docs around the tour-first path.

Feedback Wanted

I am looking for feedback from Kotlin/JVM developers, OSS maintainers, and agent infrastructure builders:

• Does the Agent = Process / Workflow = Scheduler / Tool = Syscall model make the project legible?
• Would replayable run capsules help in PR review, support, or CI?
• Is kaios review useful enough to run on a real JVM/Kotlin repo?
• Does the Kotlin Runtime API feel like the right abstraction boundary?

Feedback links:

• Discussion: #8
• Focused Kotlin/JVM feedback discussion: #17
• Kotlin/JVM evaluation path: https://github.com/morning-verlu/KAI/blob/main/docs/KOTLIN_JVM_EVALUATION.md
• Evidence feedback issue: https://github.com/morning-verlu/KAI/issues/new?template=evidence_feedback.yml

Verification

Core release verification:

./gradlew test installDist distZip distTar --no-daemon
./scripts/repository-ci-smoke.sh
./scripts/codespaces-smoke.sh
./scripts/evidence-samples-smoke.sh
build/install/kaios-cli/bin/kaios tour --json

Evidence contract verification:

build/install/kaios-cli/bin/kaios capsule --file examples/evidence-sample/change-review.capsule.json --check
build/install/kaios-cli/bin/kaios replay --file examples/evidence-sample/change-review.capsule.json
build/install/kaios-cli/bin/kaios capsule --file examples/baseline-gate/capsules/baseline.capsule.json --check
build/install/kaios-cli/bin/kaios capsule --file examples/baseline-gate/capsules/current-different.capsule.json --check
build/install/kaios-cli/bin/kaios diff examples/baseline-gate/capsules/baseline.capsule.json examples/baseline-gate/capsules/current-different.capsule.json --check

kaios diff ... --check intentionally exits 1 for the baseline-gate sample because the gate detects valid changed runtime behavior.

Docker note: docker build --check . has been verified locally. Full Docker smoke depends on Docker Hub and Alpine package download speed in the local environment.

KAI OS v0.3.0 Evidence Core

KAI OS v0.3.0 makes Evidence Core the product center: agent work is now recoverable, schedulable, auditable, packageable, and CI-gateable without an API key.

Highlights:

• Process recovery evidence: runtime crashes are recorded as FAILED + RUNTIME_CRASH, recovered attempts get new PIDs, and kaios recover <run-id> --dry-run explains recovery status without mutating saved runs.
• Priority scheduler evidence: ready DAG nodes run by priority with stable ordering, event-triggered nodes wait for matching runtime events, and local worker execution records workerId.
• Syscall ledger: every tool call, including denied calls, records audit data with redacted arguments, duration, denied status, and estimated cost.
• Evidence JSON expansion: trace, capsule, review, evidence, and kaios ps --json expose scheduler, syscalls, cost, and recovery fields while keeping existing schema meanings stable.
• Config expansion: kaios.json supports priority, recovery, triggers, capabilities, executorHint, and memoryIsolation.

Try it:

curl -fsSL https://morning-verlu.github.io/KAI/install.sh | sh
export PATH="$HOME/.kaios/bin:$PATH"
kaios quickstart
kaios review
kaios evidence --summary

Verification:

• ./gradlew test installDist distZip distTar --no-daemon
• Release ZIP smoke: --version, quickstart --no-ci, review --json, ps --json, evidence --summary, recover --dry-run --json

KAI OS v0.1.84

KAI OS v0.1.84 hardens the product path with a read-only workspace compass.

Highlights:

• Added kaios next, a workspace-aware command that prints the single best next command.
• Added kaios.next/v1 JSON with status, action, fixFirst, signals, next, and nextActions for docs, support bots, CI helpers, and future UI surfaces.
• kaios next prioritizes repair, then Agent Gate verification, then process inspection once evidence is healthy.
• Updated README, install docs, launch kit, support docs, website, JSON contracts, and CLI help to make kaios next the first low-friction entry point.

Verification:

• ./gradlew :kaios-cli:test --tests ai.kaios.cli.KaiosCliSmokeTest --no-daemon
• ./gradlew test installDist distZip distTar --no-daemon
• scripts/quickstart-smoke.sh
• Manual kaios next smoke for empty, setup-only, and ready workspaces.

KAI OS v0.1.83

Product hardening\n\n- Added fixFirst to kaios.bug-report/v1 so support automation and future UI surfaces can read the first recovery action directly.\n- Added a ## Fix First section to Markdown bug reports before the detailed diagnostics.\n- For missing or invalid configs, bug reports now prioritize kaios doctor --fix --dry-run ...; for valid configs without evidence, they point to kaios gate --config ....\n- Updated support docs and JSON contracts for the bug-report first-action flow.\n\n## Verification\n\n- ./gradlew test installDist distZip distTar --no-daemon\n- scripts/quickstart-smoke.sh\n- strict bug-report fixFirst smoke\n- YAML and diff whitespace checks

KAI OS v0.1.82

Product hardening\n\n- Unified config failure recovery around kaios doctor --fix --dry-run ... so gate/verify summaries preview writes before suggesting repair commands.\n- Updated kaios.verify/v1 diagnosis.fixFirst to prefer the stable repair-project action for project config failures.\n- Kept kaios setup ... recovery commands in next for compatibility while making repair preview the first action.\n- Updated docs and smoke tests for missing and invalid config recovery paths.\n\n## Verification\n\n- ./gradlew test installDist distZip distTar --no-daemon\n- scripts/quickstart-smoke.sh\n- gate/verify invalid-config repair-preview smoke\n- YAML and diff whitespace checks

KAI OS v0.1.81

Product hardening\n\n- Added kaios doctor --fix to apply the safest local repair path for missing or invalid project workflow files.\n- Added kaios doctor --fix --dry-run and kaios.doctor-fix/v1 JSON so users and automation can preview config/CI writes before touching project files.\n- Kept existing config and CI files protected by default; invalid existing configs require an explicit --force repair.\n- Updated docs, JSON contracts, and smoke coverage for the repair workflow.\n\n## Verification\n\n- ./gradlew test installDist distZip distTar --no-daemon\n- scripts/quickstart-smoke.sh\n- doctor-fix dry-run, fix, and force-safety smoke checks\n- YAML and diff whitespace checks

KAI OS v0.1.80

What's Changed

• Added kaios quickstart --dry-run to preview generated files, commands, CI behavior, and evidence paths before writing anything.
• Added additive plan data to kaios.quickstart/v1 so docs, bots, and future UI surfaces can inspect onboarding writes without parsing terminal text.
• Updated quickstart docs and smoke tests so the no-key first-run path remains predictable.

Verification

• ./gradlew :kaios-cli:test --tests ai.kaios.cli.KaiosCliSmokeTest --no-daemon
• ./gradlew test installDist distZip distTar --no-daemon
• scripts/quickstart-smoke.sh
• Manual dry-run smoke: status=planned, no kaios.json, no workflow file, no runtime files written.

KAI OS v0.1.79

What's Changed

• Made invalid-config recovery commands directly executable: KAI OS now recommends kaios setup ... --force instead of a human-readable fix ... or rerun ... instruction.
• Added stable regenerate-config nextAction id for automation that wants a concrete repair command.
• Updated Agent Gate diagnosis so diagnosis.fixFirst points at the executable recovery command for invalid configs.

Verification

• ./gradlew :kaios-cli:test --tests ai.kaios.cli.KaiosCliSmokeTest --no-daemon
• ./gradlew test installDist distZip distTar --no-daemon
• scripts/quickstart-smoke.sh
• Manual invalid-config gate smoke: fix_first_id=regenerate-config, fix_first_command=kaios setup --ci --force, exit code 2.

KAI OS v0.1.78

What's Changed

• Added structured Agent Gate diagnosis to kaios.verify/v1: diagnosis.status, diagnosis.verdict, diagnosis.reasons, diagnosis.fixFirst, and diagnosis.diffChanges.
• Mirrored the PR-visible failure and baseline drift summary in JSON so CI bots, dashboards, and future Agent Desktop views do not need to parse Markdown.
• Kept the change additive and backward-compatible for existing kaios.verify/v1 consumers.

Verification

• ./gradlew :kaios-cli:test --tests ai.kaios.cli.KaiosCliSmokeTest --no-daemon
• ./gradlew test installDist distZip distTar --no-daemon
• scripts/quickstart-smoke.sh
• Manual baseline diff JSON smoke: diagnosis.status=different, fixFirst.id=diff-capsules, exit code 1.

KAI OS v0.1.77

Agent Gate product hardening release.\n\n- Added evidence.diff.changes so machine-readable evidence includes sampled stable runtime differences.\n- Agent Gate Step Summary now includes a What Changed table when --baseline ... --check detects behavior changes.\n- Evidence text output now prints sampled changes alongside the difference count.\n\nVerification: ./gradlew test installDist distZip distTar --no-daemon; scripts/quickstart-smoke.sh; YAML parse; manual baseline-diff gate --summary-out smoke.