janction-render 0.4.83__tar.gz → 0.4.87__tar.gz

{janction_render-0.4.83 → janction_render-0.4.87}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: janction-render
-Version: 0.4.83
+Version: 0.4.87
 Summary: Cloud GPU rendering for Blender scenes and 3D models (GLB, FBX, USD, OBJ, STL), built for AI agents: MCP server (remote + stdio), HTTP API and CLI. Product shots, 360-degree turntables, animations and stills; preview in seconds with a critic; no local Blender or GPU. A 100 JPY welcome credit for a new key (one per network every 30 days) and 400 JPY more for saving a card that is not charged, then 0.1 JPY per GPU-second.
 Author-email: "JasmyLab Inc." <contact@jasmylab.com>
 License-Expression: MIT
@@ -57,7 +57,7 @@ Claude Code, Codex, Cursor or any MCP client**: an MCP server (remote and stdio)
 JANCTION Render is a separate product from SmartRender (JasmyLab's distributed rendering for people at a desktop) and is
 not affiliated with Render.com or the Render Network. The numbers, as of 2026-10-08 (the primary source is
 https://render.janction.jp/facts.json): a 100 JPY welcome credit (1,000 GPU-seconds, 14 days, no card) for a new key, one per network every 30 days (further keys on the same network share what is left), plus 400 JPY more once a card is saved on a page that charges nothing (once per card); after that, previews
-are free up to 2 GPU-minutes a day and finals cost 0.1 JPY per GPU-second from prepaid credit (paid use started on 2026-10-08); finals up to 1,000 frames per instruction (run as parts of 240 under one job id); 1080p on free time, up to 4K on prepaid credit; inputs and results deleted
+are free up to 2 GPU-minutes a day and finals cost 0.1 JPY per GPU-second from prepaid credit (paid use started on 2026-10-08); finals up to 3,000 frames per instruction (run as parts of 240 under one job id); 1080p on free time, up to 4K on prepaid credit; inputs and results deleted
 24 hours after last use (7 days for keys that have topped up).
 
 Official site: https://render.janction.jp · MCP endpoint: `https://render.janction.jp/mcp` · Fact sheet: https://render.janction.jp/facts
@@ -116,7 +116,7 @@ Keep 3D files in a repository? Add a button that lets anyone render them from th
   errors with the fix, spending caps per key, safe retries with `Idempotency-Key`.
 - **3D files and turntables.** glTF / GLB, FBX, USD, OBJ, STL, PLY, Alembic are imported into an empty scene with a camera and
   HDRI lighting; `orbit=True` makes a turntable. Fixed-price outcomes: `POST /v1/outcomes/turntable` and `/product-shot`.
-- **Honest limits.** a 100 JPY welcome credit for a new key with no card (14 days; one per network every 30 days) and 400 JPY more after saving a card that is not charged, then 0.1 JPY per GPU-second (previews free up to 2 GPU-minutes a day), finals up to 1,000 frames per instruction as parts of 240 (1080p on free time, up to 4K on prepaid credit), one GPU shared with another
+- **Honest limits.** a 100 JPY welcome credit for a new key with no card (14 days; one per network every 30 days) and 400 JPY more after saving a card that is not charged, then 0.1 JPY per GPU-second (previews free up to 2 GPU-minutes a day), finals up to 3,000 frames per instruction as parts of 240 (1080p on free time, up to 4K on prepaid credit), one GPU shared with another
   workload (jobs can wait; the ETA says so). Inputs and results are deleted 24 hours after last use (7 days for keys that have topped up) and never used for training.
 - **Not Render.com.** Same word, different product.
 
@@ -421,7 +421,7 @@ nothing adds 400 JPY more, once per card (up to 500 JPY in all). After it is use
 free up to 2 GPU-minutes a day and finals cost 0.1 JPY per
 GPU-second from prepaid credit, bought by card at a Stripe Checkout page (from 500 JPY; 2,000 JPY gives 2,200 JPY of credit,
 5,000 gives 5,750, 10,000 gives 12,000; the first top-up in the welcome period counts 1.5x, bonus up to 500 JPY). Only the part
-the free time does not cover is charged. Finals go up to 1,000 frames per instruction (run as parts of 240 under one job id): 1080p on free time, up to 4K (3840x2160) on prepaid credit. `POST /v1/estimate` quotes what will be charged
+the free time does not cover is charged. Finals go up to 3,000 frames per instruction (run as parts of 240 under one job id): 1080p on free time, up to 4K (3840x2160) on prepaid credit. `POST /v1/estimate` quotes what will be charged
 (`cost.estimated_yen`), the currency and an expiry before any GPU time is used: https://render.janction.jp/pricing
 
 ## Troubleshooting

{janction_render-0.4.83 → janction_render-0.4.87}/README.md

@@ -17,7 +17,7 @@ Claude Code, Codex, Cursor or any MCP client**: an MCP server (remote and stdio)
 JANCTION Render is a separate product from SmartRender (JasmyLab's distributed rendering for people at a desktop) and is
 not affiliated with Render.com or the Render Network. The numbers, as of 2026-10-08 (the primary source is
 https://render.janction.jp/facts.json): a 100 JPY welcome credit (1,000 GPU-seconds, 14 days, no card) for a new key, one per network every 30 days (further keys on the same network share what is left), plus 400 JPY more once a card is saved on a page that charges nothing (once per card); after that, previews
-are free up to 2 GPU-minutes a day and finals cost 0.1 JPY per GPU-second from prepaid credit (paid use started on 2026-10-08); finals up to 1,000 frames per instruction (run as parts of 240 under one job id); 1080p on free time, up to 4K on prepaid credit; inputs and results deleted
+are free up to 2 GPU-minutes a day and finals cost 0.1 JPY per GPU-second from prepaid credit (paid use started on 2026-10-08); finals up to 3,000 frames per instruction (run as parts of 240 under one job id); 1080p on free time, up to 4K on prepaid credit; inputs and results deleted
 24 hours after last use (7 days for keys that have topped up).
 
 Official site: https://render.janction.jp · MCP endpoint: `https://render.janction.jp/mcp` · Fact sheet: https://render.janction.jp/facts
@@ -76,7 +76,7 @@ Keep 3D files in a repository? Add a button that lets anyone render them from th
   errors with the fix, spending caps per key, safe retries with `Idempotency-Key`.
 - **3D files and turntables.** glTF / GLB, FBX, USD, OBJ, STL, PLY, Alembic are imported into an empty scene with a camera and
   HDRI lighting; `orbit=True` makes a turntable. Fixed-price outcomes: `POST /v1/outcomes/turntable` and `/product-shot`.
-- **Honest limits.** a 100 JPY welcome credit for a new key with no card (14 days; one per network every 30 days) and 400 JPY more after saving a card that is not charged, then 0.1 JPY per GPU-second (previews free up to 2 GPU-minutes a day), finals up to 1,000 frames per instruction as parts of 240 (1080p on free time, up to 4K on prepaid credit), one GPU shared with another
+- **Honest limits.** a 100 JPY welcome credit for a new key with no card (14 days; one per network every 30 days) and 400 JPY more after saving a card that is not charged, then 0.1 JPY per GPU-second (previews free up to 2 GPU-minutes a day), finals up to 3,000 frames per instruction as parts of 240 (1080p on free time, up to 4K on prepaid credit), one GPU shared with another
   workload (jobs can wait; the ETA says so). Inputs and results are deleted 24 hours after last use (7 days for keys that have topped up) and never used for training.
 - **Not Render.com.** Same word, different product.
 
@@ -381,7 +381,7 @@ nothing adds 400 JPY more, once per card (up to 500 JPY in all). After it is use
 free up to 2 GPU-minutes a day and finals cost 0.1 JPY per
 GPU-second from prepaid credit, bought by card at a Stripe Checkout page (from 500 JPY; 2,000 JPY gives 2,200 JPY of credit,
 5,000 gives 5,750, 10,000 gives 12,000; the first top-up in the welcome period counts 1.5x, bonus up to 500 JPY). Only the part
-the free time does not cover is charged. Finals go up to 1,000 frames per instruction (run as parts of 240 under one job id): 1080p on free time, up to 4K (3840x2160) on prepaid credit. `POST /v1/estimate` quotes what will be charged
+the free time does not cover is charged. Finals go up to 3,000 frames per instruction (run as parts of 240 under one job id): 1080p on free time, up to 4K (3840x2160) on prepaid credit. `POST /v1/estimate` quotes what will be charged
 (`cost.estimated_yen`), the currency and an expiry before any GPU time is used: https://render.janction.jp/pricing
 
 ## Troubleshooting

{janction_render-0.4.83 → janction_render-0.4.87}/janction_render/__init__.py

@@ -1,3 +1,3 @@
 """JANCTION Render — ターミナルの AI エージェントから GPU で 3DCG を描く。"""
 
-__version__ = "0.4.83"
+__version__ = "0.4.87"

{janction_render-0.4.83 → janction_render-0.4.87}/janction_render/client.py

@@ -28,6 +28,8 @@ ARCHIVE_MIN_FILES = 8   # これ以上の数のファイルは zip で 1 回で
 # 届かず HTML の 413 になるので、送る前に断ってリンク（scene_url / asset_urls）で渡すよう言う。素材は 100 MB ずつに分けて送る
 # （10/12）。手元や自前の受付には掛けない。JANCTION_RENDER_MAX_UPLOAD_MB で変えられる（0 で見ない）
 EDGE_UPLOAD_MB = 100
+# この大きさを超えるシーンは分けて送る（10/13）。Cloudflare の 100 MB が 10^8 バイトなら 95.4 MiB から先は 1 回では通らない
+RESUMABLE_ABOVE = 64 * 1024 * 1024
 
 
 def select_artifacts(artifacts: list[dict[str, Any]], only: Optional[str] = None) -> list[dict[str, Any]]:
@@ -428,13 +430,18 @@ class Client:
             out = None
         if out is None:
             cap = self.upload_cap_bytes()
-            if cap and p.stat().st_size > cap:
-                raise self._too_large(f"the scene file {p.name}", p.stat().st_size, cap, "scene_url")
-            with p.open("rb") as fh:
-                r = self.s.post(f"{self.server}/v1/files", headers=self._headers(),
-                                files={"file": (p.name, fh, "application/octet-stream")}, timeout=1800)
-            self._raise(r)
-            out = r.json()
+            if cap and p.stat().st_size > min(cap, RESUMABLE_ABOVE):
+                # 1 回で送れる大きさ（Cloudflare の 100 MB）に近い・超える: 分けて送る（/v1/uploads、10/13）。受付が古ければ前のとおり断る
+                done = self.upload_resumable(p, digest)
+                if done is None:
+                    raise self._too_large(f"the scene file {p.name}", p.stat().st_size, cap, "scene_url")
+                out = done.get("scene") or done
+            else:
+                with p.open("rb") as fh:
+                    r = self.s.post(f"{self.server}/v1/files", headers=self._headers(),
+                                    files={"file": (p.name, fh, "application/octet-stream")}, timeout=1800)
+                self._raise(r)
+                out = r.json()
         items: list[tuple[str, Path]] = []
         missing: list[str] = []
         if p.suffix.lower() == ".blend" and trace_assets:
@@ -485,6 +492,113 @@ class Client:
         out["missing_assets"] = missing
         return out
 
+    # ---- 分けて送る（10/13、docs/67。最大 5 GiB の .blend）----------------------------------------
+
+    _RESUME_FILE = Path.home() / ".janction-render-uploads.json"
+
+    def _resume_get(self, ident: str) -> Optional[str]:
+        try:
+            return json.loads(self._RESUME_FILE.read_text(encoding="utf-8")).get(ident)
+        except (OSError, ValueError, AttributeError):
+            return None
+
+    def _resume_set(self, ident: str, url: Optional[str]) -> None:
+        try:
+            data = json.loads(self._RESUME_FILE.read_text(encoding="utf-8")) if self._RESUME_FILE.exists() else {}
+            if not isinstance(data, dict):
+                data = {}
+        except (OSError, ValueError):
+            data = {}
+        if url:
+            data[ident] = url
+        else:
+            data.pop(ident, None)
+        try:
+            self._RESUME_FILE.write_text(json.dumps(data), encoding="utf-8")
+        except OSError:
+            pass
+
+    def upload_resumable(self, path: str | Path, digest: str = "",
+                         progress: Optional[Callable[[int, int], None]] = None, part_bytes: int = 0) -> Optional[dict[str, Any]]:
+        """100 MB を超えるファイルを部分に分けて送る（POST /v1/uploads → PATCH、tus 1.0）。返り値は受付の状態（scene_id・scene）。
+        通信が切れたら位置を聞き直して続きから送る。途中で止めても、同じコマンドをもう一度打てば続きから（~/.janction-render-uploads.json）。
+        受付が分けて送る道を持っていなければ None。"""
+        import base64
+
+        p = Path(path)
+        size = p.stat().st_size
+        digest = digest or sha256_of(p)
+        ident = f"{self.server}|{digest}|{size}"
+        tus = {"Tus-Resumable": "1.0.0"}
+        url, off = self._resume_get(ident), 0
+        if url:
+            try:
+                h = self.s.head(url, headers={**self._headers(), **tus}, timeout=30)
+                if h.status_code == 200 and int(h.headers.get("Upload-Length") or -1) == size:
+                    off = int(h.headers.get("Upload-Offset") or 0)
+                else:
+                    url = None
+            except (ClientError, ValueError):
+                url = None
+        part = part_bytes
+        if not url:
+            meta = (f"filename {base64.b64encode(p.name.encode('utf-8')).decode()},"
+                    f"sha256 {base64.b64encode(digest.encode()).decode()}")
+            r = self.s.post(f"{self.server}/v1/uploads", timeout=60,
+                            headers={**self._headers(), **tus, "Upload-Length": str(size), "Upload-Metadata": meta})
+            if r.status_code in (404, 405):
+                return None
+            self._raise(r)
+            url = f"{self.server}{r.headers['Location']}"
+            part = part or int(r.json().get("part_bytes") or 0)
+            self._resume_set(ident, url)
+        part = part or 32 * 1024 * 1024
+        tries = 0
+        if progress:
+            progress(off, size)
+        with p.open("rb") as fh:
+            while True:
+                fh.seek(off)
+                data = fh.read(part)
+                try:
+                    r = self.s.patch(url, data=data, timeout=600, headers={
+                        **self._headers(), **tus, "Upload-Offset": str(off), "Content-Type": "application/offset+octet-stream"})
+                except ClientError:
+                    r = None                       # つながらない・切れた: 待って位置を聞き直す
+                if r is not None and r.status_code == 200:
+                    self._resume_set(ident, None)
+                    if progress:
+                        progress(size, size)
+                    return r.json()
+                if r is not None and r.status_code == 204:
+                    off, tries = int(r.headers.get("Upload-Offset") or off), 0
+                    if progress:
+                        progress(off, size)
+                    continue
+                if r is not None and r.status_code == 409 and r.headers.get("Upload-Offset") is not None:
+                    off = int(r.headers["Upload-Offset"])
+                    continue
+                if r is not None and r.status_code < 500 and r.status_code not in (408, 409, 429) or                         r is not None and r.status_code == 507:
+                    if r.status_code == 404:
+                        self._resume_set(ident, None)
+                    self._raise(r)
+                tries += 1
+                if tries > 8:
+                    raise ClientError(0, "upload_interrupted",
+                                      f"the upload of {p.name} kept failing at {off} of {size} bytes; run the same command again "
+                                      "to continue from there", extra={"retryable": True, "offset": off})
+                time.sleep(min(30.0, 2.0 ** tries))
+                try:
+                    h = self.s.head(url, headers={**self._headers(), **tus}, timeout=30)
+                    if h.status_code == 200:
+                        off = int(h.headers.get("Upload-Offset") or off)
+                    elif h.status_code == 404:
+                        self._resume_set(ident, None)
+                        self._raise(h)
+                except ClientError as exc:
+                    if exc.status:
+                        raise
+
     def upload_cap_bytes(self) -> int:
         """1 回の送信の上限（バイト）。0 は見ない（手元・自前の受付）。"""
         raw = (os.environ.get("JANCTION_RENDER_MAX_UPLOAD_MB") or "").strip()

{janction_render-0.4.83 → janction_render-0.4.87}/janction_render/mcp_common.py

@@ -8,6 +8,7 @@
 受付の server/jobs.py もプリセットの名前はここから取る。"""
 from __future__ import annotations
 
+import os
 from typing import Any, Optional
 
 PURPOSE_PRESETS = ("product", "turntable", "animation")
@@ -89,8 +90,10 @@ def estimate_next(est: dict[str, Any]) -> str:
     q = est.get("quota") or {}
     cost = est.get("cost") or {}
     if q and q.get("fits_size") is False:
-        return (f"too big for one job: at most {q.get('max_frames_per_job')} frames per job part (1920x1080 on free time, 3840x2160 "
-                "on prepaid credit); the service splits longer instructions into parts under one job id, or lower the size")
+        most = q.get("max_frames_per_instruction") or q.get("max_frames_per_job")
+        return (f"too big: at most {most} frames per render_final call (longer than {q.get('max_frames_per_job')} frames runs as "
+                "parts under one job id) and 1920x1080 on free time, 3840x2160 for keys with prepaid credit; split the frame range "
+                "into several calls or lower the size")
     if q and not q.get("fits_today", True) and q.get("mode") == "free_allowance":
         # 有料モード: 無料枠を超えた分は残高から（10/8）。額と残高を言ってから
         return (f"this goes past the free GPU time left: about {cost.get('estimated_yen')} JPY would be charged from credit "
@@ -223,3 +226,34 @@ def docs(flavor: str) -> dict[str, str]:
     f = "remote" if flavor == "remote" else "stdio"
     return {"ENV_DOC": _ENV_DOC[f] + _POLYHAVEN_DOC[f], "POLYHAVEN_DOC": _POLYHAVEN_DOC[f], "ORBIT_DOC": _ORBIT_DOC,
             "CAMERAS_DOC": _CAMERAS_DOC, "OUTPUT_DOC": _OUTPUT_DOC[f], "TRANSPARENT_DOC": _TRANSPARENT_DOC, "ENGINE_DOC": _ENGINE_DOC}
+
+
+def auto_idempotency_key(tool: str, body: dict[str, Any], now: Optional[float] = None, window_s: int = 120) -> str:
+    """render_final にキーが無いときの自動の Idempotency-Key（10/13）。MCP の呼び出しが時間切れになって AI が同じ
+    render_final を呼び直すと、同じ仕事が 2 本描かれて 2 回引かれていた。同じ中身・同じ 2 分の枠なら同じキーになり、
+    受付は最初の仕事を返す。中身が 1 つでも違えば別のキー（＝別の仕事）。"""
+    import hashlib
+    import json as _json
+    import time as _time
+
+    t = _time.time() if now is None else now
+    blob = _json.dumps({"tool": tool, "body": body}, sort_keys=True, separators=(",", ":"), default=str)
+    return f"auto:{int(t // window_s)}:" + hashlib.sha256(blob.encode()).hexdigest()[:32]
+
+
+def digest(text: str) -> str:
+    """長い引数（scene_script）をキーに入れるための短い要約。"""
+    import hashlib
+
+    return hashlib.sha256((text or "").encode("utf-8", "replace")).hexdigest()[:24] if text else ""
+
+
+def file_stamp(path: str) -> str:
+    """手元のファイルの「同じもの」の目印（パス・大きさ・更新時刻）。中身を読まずに済ませる。"""
+    if not path:
+        return ""
+    try:
+        st = os.stat(os.path.expanduser(path))
+        return f"{os.path.abspath(os.path.expanduser(path))}:{st.st_size}:{int(st.st_mtime)}"
+    except OSError:
+        return path

{janction_render-0.4.83 → janction_render-0.4.87}/janction_render/mcp_server.py

@@ -658,7 +658,13 @@ def render_final(
                       camera=camera or None, transparent=(True if transparent else None), notify_url=notify_url or None,
                       max_yen=(max_yen if max_yen >= 0 else None), idempotency_key=idempotency_key or None,
                       **_job_params(environment, environment_strength, environment_visible, blender, engine=engine), **extra)
-        j = c.submit(sid, kind="final", **_with_preset(params, preset, "final"))
+        params = _with_preset(params, preset, "final")
+        if not params.get("idempotency_key"):
+            # 時間切れのあとの呼び直しで 2 本描かない（10/13、mcp_common.auto_idempotency_key）
+            params["idempotency_key"] = mcp_common.auto_idempotency_key(
+                "render_final", {"scene_id": scene_id, "scene_path": mcp_common.file_stamp(scene_path), "scene_url": scene_url,
+                                 "scene_script": mcp_common.digest(scene_script), "assets": assets, **params})
+        j = c.submit(sid, kind="final", **params)
     except ClientError as exc:
         block = _payment_block(exc) or _quota_block(exc)
         return _j(block if block else _fail(exc))

{janction_render-0.4.83 → janction_render-0.4.87}/janction_render/prompts.py

@@ -96,7 +96,7 @@ def blend_to_mp4_text(scene: str, frames: str, resolution: str, remote: bool) ->
     span = f"frames {fr}" if fr else "the scene's own frame range"
     return (f"Render this Blender animation to an MP4 with JANCTION Render: {scene.strip()}. " + _scene_hint(remote)
             + f"Use {span} at {res}. Start with scene_info (frame range, cameras, missing files), then render_preview with "
-            "four frames spread over the range (frames='1-N' tiles them in one image). One instruction takes up to 1,000 frames "
+            "four frames spread over the range (frames='1-N' tiles them in one image). One instruction takes up to 3,000 frames "
             "(the service runs them as parts of 240 under one job id; 1080p on free time, 4K on prepaid credit): tell me the plan. " + _CONFIRM)
 
 
@@ -151,7 +151,7 @@ def register(mcp: Any, remote: bool = False) -> list[str]:
 
     @mcp.prompt(name="blend_to_mp4", title="Render a Blender animation to MP4",
                 description="An animated .blend to MP4 on cloud GPUs: reads the scene, previews four frames, plans the render "
-                            "(up to 1,000 frames per instruction, run as parts of 240), then renders after you confirm.")
+                            "(up to 3,000 frames per instruction, run as parts of 240), then renders after you confirm.")
     def blend_to_mp4(scene: Annotated[str, Field(description=_SCENE)],
                      frames: Annotated[str, Field(description="Frame range such as 1-120 (empty: the scene's own range)")] = "",
                      resolution: Annotated[str, Field(description="WIDTHxHEIGHT, up to 1920x1080 (default 1920x1080)")] = "1920x1080") -> str:

{janction_render-0.4.83 → janction_render-0.4.87}/janction_render/tooldocs.py

@@ -30,6 +30,7 @@ FREE_GPU_MINUTES_PER_KEY_PER_DAY = 10
 FREE_MAX_FRAMES = 240
 FREE_MAX_SIZE = "1920x1080"
 RETENTION_HOURS = 24
+PAID_RETENTION_DAYS = 7       # 一度でもチャージした鍵（core.paid_ttl_seconds の既定 168 時間）
 YEN_PER_GPU_SECOND = 0.1      # 無料枠を超えた分の単価（10/8 から有料。billing.yen_per_gpu_second の既定値）
 # ようこそクレジット（docs/46、JR_FREE_MODEL=welcome）。billing の既定値と同じ（tests/test_tooldocs.py が突き合わせる）
 WELCOME_YEN = 500                 # network の形: 新しい鍵に 500 円分
@@ -98,13 +99,13 @@ def _free_offer(short: bool = False) -> str:
             "GPU-second (prepaid)")
 
 
-MAX_FRAMES_PER_INSTRUCTION = 1000      # billing.max_frames_per_instruction の既定値（240 ずつの部分に分けて 1 つの仕事 ID で）
+MAX_FRAMES_PER_INSTRUCTION = 3000      # billing.max_frames_per_instruction の既定値（240 ずつの部分に分けて 1 つの仕事 ID で）
 PAID_MAX_SIZE = "3840x2160"            # 先払いのクレジットでの上限（無料の時間は 1080p まで）
 
 
 def _limits() -> str:
-    return (f"LIMITS: {_free_offer(short=True)}; up to {MAX_FRAMES_PER_INSTRUCTION:,} frames per instruction (parts of {FREE_MAX_FRAMES}), "
-            f"{FREE_MAX_SIZE} free / {PAID_MAX_SIZE} prepaid; kept {RETENTION_HOURS} h.")
+    return (f"LIMITS: {_free_offer(short=True)}; up to {MAX_FRAMES_PER_INSTRUCTION:,} frames per call (parts of {FREE_MAX_FRAMES}), "
+            f"{FREE_MAX_SIZE} free / {PAID_MAX_SIZE} prepaid; kept {RETENTION_HOURS} h ({PAID_RETENTION_DAYS} d if topped up).")
 
 
 def _presets(flavor: str) -> str:
@@ -129,7 +130,7 @@ def _free_left() -> str:
                 "after it)")
     return "today's free quota"
 _SCRIPTS = ("SCRIPTS: Blender 5.0 bpy; build only the scene, camera and frame range (the service sets resolution and samples); "
-            "`import jr_assets; jr_assets.frame_camera()` fits every object; a script that raises fails the job (nothing rendered).")
+            "`import jr_assets; jr_assets.frame_camera()` fits every object; a script that raises renders nothing.")
 _MORE = f"Examples and docs: {SITE}/examples, {SITE}/llms.txt"
 
 
@@ -185,12 +186,12 @@ def descriptions(flavor: str, wait_cap_s: int = 50) -> dict[str, str]:
            "Returns job_id and estimate.human at once; then render_status until done, and render_download(job_id, only='mp4') "
            "for the video. ")
         + _NOT_HERE + " Use the same environment and blender as the approved preview. "
-        f"Limits: up to {FREE_MAX_FRAMES} frames and {FREE_MAX_SIZE} per job; {_free_offer()}; ask the user before a render "
-        "that the free time does not cover; files are "
-        f"deleted {RETENTION_HOURS} hours after last use. preset='product' | 'turntable' | 'animation' fills the settings for "
-        "that purpose. Options on each parameter: output (png, exr, mp4, webm, prores, gif, webp), orbit=True (turntable MP4), "
-        f"cameras (one image per named camera), transparent, engine='eevee', notify_url (one POST when done), {a} with "
-        "'polyhaven:<id>'. "
+        # 10/13: ここが「240 コマ・1080p まで」のままで、AI が長いアニメや 4K を断っていた（受付は 3,000 コマ・先払いは 4K）
+        f"Limits: up to {MAX_FRAMES_PER_INSTRUCTION:,} frames per call (over {FREE_MAX_FRAMES} runs as parts under one job id); "
+        f"{FREE_MAX_SIZE} on free time, {PAID_MAX_SIZE} with prepaid credit; {_free_offer()}; ask before a render the "
+        f"free time does not cover; files are deleted {RETENTION_HOURS} h after last use ({PAID_RETENTION_DAYS} days once topped "
+        "up). preset='product' | 'turntable' | 'animation' fills the settings for that purpose. Options per parameter: "
+        f"output (png, exr, mp4, webm, prores, gif, webp), orbit, cameras, transparent, engine, notify_url, {a} ('polyhaven:<id>'). "
         + ("Pass scene_id from render_preview, or scene_script / scene_url for a new scene." if remote else
            "Pass scene_id from render_preview, or scene_path / scene_script / scene_url; files are saved locally."))
     scene_info = (
@@ -212,7 +213,8 @@ def descriptions(flavor: str, wait_cap_s: int = 50) -> dict[str, str]:
            "call render_download once status is done."))
     if remote:
         download = (
-            "Get the download links of a job (no key needed, valid about 24 hours) and show them to the user; in a terminal, "
+            "Get the download links of a job (no key needed, valid about 24 hours; call again for fresh links while the files are "
+            f"kept, {RETENTION_HOURS} h after last use or {PAID_RETENTION_DAYS} days for keys that have topped up) and show them to the user; in a terminal, "
             "fetch them with curl -o. A finished preview also comes back as an inline image. wait_seconds > 0 first waits for "
             f"the job (the service waits at most {wait_cap_s} s per call; call again while it is still running). only picks "
             "the files: 'output' (default: the MP4 or the preview sheet, plus the first frames), 'mp4', 'sheet', 'frames' or "

{janction_render-0.4.83 → janction_render-0.4.87}/janction_render.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: janction-render
-Version: 0.4.83
+Version: 0.4.87
 Summary: Cloud GPU rendering for Blender scenes and 3D models (GLB, FBX, USD, OBJ, STL), built for AI agents: MCP server (remote + stdio), HTTP API and CLI. Product shots, 360-degree turntables, animations and stills; preview in seconds with a critic; no local Blender or GPU. A 100 JPY welcome credit for a new key (one per network every 30 days) and 400 JPY more for saving a card that is not charged, then 0.1 JPY per GPU-second.
 Author-email: "JasmyLab Inc." <contact@jasmylab.com>
 License-Expression: MIT
@@ -57,7 +57,7 @@ Claude Code, Codex, Cursor or any MCP client**: an MCP server (remote and stdio)
 JANCTION Render is a separate product from SmartRender (JasmyLab's distributed rendering for people at a desktop) and is
 not affiliated with Render.com or the Render Network. The numbers, as of 2026-10-08 (the primary source is
 https://render.janction.jp/facts.json): a 100 JPY welcome credit (1,000 GPU-seconds, 14 days, no card) for a new key, one per network every 30 days (further keys on the same network share what is left), plus 400 JPY more once a card is saved on a page that charges nothing (once per card); after that, previews
-are free up to 2 GPU-minutes a day and finals cost 0.1 JPY per GPU-second from prepaid credit (paid use started on 2026-10-08); finals up to 1,000 frames per instruction (run as parts of 240 under one job id); 1080p on free time, up to 4K on prepaid credit; inputs and results deleted
+are free up to 2 GPU-minutes a day and finals cost 0.1 JPY per GPU-second from prepaid credit (paid use started on 2026-10-08); finals up to 3,000 frames per instruction (run as parts of 240 under one job id); 1080p on free time, up to 4K on prepaid credit; inputs and results deleted
 24 hours after last use (7 days for keys that have topped up).
 
 Official site: https://render.janction.jp · MCP endpoint: `https://render.janction.jp/mcp` · Fact sheet: https://render.janction.jp/facts
@@ -116,7 +116,7 @@ Keep 3D files in a repository? Add a button that lets anyone render them from th
   errors with the fix, spending caps per key, safe retries with `Idempotency-Key`.
 - **3D files and turntables.** glTF / GLB, FBX, USD, OBJ, STL, PLY, Alembic are imported into an empty scene with a camera and
   HDRI lighting; `orbit=True` makes a turntable. Fixed-price outcomes: `POST /v1/outcomes/turntable` and `/product-shot`.
-- **Honest limits.** a 100 JPY welcome credit for a new key with no card (14 days; one per network every 30 days) and 400 JPY more after saving a card that is not charged, then 0.1 JPY per GPU-second (previews free up to 2 GPU-minutes a day), finals up to 1,000 frames per instruction as parts of 240 (1080p on free time, up to 4K on prepaid credit), one GPU shared with another
+- **Honest limits.** a 100 JPY welcome credit for a new key with no card (14 days; one per network every 30 days) and 400 JPY more after saving a card that is not charged, then 0.1 JPY per GPU-second (previews free up to 2 GPU-minutes a day), finals up to 3,000 frames per instruction as parts of 240 (1080p on free time, up to 4K on prepaid credit), one GPU shared with another
   workload (jobs can wait; the ETA says so). Inputs and results are deleted 24 hours after last use (7 days for keys that have topped up) and never used for training.
 - **Not Render.com.** Same word, different product.
 
@@ -421,7 +421,7 @@ nothing adds 400 JPY more, once per card (up to 500 JPY in all). After it is use
 free up to 2 GPU-minutes a day and finals cost 0.1 JPY per
 GPU-second from prepaid credit, bought by card at a Stripe Checkout page (from 500 JPY; 2,000 JPY gives 2,200 JPY of credit,
 5,000 gives 5,750, 10,000 gives 12,000; the first top-up in the welcome period counts 1.5x, bonus up to 500 JPY). Only the part
-the free time does not cover is charged. Finals go up to 1,000 frames per instruction (run as parts of 240 under one job id): 1080p on free time, up to 4K (3840x2160) on prepaid credit. `POST /v1/estimate` quotes what will be charged
+the free time does not cover is charged. Finals go up to 3,000 frames per instruction (run as parts of 240 under one job id): 1080p on free time, up to 4K (3840x2160) on prepaid credit. `POST /v1/estimate` quotes what will be charged
 (`cost.estimated_yen`), the currency and an expiry before any GPU time is used: https://render.janction.jp/pricing
 
 ## Troubleshooting

{janction_render-0.4.83 → janction_render-0.4.87}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "janction-render"
-version = "0.4.83"
+version = "0.4.87"
 description = "Cloud GPU rendering for Blender scenes and 3D models (GLB, FBX, USD, OBJ, STL), built for AI agents: MCP server (remote + stdio), HTTP API and CLI. Product shots, 360-degree turntables, animations and stills; preview in seconds with a critic; no local Blender or GPU. A 100 JPY welcome credit for a new key (one per network every 30 days) and 400 JPY more for saving a card that is not charged, then 0.1 JPY per GPU-second."
 readme = "README.md"
 license = "MIT"