v0.0.10 shipped on 2026-05-23: Task Worlds landed earlier that day then was ripped out in favor of tighter execution state and compensation wiring.
Current alpha is v0.1.2.
v0.0.10
Changelog milestone (alpha)
Historical
Alpha slice
22
Eval scenario packs
2
UIs (TUI + Web)
Shipped in this slice
liminalCLI (dev) —setup,doctor,web,tui,update,path- Install scripts:
scripts/install.sh,scripts/install.ps1(alpha onboarding — not a GA release) - Recipe library v2 — compounds on strategy shape, not generic “recipe:” notes
- Harness power pass — execution state, compensation playback,
breakdown, tool DAGdepends_on, contract→tool mapping - Workflow evals + additional core baseline tests
- Web fetch — Wikipedia extracts, pagination, distill bypass, model
max_charsfloor - Persona UI themes — distinct deterministic palettes, hue-preserving normalization, HudShell theme wiring
- Closed-loop tuning hooks, per-harness workspace root, streaming edit previews in web
- OpenRouter attribution unified across sidecar calls
Removed (same day)
What you can run today
Clone the repo, set AGENT_API_KEY in .env, build core + tools, then:
npm run tui
# or
npm run web
Optional: run the install script or liminal setup after building.
Alpha capability map (cumulative)
| Area | Status on main |
|---|---|
| Harness | ReAct loop, retries, circuit breaker, compression, approvals, optional self-heal lint |
| Writes | Streaming large-file writes + integrity checks |
| Tools | Files, shell, git, code intel, web, browser + CAPTCHA, memory, vault, documents, orchestration, dynamic tools, MCP |
| Reasoning | Intent routing, two-tier models, external think / reason, effort learning |
| Persona | Bootstrap, generated themes, four web shells |
| Mission state | Execution state + epistemic subgoals (not Task Worlds) |
| Observability | JSONL traces, self_telemetry, recipe stats |
| Eval | 22 packs → .agent_eval_runs/ sink |
Comparison table
Roadmap (not shipped)
| Stage | Meaning | Status |
|---|---|---|
| v0.0.x alpha | Rapid iteration on main | Superseded — see v0.1.2 |
| Beta | API/default stability pass, broader dogfooding | Not started |
| RC | Release checklist, docs freeze candidate | Not started |
| v0.1.0 | Public preview tag + npm/install GA story | Not started |
Comparison table
Pre-release history
| Version | Date | Summary |
|---|---|---|
| v0.0.1 | 2026-05-03 | Monorepo + first loop |
| v0.0.2 | 2026-05-03 | Persona generation |
| v0.0.3 | 2026-05-05 | Lazy tools + hybrid recall |
| v0.0.4 | 2026-05-08 | Vision, markets, doc engine |
| v0.0.5 | 2026-05-09 | Upgrade V / DAG |
| v0.0.6 | 2026-05-10 | MCP + research tiers |
| v0.0.7 | 2026-05-12 | JSONL + world context |
| v0.0.8 | 2026-05-15 | Settings + persona UX |
| v0.0.9 | 2026-05-22 | Browser + reasoning + writes |
| v0.0.10 | 2026-05-23 | Execution state pass + dev CLI |
| v0.1.2 | 2026-05-26 | Research ledger, identity memory, web chat chrome (current) |
| v0.0.11 | 2026-05-24 | Storage split, multi-chat, audio, connectors |
Comparison table
Still out of scope for alpha
- Hosted team plane or cloud sessions
- VS Code / JetBrains extensions
- npm global package publish
- Official Docker image
- Task Worlds (removed; no current replacement product surface)
Try the alpha on your machine
Clone from GitHub. FSL-1.1-MIT. Node 22+ and your own API key. Beta and public preview coming later.