ltcai 9.5.0 → 9.6.0

package/README.md

@@ -1,9 +1,17 @@
 # Lattice AI
 
-**Lattice AI 9.5.0 is the local-first Digital Brain platform. The Command Center puts everything one keystroke away: a Cmd+K palette that searches knowledge, past conversations, automations, and app pages in one query, and a daily briefing that condenses recent knowledge, automation state, pending reviews, Brain health, and suggested next steps into one glance with one-click quick actions.**
+**Lattice AI 9.6.0 is the local-first Digital Brain platform. The Trusted Agent Loop makes autonomous work transparent and git-like: the reasoning loop reports every model call, format repair, and tool outcome; a deterministic evaluation harness gates every release on weak-model robustness; and change governance is proposal-first — creating new files is frictionless while edits and deletions of existing content become reviewable proposals you approve or reject before anything changes.**
 
 **Lattice AI는 모델이 바뀌어도 내 지식과 맥락을 보존하는 로컬 우선 AI 브레인입니다.**
 
+> The 9.6.0 release engineers trust into the agent loop: a structured
+> LoopTrace makes every run observable (llm calls, parse recoveries, format
+> repairs, corrections, tool outcomes — returned as `loop` in the agent API),
+> weak-model tolerance gains python-literal repair and escalating format
+> corrections, a deterministic agent evaluation harness gates CI, and a
+> central change governor stages edits/deletions of existing files as review
+> proposals (`/api/proposals`) applied exactly as reviewed — while new-file
+> creation now runs without approval friction.
 > The 9.5.0 release adds the Command Center: a deterministic, local
 > `/api/command` surface serving a daily briefing (recent knowledge,
 > conversation activity, automation state, pending reviews, health snapshot,
@@ -108,11 +116,11 @@ You need Lattice AI when:
 The screenshots below are the latest checked-in visual evidence captures. They
 keep the first-run Brain flow, memory graph, source capture, model library,
 system view, admin console, and review center visible as release gates while
-9.5.0 puts the whole Brain one keystroke away: the Cmd+K Command Palette
-searches knowledge, conversations, automations, and pages in one query, and
-Today's Briefing condenses the Brain's state into one glance with one-click
-next steps. The captures below are the checked-in 9.5.0 visual release
-evidence for that product flow.
+9.6.0 makes autonomous work trustworthy: agents create new things freely,
+but every change to existing content becomes a reviewable proposal with a
+diff, and the reasoning loop itself is observable and regression-gated. The
+captures below are the checked-in 9.6.0 visual release evidence for that
+product flow.
 
 ### 1. Wake Brain
 
@@ -124,21 +132,21 @@ confirm owner, check the computer, choose the Brain voice.
 Choose the owner of the Brain. The profile is not a SaaS account by default; it
 is the local identity for the knowledge you keep.
 
-![Login](output/release/v9.5.0/screenshots/01-login.png)
+![Login](output/release/v9.6.0/screenshots/01-login.png)
 
 ### 3. Recommended Models
 
 Start with a short list: safest recommendation, faster model, stronger model.
 Advanced details stay available without overwhelming first-time users.
 
-![Recommended Models](output/release/v9.5.0/screenshots/02-recommended-models.png)
+![Recommended Models](output/release/v9.6.0/screenshots/02-recommended-models.png)
 
 ### 4. Install And Load
 
 Download and load only after consent. Lattice explains model size, local
 execution, and network use before work starts.
 
-![Install and Load](output/release/v9.5.0/screenshots/03-install-load-progress.png)
+![Install and Load](output/release/v9.6.0/screenshots/03-install-load-progress.png)
 
 ### 5. Brain Chat
 
@@ -150,14 +158,14 @@ and its visible life signal follow real listening, recall, synthesis, and action
 state. Detailed memory rings, provenance, conversation history, and
 model/runtime proof open as overlays only when requested.
 
-![One-viewport Living Brain Home](output/release/v9.5.0/screenshots/04-brain-chat-home.png)
+![One-viewport Living Brain Home](output/release/v9.6.0/screenshots/04-brain-chat-home.png)
 
 ### 6. Review Center
 
 Automation results are staged for review before they become durable decisions.
 Snooze, unsnooze, run now, approve, and dismiss actions stay explicit.
 
-![Review Center](output/release/v9.5.0/screenshots/12-review-center.png)
+![Review Center](output/release/v9.6.0/screenshots/12-review-center.png)
 
 ## Brain Depths
 
@@ -173,10 +181,10 @@ The user travels inward from everyday memory to deeper structure:
 
 Walkthrough:
 
-![v9.5.0 Living Brain walkthrough](output/release/v9.5.0/gifs/v9.5.0-living-brain-walkthrough.gif)
+![v9.6.0 Living Brain walkthrough](output/release/v9.6.0/gifs/v9.6.0-living-brain-walkthrough.gif)
 
 Screenshot index and capture notes:
-[output/release/v9.5.0/SCREENSHOT_INDEX.md](output/release/v9.5.0/SCREENSHOT_INDEX.md)
+[output/release/v9.6.0/SCREENSHOT_INDEX.md](output/release/v9.6.0/SCREENSHOT_INDEX.md)
 
 ## Install
 
@@ -261,45 +269,50 @@ See [docs/DEVELOPMENT.md](docs/DEVELOPMENT.md) for developer workflow details.
 
 ## Current Release
 
-The current release is **9.5.0 — Command Center**:
-
-- A new `CommandCenterService` (`/api/command/briefing`, `/api/command/search`)
-  aggregates every Brain surface read-only and deterministically: recent
-  knowledge from the scoped graph, conversation activity, automation
-  enable/draft counts, pending review items, a Brain-health snapshot, and the
-  top automation suggestions — each section degrades independently when a
-  backend is unavailable.
-- State-derived quick actions with stable ids ("N items waiting for review",
-  "enable your draft automations", "connect a knowledge folder") give the UI
-  one-click jumps that always reflect the actual product state.
-- Universal search groups results across knowledge nodes (scoped keyword
-  search), the user's own conversations (deduped per conversation, newest
-  first), and installed automations — all scoped to the requesting user and
-  workspace.
-- The frontend gains a Cmd+K Command Palette (grouped results, keyboard
-  navigation, page jumps, fully ko/en localized) and a Today's Briefing panel
-  on the Brain home with stat chips, recent knowledge, and quick actions.
-- Everything is read-only, local, and model-free: no model calls, no writes,
-  no external actions.
-
-Expected artifacts for 9.5.0 release must use exact filenames:
-
-- `dist/ltcai-9.5.0-py3-none-any.whl`
-- `dist/ltcai-9.5.0.tar.gz`
-- `ltcai-9.5.0.tgz`
-- `dist/ltcai-9.5.0.vsix`
-- `src-tauri/target/release/bundle/dmg/Lattice AI_9.5.0_aarch64.dmg`
+The current release is **9.6.0 — Trusted Agent Loop**:
+
+- The single-agent reasoning loop is fully observable: a structured
+  `LoopTrace` records every llm call, parse error (recovered or not), format
+  repair, correction, tool outcome, retry, and rollback, and its summary is
+  returned as `loop` with every agent API response.
+- Weak-model robustness is stronger and measured: the action parser now also
+  repairs Python-literal dicts (single quotes, True/False/None), repairs are
+  reported by name, and a second formatting slip escalates the correction
+  with the exact list of valid tools.
+- A deterministic agent evaluation harness (`scripts/agent_eval.py`, CI
+  gate) drives the real state machine through scripted scenarios — happy
+  path, weak-model format gauntlet, correction escalation, destructive
+  block, loop detection, critic retry, unrecoverable garbage — and must stay
+  at 100% scenario pass.
+- Change governance is proposal-first (`latticeai/core/tool_governor.py`):
+  additive creates run with minimal friction, while agent edits/deletions of
+  existing files are staged as review proposals with a unified diff and a
+  small/large tier, listed at `GET /api/proposals`, applied exactly as
+  reviewed on approve, discarded on reject. The Brain home shows a
+  "변경 제안 / Change proposals" panel with diff previews and one-click
+  approve/reject.
+- Housekeeping: Ruff per-file lint ignores trimmed from 9 entries to 3,
+  AGENTS.md now carries a machine-checked current-release marker inside the
+  docs gate, and the review queue gained the `change_proposal` source.
+
+Expected artifacts for 9.6.0 release must use exact filenames:
+
+- `dist/ltcai-9.6.0-py3-none-any.whl`
+- `dist/ltcai-9.6.0.tar.gz`
+- `ltcai-9.6.0.tgz`
+- `dist/ltcai-9.6.0.vsix`
+- `src-tauri/target/release/bundle/dmg/Lattice AI_9.6.0_aarch64.dmg`
 
 Do not use wildcard artifact uploads. Package registry publishing remains owner-run.
 
 See [docs/ROADMAP_RECOMMENDATIONS.md](docs/ROADMAP_RECOMMENDATIONS.md) for the
-strategic roadmap slices applied through 9.5.0 and the follow-up tracks.
+strategic roadmap slices applied through 9.6.0 and the follow-up tracks.
 
 ## Known Limitations
 
 - External package registries are owner-published and can lag behind GitHub.
 - PostgreSQL/pgvector is optional scale/migration tooling. SQLite remains the
-  live local Brain store in 9.5.0.
+  live local Brain store in 9.6.0.
 - Docker, model downloads, cloud model calls, Telegram, Brain Network, and update
   checks require explicit user action.
 - Conversation does not fabricate answers when no model is loaded.
@@ -311,6 +324,7 @@ strategic roadmap slices applied through 9.5.0 and the follow-up tracks.
 
 | Version | Theme |
 | --- | --- |
+| 9.6.0 | Trusted Agent Loop: observable agent reasoning (LoopTrace + loop API payload), python-literal/fence/think weak-model repairs with escalating corrections, a deterministic CI agent-eval harness, and proposal-first change governance where edits/deletions of existing files become reviewable diffs |
 | 9.5.0 | Command Center: a Cmd+K palette searching knowledge, conversations, automations, and pages in one query, plus a daily briefing with Brain state at a glance and state-derived one-click quick actions |
 | 9.4.0 | Question-Driven Everyday Automation: the Brain mines recurring user questions and connected knowledge folders into one-click, consent-first automation suggestions installed as review-gated drafts |
 | 9.3.0 | Proactive Brain Intelligence: the Brain diagnoses its own health, surfaces contradictions and stale knowledge, proposes consent-first duplicate consolidation, and recalls with hybrid lexical+semantic evidence behind an honest quality gate |

package/docs/CHANGELOG.md

@@ -4,6 +4,38 @@ The top entry is either the current unreleased main-branch work or the current
 release line. Older entries are historical and may describe behavior as it
 existed at that release.
 
+## [9.6.0] - 2026-07-20
+
+### Added
+- Added `latticeai/core/agent_trace.py` (`LoopTrace`): typed observability
+  for the single-agent reasoning loop, returned as `loop` with every agent
+  API response.
+- Added `latticeai/core/agent_eval.py` + `scripts/agent_eval.py`: a
+  deterministic scripted-model evaluation harness over the real agent state
+  machine, wired into CI as a release gate.
+- Added `latticeai/core/tool_governor.py` and
+  `latticeai/services/change_proposals.py` + `/api/proposals` router:
+  central read/additive/mutation/destructive classification with
+  proposal-first governance — edits/deletions of existing files are staged
+  as review-queue proposals (source `change_proposal`) with unified diffs
+  and applied exactly as reviewed on approval.
+- Added the "변경 제안 / Change proposals" Brain home panel with diff
+  previews and one-click approve/reject.
+
+### Changed
+- The action parser tolerates Python-literal dicts and reports every repair
+  by name; repeated formatting slips escalate the correction with the valid
+  tool list.
+- Additive file creates in the agent loop run without plan-approval
+  friction; plan approval no longer hard-blocks governor-managed tools.
+- Ruff per-file lint ignores trimmed from 9 entries to 3.
+
+### Security
+- Mutations and deletions of existing workspace files by the agent can no
+  longer apply silently: they stage as reviewable proposals, and approve
+  applies the exact reviewed content with full audit events.
+- AGENTS.md is now inside the machine-checked release documentation gate.
+
 ## [9.5.0] - 2026-07-20
 
 ### Added

package/docs/COMMUNITY_AND_PLUGINS.md

@@ -1,13 +1,13 @@
 # Community And Plugins
 
-Current release: **9.5.0 — Command Center**.
+Current release: **9.6.0 — Trusted Agent Loop**.
 
 LatticeAI defines the path from a strong local-first framework (8.4.0
 action-aware baseline, 8.5.0 registry+DI hardening, 8.6.0 capture/navigation
 reliability, 8.7.0 runtime-state hygiene, 8.8.0 Brain Core extraction
 readiness, 8.9.0 scoped Tool API hardening, 9.0.0 cleanup closure, 9.1.0
 fail-closed review completion, 9.2.0 model-agnostic file generation, 9.3.0
-proactive Brain intelligence, 9.4.0 question-driven everyday automation, and 9.5.0 Command Center)
+proactive Brain intelligence, 9.4.0 question-driven everyday automation, 9.5.0 Command Center, and 9.6.0 Trusted Agent Loop)
 to a product ecosystem. The
 immediate goal is small and practical: make it clear how
 contributors can extend the Brain without weakening local-first trust,

package/docs/DEVELOPMENT.md

@@ -1,10 +1,10 @@
 # Lattice AI Development
 
-Current release: **9.5.0 — Command Center**.
+Current release: **9.6.0 — Trusted Agent Loop**.
 
 This document is for contributors working on the local-first Digital Brain
 codebase. Product positioning and quick start stay in `README.md`; release
-history is intentionally limited to 8.0.0-9.5.0 in `docs/CHANGELOG.md` and
+history is intentionally limited to 8.0.0-9.6.0 in `docs/CHANGELOG.md` and
 `RELEASE.md`.
 
 ## Product Contract
@@ -153,10 +153,10 @@ For user-facing, API, runtime, release, or packaging changes, check:
 Release/publish examples must use exact target-version filenames. Do not
 document wildcard artifact upload commands.
 
-For 9.5.0 release work, exact artifacts are:
+For 9.6.0 release work, exact artifacts are:
 
-- `dist/ltcai-9.5.0-py3-none-any.whl`
-- `dist/ltcai-9.5.0.tar.gz`
-- `ltcai-9.5.0.tgz`
-- `dist/ltcai-9.5.0.vsix`
-- `src-tauri/target/release/bundle/dmg/Lattice AI_9.5.0_aarch64.dmg`
+- `dist/ltcai-9.6.0-py3-none-any.whl`
+- `dist/ltcai-9.6.0.tar.gz`
+- `ltcai-9.6.0.tgz`
+- `dist/ltcai-9.6.0.vsix`
+- `src-tauri/target/release/bundle/dmg/Lattice AI_9.6.0_aarch64.dmg`

package/docs/LEGACY_COMPATIBILITY.md

@@ -1,6 +1,6 @@
 # Legacy Compatibility Map
 
-Current target: **9.5.0 — Command Center**.
+Current target: **9.6.0 — Trusted Agent Loop**.
 
 Lattice AI is moving toward a smaller, modular architecture centered on
 `lattice_brain`, `latticeai.services`, `latticeai.api`, and `latticeai.runtime`.

package/docs/ONBOARDING.md

@@ -1,6 +1,6 @@
 # Lattice AI Onboarding
 
-Current release: **9.5.0 — Command Center**.
+Current release: **9.6.0 — Trusted Agent Loop**.
 
 The first-run goal is a five-minute path from "I opened the app" to "my Brain
 has a source, a question, and proof." This page is the product contract behind
@@ -40,7 +40,7 @@ read the docs first.
 
 ## Release Gate
 
-9.5.0 treats onboarding as a release gate, not marketing copy. The current
+9.6.0 treats onboarding as a release gate, not marketing copy. The current
 machine-checkable product readiness report requires this five-minute contract,
 the Brain Home surface, setup helpers, graph ingestion tests, and exact release
 artifact documentation before the release can be called complete.

package/docs/OPERATIONS.md

@@ -1,4 +1,4 @@
-# Lattice AI — Operations Guide (v9.5.0)
+# Lattice AI — Operations Guide (v9.6.0)
 
 ## 1. 데이터 파일 위치
 

package/docs/TRUST_MODEL.md

@@ -1,6 +1,6 @@
 # Lattice AI Trust Model
 
-Current release: **9.5.0 — Command Center**.
+Current release: **9.6.0 — Trusted Agent Loop**.
 
 Lattice AI is local-first, explicit about external communication, and honest
 when a capability is unavailable.

package/docs/WHY_LATTICE.md

@@ -1,6 +1,6 @@
 # Why Lattice AI Exists
 
-Current release: **9.5.0 — Command Center**.
+Current release: **9.6.0 — Trusted Agent Loop**.
 
 **Lattice AI is a local-first Digital Brain that keeps your knowledge durable
 across any AI model.**

package/docs/kg-schema.md

@@ -1,6 +1,6 @@
 # Knowledge Graph Schema
 
-Current release: **9.5.0 — Command Center**.
+Current release: **9.6.0 — Trusted Agent Loop**.
 
 명세 출처: `lattice_ai_full_spec.pptx` 슬라이드 20·21·22
 구현: `kg_schema.py`

package/docs/mcp-tools.md

@@ -1,4 +1,4 @@
-# MCP 도구 카탈로그 (v9.5.0)
+# MCP 도구 카탈로그 (v9.6.0)
 
 Lattice AI는 MCP(Model Context Protocol) 서버로 동작하여 Claude Desktop, Cursor 등에서 직접 도구를 사용할 수 있습니다.
 

package/lattice_brain/__init__.py

@@ -26,7 +26,7 @@ from .storage import (
     storage_from_env,
 )
 
-__version__ = "9.5.0"
+__version__ = "9.6.0"
 
 __all__ = [
     "AgentRuntime",

package/lattice_brain/runtime/multi_agent.py

@@ -21,7 +21,7 @@ from .contracts import multi_agent_contract
 from ..utils import now_iso as _now
 
 
-MULTI_AGENT_VERSION = "9.5.0"
+MULTI_AGENT_VERSION = "9.6.0"
 
 AGENT_ROLES = ("researcher", "planner", "executor", "reviewer", "release")
 CORE_PIPELINE = ("planner", "executor", "reviewer")

package/latticeai/__init__.py

@@ -1,3 +1,3 @@
 """Lattice AI - modular server package."""
 
-__version__ = "9.5.0"
+__version__ = "9.6.0"

package/latticeai/api/change_proposals.py

@@ -0,0 +1,60 @@
+"""Change proposal API router (v9.6.0).
+
+Read/approve/reject surface over
+:class:`~latticeai.services.change_proposals.ChangeProposalService`. The
+items live in the shared review queue (source ``change_proposal``), so the
+Act review center shows them too; this router adds the proposal-specific
+semantics: **approve applies the staged content exactly as reviewed**,
+reject discards it, and nothing touches disk while a proposal is pending.
+"""
+
+from __future__ import annotations
+
+from typing import Callable, Optional
+
+from fastapi import APIRouter, HTTPException, Request
+
+from latticeai.services.change_proposals import ChangeProposalService
+
+
+def create_change_proposals_router(
+    *,
+    service: ChangeProposalService,
+    require_user: Callable[[Request], str],
+    gate_read: Callable[[Request], Optional[str]],
+    gate_write: Callable[[Request], Optional[str]],
+) -> APIRouter:
+    router = APIRouter()
+
+    @router.get("/api/proposals")
+    async def list_proposals(request: Request):
+        user = require_user(request)
+        scope = gate_read(request)
+        return service.pending(user_email=user, workspace_id=scope)
+
+    @router.post("/api/proposals/{item_id}/approve")
+    async def approve_proposal(item_id: str, request: Request):
+        user = require_user(request)
+        scope = gate_write(request)
+        try:
+            return service.approve_and_apply(
+                item_id, user_email=user, workspace_id=scope
+            )
+        except (KeyError, FileNotFoundError) as exc:
+            raise HTTPException(status_code=404, detail=str(exc))
+        except ValueError as exc:
+            raise HTTPException(status_code=400, detail=str(exc))
+
+    @router.post("/api/proposals/{item_id}/reject")
+    async def reject_proposal(item_id: str, request: Request):
+        user = require_user(request)
+        scope = gate_write(request)
+        try:
+            return service.reject(item_id, user_email=user, workspace_id=scope)
+        except (KeyError, FileNotFoundError) as exc:
+            raise HTTPException(status_code=404, detail=str(exc))
+
+    return router
+
+
+__all__ = ["create_change_proposals_router"]

package/latticeai/api/chat_agent_http.py

@@ -229,6 +229,7 @@ class AgentHTTPController:
                 "planning_model": req.planning_model or self.model_router.current_model_id,
                 "executing_model": req.executing_model or self.model_router.current_model_id,
                 "reviewing_model": req.reviewing_model or self.model_router.current_model_id,
+                "loop": ctx.trace.summary(),
             }
 
         self.runtime.approve(ctx, current_user, approved_by_human=False)
@@ -306,6 +307,7 @@ class AgentHTTPController:
             "state_history": ctx.state_history,
             "final_state": ctx.state.value,
             "created_files": collect_created_files(ctx.transcript),
+            "loop": ctx.trace.summary(),
         }
 
     async def resume(

package/latticeai/app_factory.py

@@ -169,6 +169,9 @@ def _build(config: "Optional[Config]" = None) -> Dict[str, Any]:
     from latticeai.api.brain_intelligence import create_brain_intelligence_router
     from latticeai.api.command_center import create_command_center_router
     from latticeai.services.command_center import CommandCenterService
+    from latticeai.api.change_proposals import create_change_proposals_router
+    from latticeai.services.change_proposals import ChangeProposalService
+    from latticeai.tools import resolve_workspace_path
     from latticeai.api.memory import create_memory_router
     from latticeai.api.browser import create_browser_router
     from latticeai.api.portability import create_portability_router
@@ -1051,6 +1054,23 @@ def _build(config: "Optional[Config]" = None) -> Dict[str, Any]:
             gate_read=PLATFORM.gate_read,
         )
     )
+    CHANGE_PROPOSALS = ChangeProposalService(
+        review_queue=REVIEW_QUEUE,
+        resolve_path=resolve_workspace_path,
+        audit=append_audit_event,
+    )
+    # Proposal-first mutations: the agent loop consults the governor so
+    # additive creates run with minimal friction while changes/deletions of
+    # existing files are staged for review instead of applied.
+    CHAT_AGENT_RUNTIME.deps.change_governor = CHANGE_PROPOSALS
+    app.include_router(
+        create_change_proposals_router(
+            service=CHANGE_PROPOSALS,
+            require_user=require_user,
+            gate_read=PLATFORM.gate_read,
+            gate_write=PLATFORM.gate_write,
+        )
+    )
 
 
     # ── Health & Info ──────────────────────────────────────────────────────────────

package/latticeai/cli/entrypoint.py

@@ -157,7 +157,9 @@ def _start_tunnel(port: int) -> str | None:
     log_path = Path.home() / ".latticeai" / "tunnel.log"
     log_path.parent.mkdir(parents=True, exist_ok=True)
 
-    proc = subprocess.Popen(
+    # Detached on purpose: the tunnel outlives this helper; the log file is
+    # the observable surface.
+    subprocess.Popen(
         [bin_path, "tunnel", "--url", f"http://localhost:{port}"],
         stdout=open(log_path, "w"),
         stderr=subprocess.STDOUT,