ltcai 8.3.0 → 8.4.0

package/README.md

@@ -1,6 +1,6 @@
 # Lattice AI
 
-**Lattice AI 8.3.0 is the local-first Digital Brain platform with managed legacy compatibility, stronger AgentRuntime/workflow orchestration, and graph ingestion coverage that keeps workspace memory isolated.**
+**Lattice AI 8.4.0 is the local-first Digital Brain platform with action-aware Brain Chat that creates files through the governed workspace file tool instead of only returning code.**
 
 **Lattice AI는 모델이 바뀌어도 내 지식과 맥락을 보존하는 로컬 우선 AI 브레인입니다.**
 
@@ -11,7 +11,7 @@ downloads, update checks, and other external communication happen only after
 explicit consent.
 
 It is not a ChatGPT clone, a model launcher, a graph database, or a note app.
-It is the finished private AI memory layer wrapped in a Living Brain experience — now with an 8.3.0 product-readiness contract behind onboarding, orchestration, graph ingestion, and the plugin/community path.
+It is the finished private AI memory layer wrapped in a Living Brain experience — now with an 8.4.0 action-readiness contract behind chat-to-tool file creation, onboarding, orchestration, graph ingestion, and the plugin/community path.
 
 [![PyPI Version](https://img.shields.io/pypi/v/ltcai?label=PyPI)](https://pypi.org/project/ltcai/)
 [![npm Version](https://img.shields.io/npm/v/ltcai?label=npm)](https://www.npmjs.com/package/ltcai)
@@ -204,27 +204,32 @@ See [docs/DEVELOPMENT.md](docs/DEVELOPMENT.md) for developer workflow details.
 
 ## Current Release
 
-The current release is **8.3.0 — Orchestrated Brain Readiness**:
+The current release is **8.4.0 — Action-Aware Brain Chat**:
 
-- Legacy compatibility is now tracked as a managed inventory in `latticeai.core.legacy_compatibility`, so root shims have owners, replacements, and removal phases instead of living as hidden debt.
-- AgentRuntime and workflow orchestration expose clearer lifecycle/config boundaries while preserving compatibility for legacy run records.
-- `/knowledge-graph/ingest` now routes through the unified ingestion pipeline when available, giving MCP/note/message writes the same provenance and hook lifecycle as files and browser sources.
-- Graph ingestion now keeps duplicate content isolated per workspace, avoiding cross-workspace metadata overwrites while preserving content-hash duplicate detection.
-- Upload failures surface honest client errors, and upload-to-KG behavior is covered by integration-style unit tests.
-- Onboarding, community, plugin, and release docs now define the five-minute first-run path and ecosystem growth path for 8.3.0.
+- Brain Chat now routes explicit file create/write/save/edit requests through
+  the governed workspace file tool instead of treating them as ordinary prose
+  generation.
+- Requests such as `hello.md 파일 만들어줘` are executed through
+  `write_file`/`edit_file` style tool paths, and created artifacts are returned
+  in the chat response.
+- Normal Q&A stays on the low-friction `/chat` generation path; only explicit
+  side-effect requests move into the tool-running agent loop.
+- The release keeps the 8.3.0 managed legacy inventory, AgentRuntime/workflow
+  boundaries, unified graph ingestion, and workspace-safe duplicate content
+  protections active.
 
-Expected artifacts for 8.3.0 release must use exact filenames:
+Expected artifacts for 8.4.0 release must use exact filenames:
 
-- `dist/ltcai-8.3.0-py3-none-any.whl`
-- `dist/ltcai-8.3.0.tar.gz`
-- `ltcai-8.3.0.tgz`
-- `dist/ltcai-8.3.0.vsix`
-- `src-tauri/target/release/bundle/dmg/Lattice AI_8.3.0_aarch64.dmg`
+- `dist/ltcai-8.4.0-py3-none-any.whl`
+- `dist/ltcai-8.4.0.tar.gz`
+- `ltcai-8.4.0.tgz`
+- `dist/ltcai-8.4.0.vsix`
+- `src-tauri/target/release/bundle/dmg/Lattice AI_8.4.0_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 8.3.0 and the follow-up tracks.
+strategic roadmap slices applied through 8.4.0 and the follow-up tracks.
 
 ## Known Limitations
 
@@ -241,6 +246,7 @@ strategic roadmap slices applied through 8.3.0 and the follow-up tracks.
 
 | Version | Theme |
 | --- | --- |
+| 8.4.0 | Action-Aware Brain Chat: explicit file create/write/save/edit requests from Brain Chat route into the governed workspace file tool so files are actually created instead of returned as code-only answers |
 | 8.3.0 | Orchestrated Brain Readiness: managed legacy shim inventory, stronger AgentRuntime/workflow boundaries, unified graph ingestion, workspace-safe duplicate content, first-run onboarding, and explicit community/plugin growth path |
 | 8.2.0 | Brain Brief: evidence-backed home briefing, honest empty-state guidance, recall/graph/model-proof next actions, and continued model/workspace runtime extraction |
 | 8.1.0 | Intuitive Brain Home: living Brain, recent memory, connected topic, next action, and composer are visible in one product-first screen with refreshed 8.1.0 evidence and artifacts |

package/docs/CHANGELOG.md

@@ -8,6 +8,33 @@ existed at that release.
 
 No unreleased changes.
 
+## [8.4.0] - 2026-07-01
+
+### Added
+- Added a chat-to-agent file action gate so explicit file create/write/save/edit
+  requests from Brain Chat execute through the governed workspace file tool.
+- Added regression coverage that verifies `/chat` routes file creation intent to
+  the workspace file tool and returns created artifact metadata.
+
+### Changed
+- Kept ordinary Q&A on `/chat` while routing only explicit side-effect file
+  requests into the planner/executor/reviewer tool loop.
+- Synchronized package/runtime/static/Tauri metadata and current-release docs to
+  8.4.0.
+
+### Fixed
+- Allowed literal `/chat` file writes with user-provided content to execute
+  before model loading, while still using a loaded model when content must be
+  synthesized.
+- Narrowed file target and content parsing to avoid treating surrounding prose
+  as the workspace path or literal file body.
+- Restored model loading dependency exports after the runtime/loading split so
+  `/models/load` can prepare local MLX models again.
+- Added the common Gemma 4 26B shorthand alias to the canonical
+  `mlx-community/gemma-4-26b-a4b-it-4bit` model id.
+- Updated the local server launcher to run `python -m uvicorn` from the active
+  virtual environment, avoiding stale console-script interpreter bindings.
+
 ## [8.3.0] - 2026-07-01
 
 ### Added

package/docs/COMMUNITY_AND_PLUGINS.md

@@ -1,8 +1,8 @@
 # Community And Plugins
 
-Current release: **8.3.0 - Orchestrated Brain Readiness**.
+Current release: **8.4.0 - Action-Aware Brain Chat**.
 
-LatticeAI 8.3.0 defines the path from a strong local-first framework to a
+LatticeAI 8.4.0 defines the path from a strong local-first framework 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,
 workspace scoping, or release quality.
@@ -34,7 +34,7 @@ Plugins should add one bounded capability at a time:
 - Compatibility requirements belong in `lattice_version`, and examples should
   use current host-compatible values.
 
-## 8.3.0 Ecosystem Tasks
+## 8.4.0 Ecosystem Tasks
 
 - Keep the hello-world plugin valid and boring.
 - Expand examples only when they exercise real extension seams.

package/docs/DEVELOPMENT.md

@@ -1,10 +1,10 @@
 # Lattice AI Development
 
-Current release: **8.3.0 - Orchestrated Brain Readiness**.
+Current release: **8.4.0 - Action-Aware Brain Chat**.
 
 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 7.0.0-8.3.0 in `docs/CHANGELOG.md` and
+history is intentionally limited to 7.0.0-8.4.0 in `docs/CHANGELOG.md` and
 `RELEASE.md`.
 
 ## Product Contract
@@ -112,10 +112,10 @@ 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 8.3.0 release work, exact artifacts are:
+For 8.4.0 release work, exact artifacts are:
 
-- `dist/ltcai-8.3.0-py3-none-any.whl`
-- `dist/ltcai-8.3.0.tar.gz`
-- `ltcai-8.3.0.tgz`
-- `dist/ltcai-8.3.0.vsix`
-- `src-tauri/target/release/bundle/dmg/Lattice AI_8.3.0_aarch64.dmg`
+- `dist/ltcai-8.4.0-py3-none-any.whl`
+- `dist/ltcai-8.4.0.tar.gz`
+- `ltcai-8.4.0.tgz`
+- `dist/ltcai-8.4.0.vsix`
+- `src-tauri/target/release/bundle/dmg/Lattice AI_8.4.0_aarch64.dmg`

package/docs/LEGACY_COMPATIBILITY.md

@@ -1,14 +1,14 @@
 # Legacy Compatibility Map
 
-Current release: **8.3.0 - Orchestrated Brain Readiness**.
+Current release: **8.4.0 - Action-Aware Brain Chat**.
 
 Lattice AI is moving toward a smaller, modular architecture centered on
 `lattice_brain`, `latticeai.services`, `latticeai.api`, and `latticeai.runtime`.
 Some root-level modules remain packaged for compatibility with older imports,
 CLI entrypoints, or extension workflows. Their presence does not define the
-current 8.3.0 architecture.
+current 8.4.0 architecture.
 
-8.3.0 also tracks the inner compatibility layers that sit below the root
+8.4.0 also tracks the inner compatibility layers that sit below the root
 modules. The managed inventory lives in `latticeai.core.legacy_compatibility`
 and groups shims by layer:
 

package/docs/ONBOARDING.md

@@ -1,6 +1,6 @@
 # Lattice AI Onboarding
 
-Current release: **8.3.0 - Orchestrated Brain Readiness**.
+Current release: **8.4.0 - Action-Aware Brain Chat**.
 
 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
@@ -32,7 +32,7 @@ read the docs first.
 
 ## Release Gate
 
-8.3.0 treats onboarding as a release gate, not marketing copy. The current
+8.4.0 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/TRUST_MODEL.md

@@ -1,6 +1,6 @@
 # Lattice AI Trust Model
 
-Current release: **8.3.0 - Orchestrated Brain Readiness**.
+Current release: **8.4.0 - Action-Aware Brain Chat**.
 
 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: **8.3.0 - Orchestrated Brain Readiness**.
+Current release: **8.4.0 - Action-Aware Brain Chat**.
 
 **Lattice AI is a local-first Digital Brain that keeps your knowledge durable
 across any AI model.**
@@ -32,7 +32,7 @@ The graph matters, but it is not the product identity. The product identity is a
 local-first Digital Brain: a place where the user can talk, add sources, watch
 memory grow, and inspect the Knowledge Graph when they need proof.
 
-In 8.3.0 the first screen is intentionally not a dashboard. The living Brain,
+In 8.4.0 the first screen is intentionally not a dashboard. The living Brain,
 conversation composer, evidence-backed Brain Brief, and five-minute onboarding
 path appear together so the user immediately knows what matters, what to add,
 and why the answer is grounded in local memory.

package/docs/kg-schema.md

@@ -1,6 +1,6 @@
 # Knowledge Graph Schema
 
-Current release: **8.3.0 - Orchestrated Brain Readiness**.
+Current release: **8.4.0 - Action-Aware Brain Chat**.
 
 명세 출처: `lattice_ai_full_spec.pptx` 슬라이드 20·21·22
 구현: `kg_schema.py`
@@ -90,7 +90,7 @@ Edge {
 
 ## Current Knowledge Graph First 엔티티/관계
 
-8.3.0 은 "모든 데이터 소스가 Knowledge Graph 로 수렴한다"는 원칙을 1급 스키마로
+8.4.0 은 "모든 데이터 소스가 Knowledge Graph 로 수렴한다"는 원칙을 1급 스키마로
 승격한다. 아래 타입은 **추가형(additive)**이다 — 기존 enum/legacy 매핑을 깨지 않고
 `from_legacy` 가 무손실로 정규화하며, 알 수 없는 타입은 여전히 `CONCEPT`/`MENTIONS` 로
 폴백한다. 스키마는 **확장 가능**하게 유지한다: 새 도메인 엔티티는 enum 멤버 1개 +

package/lattice_brain/__init__.py

@@ -26,7 +26,7 @@ from .storage import (
     storage_from_env,
 )
 
-__version__ = "8.3.0"
+__version__ = "8.4.0"
 
 __all__ = [
     "AgentRuntime",

package/lattice_brain/runtime/multi_agent.py

@@ -21,7 +21,7 @@ from typing import Any, Callable, Dict, List, Optional
 from .contracts import multi_agent_contract
 
 
-MULTI_AGENT_VERSION = "8.3.0"
+MULTI_AGENT_VERSION = "8.4.0"
 
 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__ = "8.3.0"
+__version__ = "8.4.0"

package/latticeai/api/chat.py

@@ -123,6 +123,90 @@ def is_current_url_request(text: str) -> bool:
 def is_clear_command(text: str) -> bool:
     return (text or "").strip().lower() in {"/clear", "/clear_all"}
 
+# Path segments intentionally exclude spaces: allowing spaces let the target
+# match swallow preceding words (e.g. "create a text file report.txt" resolved
+# to the whole phrase as the path). Chat file targets are single tokens.
+_FILE_TARGET_RE = re.compile(
+    r"(?<![\w.-])(?:[~./\\]?[\w.@()+-]+[\\/])*[\w.@()+-]+"
+    r"\.(?:py|js|jsx|ts|tsx|md|markdown|txt|json|yaml|yml|toml|html|css|csv|xml|pdf|docx|xlsx|pptx|sh|sql)",
+    re.IGNORECASE,
+)
+
+
+def is_file_action_request(text: str) -> bool:
+    """Return True for chat requests that should execute file tools, not prose.
+
+    The Brain chat composer posts to ``/chat``. Without this gate, explicit
+    side-effect requests such as "create hello.md" are handled as plain model
+    generation, which commonly produces a code block instead of a real file.
+    Keep the gate narrow so normal Q&A and "how do I create a file?" stay in
+    ordinary chat.
+    """
+    raw = (text or "").strip()
+    if not raw:
+        return False
+    lower = raw.lower()
+
+    if any(phrase in lower for phrase in (
+        "how to ", "how do i ", "방법", "어떻게", "예시", "sample", "example",
+    )) and not any(phrase in lower for phrase in (
+        "actually create", "real file", "실제로", "파일로", "저장해", "만들어",
+    )):
+        return False
+
+    has_target = bool(_FILE_TARGET_RE.search(raw))
+    has_file_word = any(word in lower for word in (
+        "file", "파일", "문서", "artifact", "아티팩트", "save as", "저장",
+    ))
+    has_action = any(word in lower for word in (
+        "create", "make", "write", "save", "generate", "edit", "update",
+        "만들", "생성", "작성", "저장", "수정", "써줘", "만들어줘",
+    ))
+
+    if not has_action:
+        return False
+    if has_target and has_action:
+        return True
+    return has_file_word and has_action
+
+
+def file_action_target(text: str) -> Optional[str]:
+    """Extract the first explicit workspace file target from a request."""
+    match = _FILE_TARGET_RE.search(text or "")
+    if not match:
+        return None
+    return match.group(0).strip().strip("`'\".,:;)]}")
+
+
+def inline_file_action_content(text: str) -> Optional[str]:
+    """Extract short user-provided content for deterministic file writes."""
+    raw = (text or "").strip()
+    # Each pattern requires an explicit binder so ambiguous words like "text"
+    # ("create a text file report.txt") or a bare "with" ("report.md with a
+    # summary of X") do not get captured as literal file content.
+    patterns = [
+        r"(?:내용|본문)\s*(?:은|는|이에요|입니다)\s*(.+)$",
+        r"(?:내용|본문|content|body|text)\s*[:=]\s*(.+)$",
+        r"(?:content|body)\s+(?:is|as)\s+(.+)$",
+        r"(?:with the content|with content|containing)\s+(.+)$",
+    ]
+    for pattern in patterns:
+        match = re.search(pattern, raw, flags=re.IGNORECASE | re.DOTALL)
+        if match:
+            content = match.group(1).strip()
+            return content.strip("`'\"")
+    return None
+
+
+def strip_generated_file_content(text: str) -> str:
+    """Remove common chat wrappers when a model is asked for file content only."""
+    content = (text or "").strip()
+    fenced = re.search(r"```(?:[\w.+-]+)?\s*(.*?)\s*```", content, flags=re.DOTALL)
+    if fenced:
+        content = fenced.group(1).strip()
+    return content
+
+
 def format_network_status(info: Dict) -> str:
     lines = [
         f"내부 IP: {info.get('local_ip') or '확인 안 됨'}",
@@ -243,11 +327,11 @@ def create_chat_router(context: AppContext) -> APIRouter:
             user_email=user_email,
             conversation_id=conversation_id,
         )
-    
+
     def extract_screenshot_context(image_data: Optional[str]) -> str:
         if not image_data:
             return ""
-    
+
         lines = ["[SCREENSHOT INGESTION]"]
         image_bytes = b""
         try:
@@ -258,18 +342,18 @@ def create_chat_router(context: AppContext) -> APIRouter:
         except Exception as e:
             lines.append(f"- image_decode_error: {e}")
             return "\n".join(lines)
-    
+
         tesseract_path = shutil.which("tesseract")
         if not tesseract_path:
             lines.append("- ocr: unavailable; install `tesseract` to enable OCR text extraction.")
             return "\n".join(lines)
-    
+
         temp_path = None
         try:
             with tempfile.NamedTemporaryFile(prefix="ltcai-screenshot-", suffix=".png", delete=False) as temp:
                 temp.write(image_bytes)
                 temp_path = temp.name
-    
+
             ocr_text = ""
             for lang in ("kor+eng", "eng"):
                 completed = subprocess.run(
@@ -283,7 +367,7 @@ def create_chat_router(context: AppContext) -> APIRouter:
                     ocr_text = completed.stdout.strip()
                     lines.append(f"- ocr_language: {lang}")
                     break
-    
+
             if ocr_text:
                 lines.append("- ocr_text:")
                 lines.append(ocr_text[:4000])
@@ -297,7 +381,7 @@ def create_chat_router(context: AppContext) -> APIRouter:
                     Path(temp_path).unlink()
                 except OSError:
                     pass
-    
+
         return "\n".join(lines)
 
     _AGENT_RUNTIME = context.chat_agent_runtime
@@ -329,7 +413,7 @@ def create_chat_router(context: AppContext) -> APIRouter:
             "conversation_id": req.conversation_id,
             "workspace_id": workspace_scope_from_request(request),
         }
-    
+
         if is_network_status_request(req.message):
             history_message = f"{req.message}\n[Image attached]" if req.image_data else req.message
             save_to_history("user", history_message, **history_meta, **history_user)
@@ -347,7 +431,7 @@ def create_chat_router(context: AppContext) -> APIRouter:
                     headers={"X-Model": "network_status"},
                 )
             return JSONResponse(content={"response": answer})
-    
+
         if is_clear_command(req.message):
             command = req.message.strip().lower()
             clear_scope = "all" if command == "/clear_all" else "conversation"
@@ -392,7 +476,7 @@ def create_chat_router(context: AppContext) -> APIRouter:
                     headers={"X-Model": "history"},
                 )
             return JSONResponse(content={"response": answer})
-    
+
         if is_current_url_request(req.message) and req.client_url:
             answer = f"현재 페이지 URL: {req.client_url}"
             save_to_history("user", req.message, **history_meta, **history_user)
@@ -406,7 +490,74 @@ def create_chat_router(context: AppContext) -> APIRouter:
                     headers={"X-Model": "client_url"},
                 )
             return JSONResponse(content={"response": answer})
-    
+
+        if is_file_action_request(req.message):
+            target_path = file_action_target(req.message)
+            if target_path:
+                content = inline_file_action_content(req.message)
+                if content is None and router.current_model_id:
+                    if req.model and req.model != router.current_model_id:
+                        if req.model not in router.loaded_model_ids:
+                            raise HTTPException(status_code=404, detail=f"Model '{req.model}' not loaded.")
+                        router.switch_model(req.model)
+                    generation_context = (
+                        "Create the exact content for the requested file. "
+                        "Return only the file bytes as plain text. "
+                        "Do not wrap the answer in Markdown fences, commentary, or explanations.\n\n"
+                        f"Target path: {target_path}\n"
+                        f"User request: {req.message}"
+                    )
+                    raw_content = await router.generate_as(
+                        router.current_model_id,
+                        message="Return only the requested file content.",
+                        context=generation_context,
+                        max_tokens=req.max_tokens,
+                        temperature=req.temperature,
+                    )
+                    content = strip_generated_file_content(str(raw_content))
+                if content is None:
+                    content = ""
+                try:
+                    result = execute_tool("write_file", {"path": target_path, "content": content})
+                except ToolError as exc:
+                    raise HTTPException(status_code=400, detail=str(exc)) from exc
+                answer = f"{result.get('path') or target_path} 파일을 만들었습니다."
+                created_files = [{
+                    "path": result.get("path") or target_path,
+                    "filename": Path(result.get("path") or target_path).name,
+                    "bytes": result.get("bytes", 0),
+                    "action": "write_file",
+                }]
+                notify_chat_message("user", req.message, req.source)
+                notify_chat_message("assistant", answer, req.source)
+                payload = {
+                    "status": "ok",
+                    "response": answer,
+                    "workspace": str(AGENT_ROOT),
+                    "steps": [{
+                        "state": AgentState.EXECUTING.value,
+                        "action": "write_file",
+                        "args": {"path": target_path},
+                        "result": result,
+                    }],
+                    "state_history": [AgentState.EXECUTING.value, AgentState.DONE.value],
+                    "final_state": AgentState.DONE.value,
+                    "created_files": created_files,
+                    "routed_to_agent": True,
+                    "action_route": "direct_write_file",
+                }
+                if req.stream:
+                    async def _stream_file_result():
+                        yield f"data: {json.dumps({'chunk': answer, 'model': router.current_model_id, 'agent': payload}, ensure_ascii=False)}\n\n"
+                        yield f"data: {json.dumps({'chunk': '', 'model': router.current_model_id, 'agent': payload}, ensure_ascii=False)}\n\n"
+                        yield "data: [DONE]\n\n"
+                    return StreamingResponse(
+                        _stream_file_result(),
+                        media_type="text/event-stream",
+                        headers={"X-Model": router.current_model_id or "tool", "X-Routed-To": "agent"},
+                    )
+                return JSONResponse(content=payload)
+
         if not router.current_model_id:
             detail = "No model loaded. Call /models/load first."
             if CONFIG.is_public:
@@ -420,12 +571,40 @@ def create_chat_router(context: AppContext) -> APIRouter:
                     "action": "load_model",
                 },
             )
-    
+
         if req.model and req.model != router.current_model_id:
             if req.model not in router.loaded_model_ids:
                 raise HTTPException(status_code=404, detail=f"Model '{req.model}' not loaded.")
             router.switch_model(req.model)
-    
+
+        if is_file_action_request(req.message):
+            agent_req = AgentRequest(
+                message=req.message,
+                conversation_id=req.conversation_id,
+                source=req.source or "web",
+                max_steps=25,
+                temperature=min(req.temperature, 0.2),
+                user_email=effective_email,
+                user_nickname=req.user_nickname,
+                workspace_id=workspace_scope_from_request(request),
+            )
+            result = await agent(agent_req, request)
+            answer = str(result.get("response") or "파일 작업을 처리했습니다.")
+            notify_chat_message("user", req.message, req.source)
+            notify_chat_message("assistant", answer, req.source)
+            result["routed_to_agent"] = True
+            if req.stream:
+                async def _stream_agent_result():
+                    yield f"data: {json.dumps({'chunk': answer, 'model': router.current_model_id, 'agent': result}, ensure_ascii=False)}\n\n"
+                    yield f"data: {json.dumps({'chunk': '', 'model': router.current_model_id, 'agent': result}, ensure_ascii=False)}\n\n"
+                    yield "data: [DONE]\n\n"
+                return StreamingResponse(
+                    _stream_agent_result(),
+                    media_type="text/event-stream",
+                    headers={"X-Model": router.current_model_id, "X-Routed-To": "agent"},
+                )
+            return JSONResponse(content=result)
+
         lang = detect_language(req.message)
         context = f"[LANGUAGE: {_LANG_HINT[lang]}]\n" + (req.context or "")
         # v4 Context System: one budgeted, provenance-carrying assembly
@@ -462,12 +641,12 @@ def create_chat_router(context: AppContext) -> APIRouter:
                     print("📝 Document generation context retrieved from knowledge graph.")
         except Exception as e:
             logging.warning("Knowledge graph reinforcement skipped: %s", e)
-    
+
         if req.image_data:
             screenshot_context = extract_screenshot_context(req.image_data)
             if screenshot_context:
                 context += f"\n\n{screenshot_context}"
-    
+
         if CONFIG.auto_read_chat_paths:
             _file_path_re = re.compile(r'(?:^|[\s\'\"(])((~|/[\w.])[^\s\'")\]]*)', re.MULTILINE)
             requested_paths = [_m.group(1).strip() for _m in _file_path_re.finditer(req.message or "")]
@@ -487,7 +666,7 @@ def create_chat_router(context: AppContext) -> APIRouter:
                             "Attach the file, upload it, or use an approved local-file tool flow."
                         ),
                     )
-    
+
         trace_seed = CHAT_SERVICE.build_graph_trace(
             req.message,
             KNOWLEDGE_GRAPH if (ENABLE_GRAPH and KNOWLEDGE_GRAPH) else None,
@@ -497,11 +676,11 @@ def create_chat_router(context: AppContext) -> APIRouter:
             # Persisted with the answer trace: 'why is this in my context?'
             # is answerable from the stored record (UI surface lands in T9b).
             trace_seed["context_assembly"] = context_trace
-    
+
         history_message = f"{req.message}\n[Image attached]" if req.image_data else req.message
         save_to_history("user", history_message, **history_meta, **history_user)
         notify_chat_message("user", req.message, req.source)
-    
+
         if is_doc_gen and ENABLE_GRAPH and KNOWLEDGE_GRAPH:
             conv_key = req.conversation_id or "default"
             session = _doc_gen_sessions.get(conv_key)
@@ -512,7 +691,7 @@ def create_chat_router(context: AppContext) -> APIRouter:
             system_prompt = session.get_system_prompt(graph_md)
             sources = (doc_gen_context_result or {}).get("sources", [])
             footnote = format_sources_footnote(sources)
-    
+
             if req.stream:
                 async def _stream_doc_gen():
                     collected = []
@@ -563,7 +742,7 @@ def create_chat_router(context: AppContext) -> APIRouter:
                 )
                 notify_chat_message("assistant", str(result), req.source)
                 return JSONResponse(content={"response": str(result), "trace_id": trace_record["id"], "trace": trace_record})
-    
+
         if req.stream:
             recent_context = recent_chat_context(user_email=effective_email, conversation_id=req.conversation_id)
             stream_context = context
@@ -593,9 +772,9 @@ def create_chat_router(context: AppContext) -> APIRouter:
             else:
                 history_context = recent_chat_context(user_email=effective_email, conversation_id=req.conversation_id)
                 full_context = f"{history_context}\n{context}" if context else history_context
-    
+
             result = await router.generate(req.message, full_context, req.max_tokens, req.temperature, req.image_data)
-    
+
             save_to_history("assistant", str(result), **history_meta, **history_user)
             trace_record = CHAT_SERVICE.record_trace(
                 question=req.message,
@@ -605,22 +784,22 @@ def create_chat_router(context: AppContext) -> APIRouter:
                 trace=trace_seed,
             )
             notify_chat_message("assistant", str(result), req.source)
-    
+
             return JSONResponse(content={"response": str(result), "trace_id": trace_record["id"], "trace": trace_record})
-    
+
     
     @api_router.get("/history")
     async def fetch_history(request: Request):
         """웹 화면에서 이전 대화를 불러올 수 있도록 히스토리를 반환합니다."""
         require_user(request)
         return get_history()
-    
+
     @api_router.get("/history/conversations")
     async def fetch_history_conversations(request: Request):
         """저장된 히스토리를 대화 단위로 묶어 반환합니다."""
         require_user(request)
         return group_history_conversations()
-    
+
     @api_router.get("/history/conversations/{conversation_id:path}")
     async def fetch_history_conversation(conversation_id: str, request: Request):
         """선택한 대화의 메시지를 반환합니다."""
@@ -629,7 +808,7 @@ def create_chat_router(context: AppContext) -> APIRouter:
         if not messages:
             raise HTTPException(status_code=404, detail="대화를 찾을 수 없습니다.")
         return {"id": conversation_id, "messages": messages}
-    
+
     
     @api_router.delete("/history/conversations/{conversation_id:path}")
     async def delete_history_conversation(conversation_id: str, request: Request):
@@ -645,7 +824,7 @@ def create_chat_router(context: AppContext) -> APIRouter:
             kept=result.get("kept", 0),
         )
         return result
-    
+
     
     @api_router.delete("/history")
     async def delete_history(request: Request, keep_last: int = 0):
@@ -659,7 +838,7 @@ def create_chat_router(context: AppContext) -> APIRouter:
             kept=result.get("kept", 0),
         )
         return result
-    
+
     @api_router.get("/history/search")
     async def search_history(q: str, request: Request):
         """키워드로 채팅 히스토리를 검색합니다."""
@@ -676,7 +855,7 @@ def create_chat_router(context: AppContext) -> APIRouter:
                 grouped[cid] = {"conversation_id": cid, "title": conversation_title(item), "messages": []}
             grouped[cid]["messages"].append(item)
         return {"results": list(grouped.values())[-30:], "query": q}
-    
+
     async def _stream_chat(
         req: ChatRequest,
         context: str = "",
@@ -696,7 +875,7 @@ def create_chat_router(context: AppContext) -> APIRouter:
                     clean_chunk = chunk.split("text='")[1].split("', token=")[0].replace('\\n', '\n').replace('\\\\n', '\n')
                 except Exception:
                     pass
-    
+
             full_response += str(clean_chunk)
             yield f"data: {json.dumps({'chunk': clean_chunk, 'model': router.current_model_id}, ensure_ascii=False)}\n\n"
         history_user = get_history_user(effective_email or req.user_email, req.user_nickname)
@@ -724,7 +903,7 @@ def create_chat_router(context: AppContext) -> APIRouter:
         schema_path = skill_dir / "schema.json"
         if not schema_path.exists():
             raise HTTPException(404, detail=f"Skill '{req.skill}' not found or missing schema.json")
-    
+
         schema = json.loads(schema_path.read_text(encoding="utf-8"))
         eval_cases = schema.get("evals", [])
         if req.case_id:
@@ -732,7 +911,7 @@ def create_chat_router(context: AppContext) -> APIRouter:
         if not eval_cases:
             return {"skill": req.skill, "total": 0, "passed": 0, "failed": 0, "results": [],
                     "message": "No eval cases defined in schema.json"}
-    
+
         action_name = schema.get("action", req.skill)
         results = []
         for case in eval_cases:
@@ -754,19 +933,19 @@ def create_chat_router(context: AppContext) -> APIRouter:
                 results.append({"id": case_id, "description": case.get("description", ""),
                                 "passed": False, "error": str(exc),
                                 "pass_criteria": case.get("pass_criteria", "")})
-    
+
         n_passed = sum(1 for r in results if r.get("passed") is True)
         return {
             "skill": req.skill, "action": action_name,
             "total": len(results), "passed": n_passed, "failed": len(results) - n_passed,
             "results": results,
         }
-    
+
     
     @api_router.post("/agent")
     async def agent(req: AgentRequest, request: Request):
         """Natural-language local agent.
-    
+
         State machine:
             IDLE → PLANNING → WAITING_APPROVAL → EXECUTING → VERIFYING
                                            ↓                     ↓
@@ -779,22 +958,22 @@ def create_chat_router(context: AppContext) -> APIRouter:
         req.workspace_id = req.workspace_id or workspace_scope_from_request(request)
         if not router.current_model_id:
             raise HTTPException(status_code=400, detail="No model loaded. Call /models/load first.")
-    
+
         ensure_agent_root()
         lang = detect_language(req.message)
         lang_hint = _LANG_HINT[lang]
         max_steps = max(1, min(req.max_steps, 50))
         max_retry = 3
-    
+
         ctx = AgentRunContext()
         ctx.executing_model = req.executing_model
         ctx.reviewing_model = req.reviewing_model
-    
+
         # PLANNING phase
         ctx.state = AgentState.PLANNING
         ctx.state_history.append(ctx.state.value)
         await _AGENT_RUNTIME.plan(ctx, req, lang_hint, current_user, model_id=req.planning_model)
-    
+
         # Human-in-the-loop: pause after planning, return plan to UI
         if req.human_in_loop:
             context_id = secrets.token_urlsafe(16)
@@ -810,11 +989,11 @@ def create_chat_router(context: AppContext) -> APIRouter:
                 "executing_model": req.executing_model or router.current_model_id,
                 "reviewing_model": req.reviewing_model or router.current_model_id,
             }
-    
+
         # Auto-approve and run to completion (default behaviour)
         _AGENT_RUNTIME.approve(ctx, current_user)
         return await _agent_finish(ctx, req, lang_hint, current_user, max_steps, max_retry)
-    
+
     
     async def _agent_finish(
         ctx: AgentRunContext, req: AgentRequest, lang_hint: str,
@@ -823,7 +1002,7 @@ def create_chat_router(context: AppContext) -> APIRouter:
         """HTTP glue: drive the runtime to a terminal state, persist, shape the response."""
         await _AGENT_RUNTIME.run_to_completion(ctx, req, lang_hint, current_user, max_steps, max_retry)
         asyncio.create_task(_AGENT_RUNTIME.memory_update(ctx, req, current_user))
-    
+
         message = ctx.final_message or "작업을 완료했습니다."
         save_to_history("user", req.message, source=req.source or "web", conversation_id=req.conversation_id, workspace_id=req.workspace_id)
         save_to_history("assistant", message, source=req.source or "web", conversation_id=req.conversation_id, workspace_id=req.workspace_id)
@@ -850,33 +1029,33 @@ def create_chat_router(context: AppContext) -> APIRouter:
             "final_state": ctx.state.value,
             "created_files": created_files,
         }
-    
+
     
     @api_router.post("/agent/resume")
     async def agent_resume(req: AgentResumeRequest, request: Request):
         """Resume a paused agent after human approval of the plan."""
         current_user = require_user(request)
-    
+
         with _pending_agents_lock:
             entry = _pending_agents.pop(req.context_id, None)
         if not entry:
             raise HTTPException(status_code=404, detail="Agent context not found or expired. Start a new request.")
-    
+
         ctx, orig_req, lang_hint, _orig_user = entry
-    
+
         if not req.approved:
             return {"status": "cancelled", "response": "사용자가 계획을 취소했습니다."}
-    
+
         if req.modified_plan:
             ctx.plan = req.modified_plan
             ctx.transcript[-1].update(ctx.plan)  # keep transcript in sync
-    
+
         # Apply model overrides from resume request (takes priority over original request)
         ctx.executing_model = req.executing_model or ctx.executing_model
         ctx.reviewing_model = req.reviewing_model or ctx.reviewing_model
-    
+
         _AGENT_RUNTIME.approve(ctx, current_user)
-    
+
         max_steps = max(1, min(orig_req.max_steps, 50))
         max_retry = 3
         return await _agent_finish(ctx, orig_req, lang_hint, current_user, max_steps, max_retry)

package/latticeai/core/legacy_compatibility.py

@@ -1,7 +1,7 @@
 """Managed legacy compatibility surface for legacy import shims.
 
 Compatibility modules are intentionally still present for old scripts, VS Code
-extension paths, historical integrations, and pre-graph-package imports. 8.3.0
+extension paths, historical integrations, and pre-graph-package imports. 8.4.0
 stops treating them as vague technical debt: every tracked shim has an owner,
 migration target, removal phase, and replacement import that can be surfaced in
 docs and tests.
@@ -14,7 +14,7 @@ from pathlib import Path
 from typing import Any, Dict, List
 
 
-LEGACY_COMPATIBILITY_VERSION = "8.3.0"
+LEGACY_COMPATIBILITY_VERSION = "8.4.0"
 
 
 @dataclass(frozen=True)

package/latticeai/core/marketplace.py

@@ -11,7 +11,7 @@ from copy import deepcopy
 from typing import Any, Dict, List, Optional
 
 
-MARKETPLACE_VERSION = "8.3.0"
+MARKETPLACE_VERSION = "8.4.0"
 TEMPLATE_KINDS = ("plugin", "workflow", "agent")
 
 

package/latticeai/core/workspace_os.py

@@ -49,7 +49,7 @@ __all__ = [
     "remove_skill_directory",
 ]
 
-WORKSPACE_OS_VERSION = "8.3.0"
+WORKSPACE_OS_VERSION = "8.4.0"
 
 # Workspace types separate single-user Personal workspaces from shared
 # Organization workspaces. Both keep the same local-first JSON store; the type

package/latticeai/services/architecture_readiness.py

@@ -1,6 +1,6 @@
 """Machine-checkable architecture readiness gates for release work.
 
-8.3.0 keeps the major architecture priorities under an explicit release
+8.4.0 keeps the major architecture priorities under an explicit release
 contract while product maturity work reduces visible beta seams. AgentRuntime, ToolRegistry,
 central Config, decomposed server runtime, and Knowledge Graph stabilization
 must remain discoverable, ordered, and backed by tests before the release can be
@@ -17,7 +17,7 @@ from typing import Any, Dict, List
 from latticeai.core.legacy_compatibility import legacy_shim_report
 
 
-ARCHITECTURE_VERSION_TARGET = "8.3.0"
+ARCHITECTURE_VERSION_TARGET = "8.4.0"
 
 PREFERRED_REFACTORING_ORDER = [
     "agent-runtime",

package/latticeai/services/model_catalog.py

@@ -102,6 +102,12 @@ MODEL_ENGINE_ALIASES = {
         "lmstudio": "ggml-org/gemma-4-12B-it-GGUF",
         "llamacpp": "ggml-org/gemma-4-12B-it-GGUF",
     },
+    "gemma-4-26b-it-4bit": {
+        "local_mlx": "mlx-community/gemma-4-26b-a4b-it-4bit",
+    },
+    "mlx-community/gemma-4-26b-a4b-it-4bit": {
+        "local_mlx": "mlx-community/gemma-4-26b-a4b-it-4bit",
+    },
     "gemma-4-31b-it-4bit": {
         "local_mlx": "mlx-community/gemma-4-31b-it-4bit",
         "ollama": "hf.co/ggml-org/gemma-4-31B-it-GGUF:Q4_K_M",

package/latticeai/services/model_runtime.py

@@ -31,6 +31,11 @@ from latticeai.models.router import (
     parse_model_ref,
 )
 from latticeai.core.model_resolution import ModelResolution as _ModelResolution
+from latticeai.core.model_compat import (
+    friendly_model_runtime_error as _friendly_model_runtime_error,
+    model_runtime_compatibility as _model_runtime_compatibility,
+    SMOKE_PROMPT as _SMOKE_PROMPT,
+)
 from .model_engines import (
     ensure_lmstudio_server as _ensure_lmstudio_server,
     ensure_ollama_server as _ensure_ollama_server,
@@ -50,6 +55,14 @@ from .model_engines import (
     LOCAL_SERVER_PROCESSES as _LOCAL_SERVER_PROCESSES,
 )
 
+# ``model_loading._get_model_runtime_deps`` imports these private names from
+# this module to preserve the historical model_runtime wiring surface.
+_MODEL_LOADING_COMPAT_EXPORTS = (
+    _friendly_model_runtime_error,
+    _model_runtime_compatibility,
+    _SMOKE_PROMPT,
+)
+
 
 def _missing_current_user(_request: Request) -> Optional[str]:
     return None

package/latticeai/services/product_readiness.py

@@ -1,7 +1,7 @@
-"""Machine-checkable *product* readiness gates for the 8.3 line.
+"""Machine-checkable *product* readiness gates for the 8.4 line.
 
 Where ``architecture_readiness`` proves the internal structure is sound, this
-module answers the product question the 8.3 release exists to settle: *does the
+module answers the product question the 8.4 release exists to settle: *does the
 app now feel like a finished product rather than only a strong framework?*
 It does so honestly: every gate is backed by
 evidence that is probed on disk, so a gate only reports ``complete`` when its
@@ -18,7 +18,7 @@ from typing import Any, Dict, List
 
 from latticeai.services.architecture_readiness import architecture_readiness
 
-PRODUCT_VERSION_TARGET = "8.3.0"
+PRODUCT_VERSION_TARGET = "8.4.0"
 
 
 @dataclass(frozen=True)
@@ -52,6 +52,15 @@ PRODUCT_GATES: List[ProductGate] = [
             "scripts/brain_quality_eval.py",
         ],
     ),
+    ProductGate(
+        id="action-aware-chat",
+        title="File action requests create artifacts instead of code-only answers",
+        evidence=[
+            "latticeai/api/chat.py::is_file_action_request",
+            "latticeai/api/chat.py::direct_write_file",
+            "tests/unit/test_chat_telegram_decoupling.py::test_chat_file_creation_intent_writes_real_file",
+        ],
+    ),
     ProductGate(
         id="local-first-trust",
         title="Local-first privacy is stated and bounded",
@@ -67,10 +76,10 @@ PRODUCT_GATES: List[ProductGate] = [
         evidence=[
             "package.json::release:artifacts",
             "package.json::release:validate",
-            "README.md::dist/ltcai-8.3.0-py3-none-any.whl",
-            "README.md::dist/ltcai-8.3.0.tar.gz",
-            "README.md::dist/ltcai-8.3.0.vsix",
-            "README.md::ltcai-8.3.0.tgz",
+            "README.md::dist/ltcai-8.4.0-py3-none-any.whl",
+            "README.md::dist/ltcai-8.4.0.tar.gz",
+            "README.md::dist/ltcai-8.4.0.vsix",
+            "README.md::ltcai-8.4.0.tgz",
             "scripts/validate_release_artifacts.py",
             "scripts/release_smoke.py",
             "Dockerfile",
@@ -86,12 +95,12 @@ PRODUCT_GATES: List[ProductGate] = [
         title="Release story is documented and honest",
         evidence=[
             "README.md",
-            "README.md::The current release is **8.3.0",
-            "SECURITY.md::8.3.x (latest)",
-            "vscode-extension/README.md::**8.3.0",
-            "docs/CHANGELOG.md::## [8.3.0]",
+            "README.md::The current release is **8.4.0",
+            "SECURITY.md::8.4.x (latest)",
+            "vscode-extension/README.md::**8.4.0",
+            "docs/CHANGELOG.md::## [8.4.0]",
             "FEATURE_STATUS.md",
-            "RELEASE_NOTES_v8.3.0.md",
+            "RELEASE_NOTES_v8.4.0.md",
             "latticeai/core/agent.py::SingleAgentRuntime",
             "latticeai/core/agent.py::AgentRuntime = SingleAgentRuntime",
             "lattice_brain/runtime/contracts.py::runtime-boundary/v1",
@@ -106,7 +115,7 @@ PRODUCT_GATES: List[ProductGate] = [
         id="ecosystem-path",
         title="Community and plugin growth path is explicit",
         evidence=[
-            "docs/COMMUNITY_AND_PLUGINS.md::LatticeAI 8.3.0",
+            "docs/COMMUNITY_AND_PLUGINS.md::LatticeAI 8.4.0",
             "docs/PLUGIN_SDK.md",
             "plugins/README.md",
             "plugins/hello-world/plugin.json",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ltcai",
-  "version": "8.3.0",
+  "version": "8.4.0",
   "description": "Lattice AI — local-first Digital Brain that keeps your knowledge durable across any AI model.",
   "homepage": "https://github.com/TaeSooPark-PTS/LatticeAI#readme",
   "repository": {

package/src-tauri/Cargo.lock

@@ -1584,7 +1584,7 @@ dependencies = [
 
 [[package]]
 name = "lattice-ai-desktop"
-version = "8.3.0"
+version = "8.4.0"
 dependencies = [
  "plist",
  "serde",

package/src-tauri/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lattice-ai-desktop"
-version = "8.3.0"
+version = "8.4.0"
 description = "Lattice AI Digital Brain desktop shell"
 authors = ["TaeSoo Park"]
 edition = "2021"

package/src-tauri/tauri.conf.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://schema.tauri.app/config/2",
   "productName": "Lattice AI",
-  "version": "8.3.0",
+  "version": "8.4.0",
   "identifier": "ai.lattice.desktop",
   "build": {
     "beforeDevCommand": "npm run frontend:dev",

package/static/app/asset-manifest.json

@@ -1,5 +1,5 @@
 {
-  "version": "8.3.0",
+  "version": "8.4.0",
   "generated_at": "vite",
   "entrypoints": {
     "app": "/static/app/assets/index-CoiuIFFP.js"

package/static/sw.js

@@ -1,7 +1,7 @@
 // Lattice Service Worker — PWA install + offline shell for the /app SPA.
 // Strategy: precache the Vite app bundle from its asset manifest,
 // cache-first for static assets, network-only for everything dynamic.
-const CACHE = "lattice-v830";
+const CACHE = "lattice-v840";
 const MANIFEST_URL = "/static/app/asset-manifest.json";
 
 // Non-manifest assets the shell needs offline.