Skip to content

daily-driver-current-status.md

Latest commit

 

History

History
205 lines (165 loc) · 17.1 KB

daily-driver-current-status.md

File metadata and controls

205 lines (165 loc) · 17.1 KB
Raw

Daily-Driver Current Status

As of 2026-06-18

Claim class: Current truth for daily-driver behavior (TUI, chat, agent mode, approval, cost, undo, resume).

Owns: What a daily user should trust today and which surface to choose.

Does not own: Roadmap horizon status (roadmap-status.md), ticket execution steps (plans/ticket-plans/index.md), or dated review evidence under docs/analysis/.

Review trigger: TUI, chat, agent mode, approval, cost, undo, or resume behavior changes. Last reviewed: 2026-06-25 (CG-16 test-harness only; no user-visible behavior change)

This page is the short daily-use entry point for TeaAgent's TUI, TUI chat, and agent mode. It is intentionally more practical than the audit corpus.

First run (safe)

Safe first command: teaagent chat (see "Recommended today" below for safe first task). Current status and operation guides: see the table below. On failure: Recovery And Continuity Guide.

Recommended roadmap: The complete daily-driver work plan is the master work plan — it ranks all active work by user value, risk reduction, feasibility, strategic leverage, and ROI. Start there when choosing what to implement or what to expect next.

Recommended today

Need Recommended surface Why
Conversational local coding with cost and undo visibility teaagent chat The REPL uses the shared chat controller for result display, cost accounting, and undo journal behavior.
Daily cockpit with setup, preflight, runs, and approvals teaagent tui --setup --root . The TUI is useful for status and operations, with unified cost tracking via ChatSessionController.
Non-interactive autonomous task teaagent agent run "<task>" Best when you want audit logs, approval gates, and a run summary without a live chat loop.
Resume/review a known run teaagent agent interactive-review <run_id> This is the currently reliable inspection path for suspended/background-style work.

Solid today

  • Approval governance, audit logging, plan-before-write gates, and run summaries remain the strongest parts of the project.
  • teaagent chat prints successful task answers and no longer marks successful tasks as failures.
  • teaagent chat /cost and /budget are wired to real session cost. Budget semantics are explicit: None means unlimited, while 0 is a real zero cap. Cost display labels whether the value is actual, estimated, or unavailable.
  • teaagent chat /undo uses the undo journal and preserves unrelated manual edits.
  • TUI setup, preflight, runs, session listing, and approval commands provide useful operational coverage.
  • TUI /cost now accumulates via ChatSessionController (CG-11 fixed).
  • TUI has adopted ChatSessionController for unified execution semantics (CG-12 fixed).
  • Exception swallowing removed from ChatSessionController (CG-13 fixed).
  • Failure-card matching has stopword filtering and relevance threshold (TASK-DD2-012 fixed).
  • Memory and run store corruption warnings surfaced in preflight/daily (TASK-DD2-011 fixed).
  • TUI ask/run/cost/undo/root/resume commands all delegate to ChatSessionController (P0-A-001). Headless command-path tests (22 tests in test_tui_command_path.py) verify each command goes through the controller.
  • TUI undo output explicitly labels fallback: "journal undo completed", "checkpoint restore completed", or "nothing to undo" (P0-A-003).
  • TUI help text includes a "TUI Command Reference" section documenting controller-backed command semantics (P0-A-004).

Document governance

Use these when you need the rules behind status, risk, or document ownership:

Current planning front door

Use plans/daily-driver-complete-work-plan-risk-roi-2026-06-04.md when choosing what to implement next. It ranks the active work by user value, risk reduction, feasibility, strategic leverage, and ROI.

Latest project-level cross-review

The newest project-level review layer fact-checks the broad "late-P0 / early-P1" assessment and turns it into Phase 0 trust-repair work:

Latest documentation-state package

The current documentation optimization pass adds a curated front door, a documentation-state review, critical questioning, an operating model, a master plan, and a work-item ledger:

Latest dynamic skill and long-result package

Dynamic skill generation is structurally supported but not yet proven reliable end-to-end for daily use. The RSS failure case shows the current gap: a skill can appear in a discoverable directory without proving reviewed install, activation, long-source preservation, script execution, or verified summary output.

Latest seven-control-loop strategy package

The June 5 competitor and research pass reframes TeaAgent's next architecture step as seven control loops: spec-first direction, dynamic workflow breadth, goal-loop depth, model-routing cost and quality control, synthesis review, precise memory, and human review gates. These documents are strategy and work planning evidence only; they do not prove the runtime already implements the full loop.

Latest community pain-point package

The June 5 community pass records current user pain around routing opacity, memory pollution, review cost, long-task drift, cost surprise, hook/permission confusion, skill/MCP supply-chain risk, overeager edits, and fake success. Treat these as roadmap pressure signals, not proof that every competitor has the same bug or that TeaAgent has already fixed the class.

Known issues

Issue Practical impact Tracking
Dynamic generated skills are not yet proven end-to-end. Treat generated skills as governed candidates until a run proves activation, source preservation, and output verification. Do not rely on RSS/WebSearch skill summaries as fully reliable yet. DSK-P0-001 through DSK-P0-004
Seven-control-loop package is not daily-driver runtime proof by itself. docs/roadmap-status.md owns SCL row status, while DR-006/backlog provenance keep competitor-derived SCL-P0 scheduling on hold unless owner friction evidence, governance-gap proof, or owner override promotes it. Treat completed SCL rows as control-model/docs evidence until runtime receipts prove daily behavior. SCL-P0-001 through SCL-P0-007; DR-006
Community pain-point docs are not mitigation proof by themselves. docs/roadmap-status.md owns CPP row status. Daily-driver claims still need implementation receipts and tests before survey-derived pain points can be claimed fixed in user-visible behavior. CPP-P0-001 through CPP-P0-008

Recently fixed

Fix What changed Tracking
Cost display now labels actual vs estimated vs unavailable vs unlimited. BudgetState includes explicit cost_state field; TUI /cost, /budget, run summary, and evidence bundle all use the same 4 cost states consistently. UI never implies actual cost when only an estimate is available. P0-B-001 through P0-B-003
Cost state propagated to all surfaces. RunEvidenceSummary, RunEvidenceBundle, run_summary.py, TUI /cost and /budget commands all show cost_state label with the canonical set: actual, estimated, unavailable, unlimited. P0-B-001
Cost accumulation tests added. tests/test_tui_cost.py covers 4 cost states, cost/budget consistency, multi-task accumulation, evidence bundle fields, and run summary cost_state. P0-B-002
Path-scoped approval without a path now fails closed. p no longer falls back to a global grant when no path can be extracted, and blank scoped patterns are rejected at the store boundary. DS-12 / approval UX
Budget semantics are explicit. None means unlimited, 0 is a real zero cap, and the TUI/CLI budget displays reflect that distinction. DS-13 / budget UX
Explicit --root no longer overwritten by saved TUI state. _load_tui_state condition was inverted (checked 'root' not in data instead of finding saved root). Root restoration now guarded by _root_explicit flag, set by CLI entry points via run_tui(). TASK-DD2-002
TUI undo now uses ChatSessionController.undo_last_run() with checkpoint fallback. TUI /undo first tries undo journal (file-level restore), falls back to git-stash checkpoint. CG-15 / TICKET-12
TUI cost display now reads from ChatSessionController session state (source of truth). _handle_cost uses controller.get_session_cost() with local fallback. CG-11 / TICKET-12
Exception swallowing removed from ChatSessionController. try/except (AttributeError, TypeError): pass blocks removed from execute_task. Fault-injection test added. CG-13 / TICKET-13
Redundant audit_trail field removed from suspension data. audit_trail key removed from suspend_to_background and reference in _agent.py commented out. CG-14 / TICKET-15
TUI /cost and budget display now show real session cost. TUI migrated to use ChatSessionController for unified cost tracking. Headless TUI path tests verify accumulation. TASK-DD2-003 / TASK-DD2-013
Failure-card matching has stopword filtering and relevance threshold. Matching requires 2+ significant words in common to avoid false positives from unrelated tasks. TASK-DD2-012
Memory and run store corruption warnings surfaced. health_report() methods track corrupt entries; preflight/daily show warnings for degraded state. TASK-DD2-011
Headless TUI path tests hardened. Tests now drive through actual command paths (cost, root, initial task, undo, approvals) rather than helper functions. TASK-DD2-013
REPL suspend→resume round-trip restored. suspend_to_background() records run_started before session_suspended, so teaagent agent resume <run_id> can recover the task; test_repl_suspend_resume_roundtrip covers the path. TICKET-16 Phase 2
TUI session clear now clears persisted chat messages. The command empties the active session's messages list, saves it, and reports an error when no active session exists. TUI session UX
Run evidence summaries surfaced in agent mode payload. run_evidence field added to agent run output with commands, tests, approvals, gaps. —
Updated daily-driver status docs. Removed stale known issues, added recently-fixed section. —
TUI/CLI semantic parity (P0-A). Headless command-path tests (22 tests) verify ask/run/cost/undo/root/resume delegate to ChatSessionController. TUI undo output labels journal vs checkpoint fallback. Help text includes controller-backed command reference. P0-A-001, P0-A-002, P0-A-003, P0-A-004

Do not rely on yet

  • Do not use teaagent agent run --background <run_id> to resume; it can treat the id as a new task argument.
  • Do not treat a successful docs-only check as proof that active runtime paths were tested.
  • Do not treat newly landed stop-gaps as release-ready until the active command path is tested.
  • Do not treat a loaded skill as proof that the skill was used or that the final artifact was source-backed.
  • Do not treat the seven-control-loop strategy or work-item package as runtime proof, or as permission to schedule legacy-competitive work without the DR-006 gates: owner friction evidence, governance-gap proof, or owner override.
  • Do not treat community pain-point docs as mitigation evidence. Use docs/roadmap-status.md for row state, then require implementation/test receipts before claiming daily-driver behavior is fixed.

2026-06-25 review note

CG-16 test-harness work (TUI de-mock slices, test_tui.py doctor GraphQLite stub, runner tool-decision split) does not change user-visible daily-driver behavior. Evidence: pytest tests/test_tui*.py tests/tui/ tests/test_tui_command_path.py -q.

Read next