lhumina_code/home
0
0
Fork 0
Code Issues 34 Pull requests Projects Releases Packages Wiki Activity Actions

[ROADMAP] Phase ordering for ACTIVE board (project 13) #221

New issue
Open
opened 2026-05-06 17:45:03 +00:00 by mik-tf · 0 comments
mik-tf commented 2026-05-06 17:45:03 +00:00
Owner
Copy link

Vision

Hero OS becomes a sovereign AI-native desktop where the AI agent drives every domain (CRM, files, voice, library, calendar) through one client, one broker, one persistence layer, one theme — all reproducibly installable on any machine.

Deploy-anywhere = P1 + P2

The install code path (service_<svc> install --download --reset) is the same whether a developer or an end-user runs it. P1 produces the CI artifacts; P2 makes the install scripts honest on a fresh Ubuntu. After P2 closes, "deploy Hero OS to any VM" is a real capability, not a coping strategy. P3-P6 tighten the architecture (one persistence, one AI client, one theme, one agent) but don't change deployability.

home#121 (shared dev VM) stays useful as a convenience for new dev onboarding after P2, but it stops being the only path.

Six phases, in order

P1 — Binaries (in flight, gates everything)

Close home#212.

  • Bucket C (7 nutools-wired repos with no published release): hero_lib_rhai, hero_agent, hero_collab, hero_office, hero_planner, hero_logic, hero_mail
  • Bucket D (4 off-radar repos with old-shape naming): hero_compute, hero_ledger, hero_sync, hero_tfspores

Gate: every active lhumina_code repo publishes target-triple-named CI artifacts. service_<svc> install --download --reset works for every service.

Result: Foundation under everything else. Install becomes reproducible at the CI artifact level.

P2 — Devs unblocked (T1 blockers)

After P1 is at least 80% — engineering team can't ship while these are open.

Current state (workaround): home#121 — everyone works on the shared dev VM because individual machines aren't reproducible. P2 closes the gap so any machine works; #121 stays as the optional convenience, not the only path.

Gate: OSIS runs on a fresh dev box; service_install_all is green-on-clean-Ubuntu in CI.

Result: Any fresh Ubuntu — dev workstation, demo VM, or end-user deploy — runs the install and gets a working Hero OS. Same install code path serves both audiences. "Team out of sync" goes away; deploy-anywhere becomes real.

P3 — One foundation story (T2 + T4) — parallel to P4

Gate: zero direct redis usage in the workspace; light/dark toggle in hero_os propagates to every iframe + island + admin UI.

Result: One persistence story. One theme story.

P4 — AI client adoption (T3) — parallel to P3

home#219 META +

Gate: no app instantiates a non-hero_lib LLM client; aibroker downloads models for every provider; voice is one drop-in skill.

Result: Every Hero app talks to AI the same way. The "AI making up stuff" / "AI not responding" classes of bugs end. Voice works in 5 minutes for any new app.

P5 — CRM + Python flow shipping (T5 — the demo)

Starts when P2 + P4 are green.

Gate: agent drives CRM end-to-end on herodemo with real per-context data, multi-company contacts, and no hallucinations.

Result (the killer demo): ask the AI to add a contact in OWH context, link them to two companies with roles, and the right Person + PersonCompanyLink records appear in the right OSIS context.

P6 — Polish + growth (continuous, no hard gate)

Result: the v1 vision — ambient AI desktop with reusable surfaces.

Parallelism

  • P1 must complete first. Everything below depends on a known-good --download install.
  • P2 must complete before P5. No demo can stand on a broken OSIS or non-reproducible install.
  • P3 ⊥ P4 — fully parallel. Different repos, different owners, no shared files.
  • P5 starts when P2 + P4 are green.
  • P6 runs continuously — file under each phase as polish becomes practical.

Final state — six things true at once

  1. One install path. service_<svc> install --download --reset works on a fresh Ubuntu for every service.
  2. One AI path. app → hero_lib → hero_aibroker → provider. No app rolls its own LLM client. Voice is one skill anyone drops in.
  3. One persistence path. hero_db is the only persistent store. Encrypted at rest by default.
  4. One theme path. Toggle light/dark in hero_os, every surface follows. Theme lives in hero_proc secrets, not env.sh.
  5. One agent path. service_agent_v3 drives every Hero domain through OpenRPC + Python flows. CRM data is real per-context. The agent grounds on docs_hero and live OSIS state, not hallucinations.
  6. One demo. herodemo.gent01.grid.tf reflects all of the above. The 24-hour killer demo plan (hero_demo#53) becomes credible.

Out of scope for this roadmap

Tracked separately, not gated by these phases:

  • TFTC customer onboarding — needs separate discovery.
  • "Demo fronts" alignment — meeting agenda, not engineering work.
  • L-05 (start --reset --download cargo purge) — parallel to P1, tracked at hero_demo#64.
  • D-05 UX gates (voice WS e2e, embedder semantic-search e2e) — fall under P4.
  • Phase 1 docs_hero content — own track at docs_hero#1.

How to use this issue

  • New work goes into the right phase here AND on project 13.
  • "Are we ready for P3?" → check the gate of P2; close any unmet items first.
  • When a phase closes, update this issue with ~~Phase X — closed YYYY-MM-DD~~ and start the next.
  • This issue stays pinned in project 13's description for visibility.

Source: meeting notes 2026-05-06 + recap session.

Signed-off-by: mik-tf

## Vision Hero OS becomes a sovereign AI-native desktop where the AI agent drives every domain (CRM, files, voice, library, calendar) through one client, one broker, one persistence layer, one theme — all reproducibly installable on any machine. ## Deploy-anywhere = P1 + P2 The install code path (`service_<svc> install --download --reset`) is the same whether a developer or an end-user runs it. P1 produces the CI artifacts; P2 makes the install scripts honest on a fresh Ubuntu. After P2 closes, "deploy Hero OS to any VM" is a real capability, not a coping strategy. P3-P6 tighten the architecture (one persistence, one AI client, one theme, one agent) but don't change deployability. [home#121](https://forge.ourworld.tf/lhumina_code/home/issues/121) (shared dev VM) stays useful as a *convenience* for new dev onboarding after P2, but it stops being the *only* path. ## Six phases, in order ### P1 — Binaries (in flight, gates everything) Close [home#212](https://forge.ourworld.tf/lhumina_code/home/issues/212). - **Bucket C** (7 nutools-wired repos with no published release): `hero_lib_rhai`, `hero_agent`, `hero_collab`, `hero_office`, `hero_planner`, `hero_logic`, `hero_mail` - **Bucket D** (4 off-radar repos with old-shape naming): `hero_compute`, `hero_ledger`, `hero_sync`, `hero_tfspores` **Gate:** every active `lhumina_code` repo publishes target-triple-named CI artifacts. `service_<svc> install --download --reset` works for every service. **Result:** Foundation under everything else. Install becomes reproducible at the CI artifact level. --- ### P2 — Devs unblocked (T1 blockers) After P1 is at least 80% — engineering team can't ship while these are open. **Current state (workaround):** [home#121](https://forge.ourworld.tf/lhumina_code/home/issues/121) — everyone works on the shared dev VM because individual machines aren't reproducible. P2 closes the gap so any machine works; #121 stays as the optional convenience, not the only path. - [hero_osis#38](https://forge.ourworld.tf/lhumina_code/hero_osis/issues/38) — fresh OSIS install has no Context records - [home#216](https://forge.ourworld.tf/lhumina_code/home/issues/216) — service install reproducibility - [home#217](https://forge.ourworld.tf/lhumina_code/home/issues/217) — branch cleanup ritual - [hero_skills#218](https://forge.ourworld.tf/lhumina_code/hero_skills/issues/218) — install scripts surface errors loudly **Gate:** OSIS runs on a fresh dev box; `service_install_all` is green-on-clean-Ubuntu in CI. **Result:** Any fresh Ubuntu — dev workstation, demo VM, or end-user deploy — runs the install and gets a working Hero OS. Same install code path serves both audiences. "Team out of sync" goes away; deploy-anywhere becomes real. --- ### P3 — One foundation story (T2 + T4) — parallel to P4 - **T2 hero_db migration:** [home#218](https://forge.ourworld.tf/lhumina_code/home/issues/218) META + per-service children - **T4 theme system:** [hero_os#128](https://forge.ourworld.tf/lhumina_code/hero_os/issues/128), [hero_proc#94](https://forge.ourworld.tf/lhumina_code/hero_proc/issues/94), [hero_proc#96](https://forge.ourworld.tf/lhumina_code/hero_proc/issues/96) **Gate:** zero direct redis usage in the workspace; light/dark toggle in hero_os propagates to every iframe + island + admin UI. **Result:** One persistence story. One theme story. --- ### P4 — AI client adoption (T3) — parallel to P3 [home#219 META](https://forge.ourworld.tf/lhumina_code/home/issues/219) + - [hero_aibroker#62](https://forge.ourworld.tf/lhumina_code/hero_aibroker/issues/62) — downloads models everywhere - [hero_aibroker#55](https://forge.ourworld.tf/lhumina_code/hero_aibroker/issues/55) — fail-fast / fallback chain - [hero_skills#219](https://forge.ourworld.tf/lhumina_code/hero_skills/issues/219) — voice integration skill canonical - [hero_voice#28](https://forge.ourworld.tf/lhumina_code/hero_voice/issues/28) — switch to parakeet - [hero_lib#134](https://forge.ourworld.tf/lhumina_code/hero_lib/issues/134) — fix CI in the process - [hero_agent#18](https://forge.ourworld.tf/lhumina_code/hero_agent/issues/18) — audit Timur MCP/Python work - [home#215](https://forge.ourworld.tf/lhumina_code/home/issues/215) — assistant non-functional (resolves at broker layer via #55) **Gate:** no app instantiates a non-`hero_lib` LLM client; aibroker downloads models for every provider; voice is one drop-in skill. **Result:** Every Hero app talks to AI the same way. The "AI making up stuff" / "AI not responding" classes of bugs end. Voice works in 5 minutes for any new app. --- ### P5 — CRM + Python flow shipping (T5 — the demo) Starts when P2 + P4 are green. - [hero_logic#23](https://forge.ourworld.tf/lhumina_code/hero_logic/issues/23) — e2e validate Phase 11 on demo VM (D + E merged 2026-05-06; e2e remains) - [hero_biz#37](https://forge.ourworld.tf/lhumina_code/hero_biz/issues/37) + [hero_rpc#42](https://forge.ourworld.tf/lhumina_code/hero_rpc/issues/42) — per-context routing (same root cause) - [hero_biz#38](https://forge.ourworld.tf/lhumina_code/hero_biz/issues/38) — many-to-many contact-company - [hero_biz#15](https://forge.ourworld.tf/lhumina_code/hero_biz/issues/15) — CRM AiClient through broker (child of P4) - [hero_biz#39](https://forge.ourworld.tf/lhumina_code/hero_biz/issues/39) — AI hallucinating CRM data — root cause - [hero_skills#220](https://forge.ourworld.tf/lhumina_code/hero_skills/issues/220) — `service_biz --demo` flag **Gate:** agent drives CRM end-to-end on herodemo with real per-context data, multi-company contacts, and no hallucinations. **Result (the killer demo):** ask the AI to add a contact in OWH context, link them to two companies with roles, and the right `Person` + `PersonCompanyLink` records appear in the right OSIS context. --- ### P6 — Polish + growth (continuous, no hard gate) - **T6:** [home#220](https://forge.ourworld.tf/lhumina_code/home/issues/220) (ONNX consolidation), [hero_office#25](https://forge.ourworld.tf/lhumina_code/hero_office/issues/25) (Scott) - **T7:** [hero_os#129](https://forge.ourworld.tf/lhumina_code/hero_os/issues/129) (multi-tab), [hero_proc#95](https://forge.ourworld.tf/lhumina_code/hero_proc/issues/95) (auto-start db), [hero_skills PR#215](https://forge.ourworld.tf/lhumina_code/hero_skills/pulls/215) (skill audit), [hero_memory#1](https://forge.ourworld.tf/lhumina_code/hero_memory/issues/1) **Result:** the v1 vision — ambient AI desktop with reusable surfaces. --- ## Parallelism - **P1 must complete first.** Everything below depends on a known-good `--download` install. - **P2 must complete before P5.** No demo can stand on a broken OSIS or non-reproducible install. - **P3 ⊥ P4** — fully parallel. Different repos, different owners, no shared files. - **P5 starts when P2 + P4 are green.** - **P6 runs continuously** — file under each phase as polish becomes practical. ## Final state — six things true at once 1. **One install path.** `service_<svc> install --download --reset` works on a fresh Ubuntu for every service. 2. **One AI path.** `app → hero_lib → hero_aibroker → provider`. No app rolls its own LLM client. Voice is one skill anyone drops in. 3. **One persistence path.** `hero_db` is the only persistent store. Encrypted at rest by default. 4. **One theme path.** Toggle light/dark in hero_os, every surface follows. Theme lives in `hero_proc` secrets, not `env.sh`. 5. **One agent path.** `service_agent_v3` drives every Hero domain through OpenRPC + Python flows. CRM data is real per-context. The agent grounds on `docs_hero` and live OSIS state, not hallucinations. 6. **One demo.** [herodemo.gent01.grid.tf](https://herodemo.gent01.grid.tf/) reflects all of the above. The 24-hour killer demo plan ([hero_demo#53](https://forge.ourworld.tf/lhumina_code/hero_demo/issues/53)) becomes credible. ## Out of scope for this roadmap Tracked separately, not gated by these phases: - TFTC customer onboarding — needs separate discovery. - "Demo fronts" alignment — meeting agenda, not engineering work. - L-05 (`start --reset --download` cargo purge) — parallel to P1, tracked at [hero_demo#64](https://forge.ourworld.tf/lhumina_code/hero_demo/issues/64). - D-05 UX gates (voice WS e2e, embedder semantic-search e2e) — fall under P4. - Phase 1 docs_hero content — own track at [docs_hero#1](https://forge.ourworld.tf/lhumina_code/docs_hero/issues/1). ## How to use this issue - New work goes into the right phase here AND on project 13. - "Are we ready for P3?" → check the gate of P2; close any unmet items first. - When a phase closes, update this issue with `~~Phase X — closed YYYY-MM-DD~~` and start the next. - This issue stays pinned in project 13's description for visibility. Source: meeting notes 2026-05-06 + recap session. Signed-off-by: mik-tf
mik-tf added this to the ACTIVE project 2026-05-06 17:45:05 +00:00
Sign in to join this conversation.
No Branch/Tag specified
Branches Tags
development
main
No results found.
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
ACTIVE
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
lhumina_code/home#221
No description provided.