nsaspy/zara
1
0
Fork 0
Code Issues 95 Pull requests 18 Projects Releases Packages Wiki Activity Actions

feat(runtime): typed slots and principal/conversation-scoped clarification dialogue (#155) #284

Closed
nsaspy wants to merge 0 commits from rage/155-typed-slots into master
pull from: rage/155-typed-slots
merge into: nsaspy:master
Conversation 0 Commits - Files changed -
nsaspy commented 2026-09-04 22:44:10 +00:00
Owner
Copy link

Summary

Implements #155 against the frozen IntentFrame v1 contract (#154, docs/intentframe-contract.md), per the design in rage/155-clarification-dialogue.org (D1-D12, adversarial review, A2 test-matrix row).

  • zara/runtime/frames.py — IntentFrame v1 Python mirror: closed typed slot value set (text/number/duration/datetime/ref/boolean), slot origins, frame statuses. Frames are pure values and carry no envelope metadata (frame_id/supersedes live in session bookkeeping only).
  • zara/runtime/clarification.py — ClarificationCoordinator: one open session per (principal, conversation); deterministic question order from required slots; follow-ups fill only the active requested slot; focused retry after invalid answers with bounded attempts; corrections ("actually five minutes") rotate the frame id and record supersedes; cancellation phrases close cancelled; TTL expiry on an injectable virtual clock (no sleeps); bounded session count / value size / lifetime; uniform byte-identical stale rejection wording (no cross-principal inference); drop_all for typed degraded restart/shutdown.
  • RuntimeHost owns a coordinator (D2) and drops open sessions on restart/shutdown.
  • Wake pending path: pending(open)/pending(text)/pending(python(schedule_todo)) open dialogue templates and answer with the deterministic question; follow-ups are routed to the open session before the command gate; completion executes via the existing execute_intent/python_skills paths and closes the session. Unknown pending shapes keep the legacy prompt (bounded migration).
  • Non-goals honored: no provider execution (frames are handed back, not executed by the machine), no Android, no second ContextManager, no Prolog dynamic state.

Closes #155

Test plan

  • 26 machine-level state-machine tests (mandatory examples: timer ask/fill, "bananas" focused retry, cancel-then-late-answer not consumed, correction 20->5 with supersedes, open->app, text sarah->message, deterministic two-slot order, A/B principal isolation, two conversations, stale uniformity across expired/cancelled/restarted/none, double-hold guard, cancel-vs-fill race, capacity bound, virtual-clock timeout, attempts bound, oversized values, ambiguity choice, supersede-on-new-dialogue, frame-envelope split, typed validation).
  • 9 wake wire-up tests (pending -> question -> fill -> "Executed: ...", never-mind, retry, two-slot text flow, schedule_todo skill, command supersede, agent fallback unchanged, legacy prompt for unknown shapes, execution failure closes cleanly).
  • 3 RuntimeHost ownership tests (coordinator exposed; restart/shutdown drop sessions with typed superseded state).
  • Non-interactive dialogue script walking every mandatory example in one deterministic conversation.
  • Migrated fixtures in test_command_routing.py / test_latency_metrics.py carry the coordinator; the pending-prompt test now asserts the new required behavior.
  • Full local gate at candidate head: scripts/test-all.sh 10/10 phases, nix flake check green, nix build green.
## Summary Implements #155 against the frozen IntentFrame v1 contract (#154, docs/intentframe-contract.md), per the design in `rage/155-clarification-dialogue.org` (D1-D12, adversarial review, A2 test-matrix row). - `zara/runtime/frames.py` — IntentFrame v1 Python mirror: closed typed slot value set (text/number/duration/datetime/ref/boolean), slot origins, frame statuses. Frames are pure values and carry **no** envelope metadata (frame_id/supersedes live in session bookkeeping only). - `zara/runtime/clarification.py` — ClarificationCoordinator: one open session per (principal, conversation); deterministic question order from required slots; follow-ups fill only the active requested slot; focused retry after invalid answers with bounded attempts; corrections ("actually five minutes") rotate the frame id and record `supersedes`; cancellation phrases close cancelled; TTL expiry on an injectable virtual clock (no sleeps); bounded session count / value size / lifetime; uniform byte-identical stale rejection wording (no cross-principal inference); `drop_all` for typed degraded restart/shutdown. - `RuntimeHost` owns a coordinator (D2) and drops open sessions on restart/shutdown. - Wake pending path: `pending(open)`/`pending(text)`/`pending(python(schedule_todo))` open dialogue templates and answer with the deterministic question; follow-ups are routed to the open session before the command gate; completion executes via the existing `execute_intent`/`python_skills` paths and closes the session. Unknown pending shapes keep the legacy prompt (bounded migration). - Non-goals honored: no provider execution (frames are handed back, not executed by the machine), no Android, no second ContextManager, no Prolog dynamic state. Closes #155 ## Test plan - 26 machine-level state-machine tests (mandatory examples: timer ask/fill, "bananas" focused retry, cancel-then-late-answer not consumed, correction 20->5 with supersedes, open->app, text sarah->message, deterministic two-slot order, A/B principal isolation, two conversations, stale uniformity across expired/cancelled/restarted/none, double-hold guard, cancel-vs-fill race, capacity bound, virtual-clock timeout, attempts bound, oversized values, ambiguity choice, supersede-on-new-dialogue, frame-envelope split, typed validation). - 9 wake wire-up tests (pending -> question -> fill -> "Executed: ...", never-mind, retry, two-slot text flow, schedule_todo skill, command supersede, agent fallback unchanged, legacy prompt for unknown shapes, execution failure closes cleanly). - 3 RuntimeHost ownership tests (coordinator exposed; restart/shutdown drop sessions with typed superseded state). - Non-interactive dialogue script walking every mandatory example in one deterministic conversation. - Migrated fixtures in test_command_routing.py / test_latency_metrics.py carry the coordinator; the pending-prompt test now asserts the new required behavior. - Full local gate at candidate head: `scripts/test-all.sh` 10/10 phases, `nix flake check` green, `nix build` green.
nsaspy closed this pull request 2026-09-04 23:09:00 +00:00
Some checks failed
CI / test (pull_request) Failing after 9s
Details
CI / android skeleton gate (pull_request) Failing after 5s
Details
CI / shared mic / Arch Linux (pull_request) Failing after 10m41s
Details
CI / shared mic / Ubuntu 24.04 (pull_request) Failing after 13m15s
Details

Pull request closed

Please reopen this pull request to perform a merge.
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels   
bug
Something isn't working

  
documentation
Improvements or additions to documentation

  
duplicate
This issue or pull request already exists

  
enhancement
New feature or request

  
good first issue
Good for newcomers

  
help wanted
Extra attention is needed

  
invalid
This doesn't seem right

  
question
Further information is requested

  
wontfix
This will not be worked on

No labels
bug
documentation
duplicate
enhancement
good first issue
help wanted
invalid
question
wontfix
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
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
nsaspy/zara!284
No description provided.