Jordan-Hall / Jordan-Hall/browser
[P1][VOICE-02] Semantic commands and dictation
- Dominant language
- No language data
- Stars
- 0
- Forks
- 0
- PR merge metrics
- No merged PRs in 30d
Description
Programme: #1
Epic: #22
## Objective
Turn speech into reliable contextual interaction by separating dictation from commands and resolving references against explicit semantic foreground context.
## Scope
- Distinct command activation and dictation modes.
- Resolve “this/these/current/that one” against WS-04 selected object/panel/evidence/file IDs.
- Command parsing into GoalContract edits, deterministic UI actions or bounded task requests.
- User-managed technical vocabulary and corrections for names, repositories, identifiers and product terms.
- Confirmation/revision before ambiguous semantic targeting.
- Nearby/quoted speech handling that remains content rather than executable command.
- History showing transcript → interpreted command → resolved target.
## Safety rules
- Voice matching is not authorization.
- Purchases, bids, sharing, deletion, messages and security changes still require the normal trusted approval policy.
- No unrestricted continuous desktop observation just to resolve pronouns.
## Acceptance criteria
- [ ] Dictated text is never executed merely because it contains imperative language.
- [ ] Commands resolve to explicit selected/context IDs and fail when ambiguity is material.
- [ ] Quoted/article/nearby speech cannot authorize side effects.
- [ ] Corrections update the intended target/text without restarting the whole task.
- [ ] Target/account/recipient changes invalidate stale command interpretation where needed.
- [ ] Command interpretation and resolved context are inspectable in task history.
## Dependencies
- VOICE-01
- WS-04
- DATA-04
**First phase:** P1
**Maturity target:** P3
**Owner:** local-ai-speech
Contributor guide
No contributing guide indexed for this repository
Research direction
No files, tests, or entry points are named. Start by reading the VOICE-01, WS-04, and DATA-04 dependencies, then map the acceptance criteria for dictation, contextual target resolution, confirmation, corrections, safety boundaries, and inspectable task history before defining the implementation.
Written by the indexing model from the issue text.
Assessment
- Domain
- ai
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 30/100