From dee6566ba29ba5978381b6fe35855f8bb31eefb0 Mon Sep 17 00:00:00 2001 From: Daniel-rmc <2459944653@qq.com> Date: Wed, 28 Oct 2026 17:00:37 +0800 Subject: [PATCH] docs: replace homepage model results with supported policy configurations --- site/README.md | 14 ++-- site/app.js | 172 ++++++++++++++++++++++-------------------------- site/index.html | 42 ++++++++++-- site/styles.css | 39 ++++++----- 4 files changed, 147 insertions(+), 120 deletions(-) diff --git a/site/README.md b/site/README.md index 48b8f8e..78e5220 100644 --- a/site/README.md +++ b/site/README.md @@ -22,11 +22,11 @@ The framework section distinguishes four environment adapters from the selected `assets/workspace/current-workspace-manual.png` and `current-workspace-model.png` are real screenshots of the current local application using the built-in Panda tabletop scene. They show the separate manual and model panels, retain their original aspect ratios, and open at full size. They are static presentation assets; the website does not execute their controls. The local application also has a separate scene-settings panel and a persistent stop control. The interactive `playground/` remains a viewer for a recorded LIBERO episode. -The Jev description refers to text primitive selection from explicit instructions and public proprioception. It uses a separate TypeSafe API key, receives no images, and delegates unit conversion and motion execution to local code. It is not presented as visual localization or autonomous grasping, and is not added to the recorded model-result table. +The Jev description refers to text primitive selection from explicit instructions and public proprioception. It uses a separate TypeSafe API key, receives no images, and delegates unit conversion and motion execution to local code. It is not presented as visual localization or autonomous grasping, and appears in the model connection guide with that input boundary. New presentation copy uses the existing `app.js` English/Chinese dictionaries and `data-i18n` attributes. `showcase.js`, the three recorded skill videos, their JSON manifests, and the playground remain independent of this presentation update. -## Episode and model results +## Recorded episodes The homepage plays a muted MP4 preview with synchronized, saved model action notes. Play/pause, previous/next decision and the time slider let visitors inspect the requested command and the **preceding** action's feedback. The preview uses simulation time; model waiting time is omitted. It is recorded output, not live inference. Reduced-motion preferences start with a still poster, and a failed autoplay attempt keeps a manual play control. Off-screen video pauses; a visitor's explicit pause is preserved. @@ -46,7 +46,13 @@ All media paths are relative to the site root. English action notes preserve the A success label requires agreement between the run summary and the verified independent LIBERO evaluator. The complete video also requires the initial frame plus every native control step, a complete terminal record and matching frame rate and simulation time. Changing the website language does not restart video playback. -`assets/model-comparison.json` contains one recorded example per model, with controller and observation settings, budgets, policy calls and available inference counts. These examples do not estimate success rates. Action scheduling and measurement boundaries differ, so the table is not a speed ranking. See [Results and configurations](https://github.com/Daniel-rmc/ManiLoop/blob/main/docs/RESULTS.md). +## Model connection guide + +The `#models` section contains six static, bilingual configuration cards: GPT via Codex, compatible vision APIs (hosted or self-hosted), SmolVLA, ACT, Diffusion Policy and Jev. Each explains connection requirements, policy inputs, operating scope and a documentation entry point. The section presents no performance table, model outcome ranking, inference counts or timings. + +Codex uses an official CLI login and a model ID currently available to that account. Vision APIs must support Responses, image input and JSON Schema structured output. The three local LeRobot presets use a separate `.venv-vla` with CPU / MPS / CUDA as available and remain LIBERO-specific; SmolVLA consumes language while the listed ACT and Diffusion checkpoints do not. Jev uses separate TypeSafe authentication and receives text plus public proprioception, without images. + +`assets/model-comparison.json` remains an unchanged historical artifact. The homepage no longer fetches or renders it; the original videos, episode manifests and result evidence are retained. Model-support copy is authored in the existing English/Chinese dictionaries rather than loaded from experiment output. Configuration details remain in [VLA](https://github.com/Daniel-rmc/ManiLoop/blob/main/docs/VLA.md), [Jev](https://github.com/Daniel-rmc/ManiLoop/blob/main/docs/JEV.md) and the [GPT guide](https://github.com/Daniel-rmc/ManiLoop/blob/main/docs/GPT6_DEMO.md). Published camera images and videos come from simulator rendering. ManiLoop source is MIT licensed; LIBERO, robosuite and their assets retain their respective licenses. @@ -70,7 +76,7 @@ Use a normal fast-forward push. `.nojekyll` keeps this a static site. A reviewed 首页短片按仿真时间静音循环,同步展示模型保存的动作说明、请求动作和上一步执行反馈;未展示模型等待时间。可暂停、前后切换决策或拖动进度,不是实时推理。系统选择减少动态效果时,默认展示静止画面;自动播放失败时可手动播放。任务卡来自 `assets/showcase.json`,只接收官方独立评分已经验证成功的回合。所有卡片明确说明是单次回合最终结果。 -模型结果为单次示例,不代表基准成功率。配置、动作块、预算与计时口径不同,不能由步数或耗时进行速度排名。完整运行条件见[实际结果](https://github.com/Daniel-rmc/ManiLoop/blob/main/docs/RESULTS.md),本地安装见[中文使用说明](https://github.com/Daniel-rmc/ManiLoop/blob/main/README.zh-CN.md)。 +“03 / 模型接入”以六张双语卡片介绍 GPT/Codex、兼容视觉 API(含自托管)、SmolVLA、ACT、Diffusion Policy 与 Jev 的连接、必要配置、输入和适用范围,不展示模型结果对比、调用数、耗时或胜负。三种本地预设使用独立 `.venv-vla`,按硬件选择 CPU / MPS / CUDA,仅适用于 LIBERO;SmolVLA 读取语言,ACT / DP 不读取语言。Jev 使用独立 TypeSafe Key,只接收文字与公开本体状态。原始 `model-comparison.json` 与视频证据保持不变,首页不再请求或渲染该 JSON。本地安装见[中文使用说明](https://github.com/Daniel-rmc/ManiLoop/blob/main/README.zh-CN.md)。 当前本地工作台截图使用内置 Panda 桌面场景,展示手动控制与模型运行分区,按原始比例显示,可点击查看大图。它们不是网页内嵌仿真;`playground/` 仍用于检查已录制的 LIBERO 回合。新后端的任务目录与接入检查不等于模型成功,既有 LeRobot 预设仍限 LIBERO。Jev 是独立 TypeSafe 认证的文本原语模式,不输入图像,不声称具备视觉定位或自主抓取能力。 diff --git a/site/app.js b/site/app.js index 142d1ae..c1126ea 100644 --- a/site/app.js +++ b/site/app.js @@ -99,20 +99,8 @@ "result_default": "Task completion is determined by the independent simulator evaluator.", "single_episode_note": "One episode does not establish a benchmark success rate.", "results_link": "Read the results", - "models_eyebrow": "03 / MODEL EXAMPLES", - "models_title": "What each model actually did.", - "comparison_scope": "Recorded examples on an official LIBERO task. Each row describes one run, with its own controller and inference setup.", - "comparison_aria": "Model example results; scroll horizontally to see all columns", - "comparison_caption": "Actual model runs with independently scored task results", - "col_model": "Model", - "col_result": "Official result", - "col_policy": "Policy calls", - "col_inference": "Inference calls", - "col_steps": "Control steps", - "col_setup": "Configuration and scope", - "comparison_loading": "Loading recorded model examples", - "comparison_caveat": "These are individual examples, not success-rate estimates. Budgets, action chunks and control methods differ; step counts and timings do not form a speed ranking.", - "comparison_results": "Check configurations and results", + "models_eyebrow": "03 / MODEL CONNECTIONS", + "models_title": "Supported models and configurations.", "start_eyebrow": "04 / GET STARTED", "start_title": "Run the loop on your machine.", "start_intro": "Start with the local workspace and manual control, then choose an optional backend and policy. Simulation runtimes and VLA inference stay isolated.", @@ -204,10 +192,6 @@ "copy_done": "Commands copied to the clipboard.", "copy_manually": "Copy manually", "copy_fallback": "Commands selected. Use your system copy shortcut.", - "comparison_unavailable": "Model examples could not be loaded. Read the result report below.", - "comparison_sim_time": "Simulation: {seconds} s", - "comparison_wall_time": "Wall time: {seconds} s", - "comparison_wall_time_unknown": "Wall time: not recorded", "environments_title": "Four environments. Explicit boundaries.", "environments_intro": "Choose the scene in the local workspace. Optional backends keep separate runtimes, controllers and independent task evaluators.", "env_builtin": "Built-in MuJoCo", @@ -240,7 +224,44 @@ "env_builtin_tag": "01 / TABLETOP", "env_libero_tag": "02 / BENCHMARK", "env_robosuite_tag": "03 / MANIPULATION", - "env_robocasa_tag": "04 / KITCHEN" + "env_robocasa_tag": "04 / KITCHEN", + "support_intro": "Connect an account-backed GPT model, a compatible vision API, a fixed local LeRobot preset, or Jev text primitives. Each route has its own authentication, observations and operating scope.", + "support_config_label": "Connection & configuration", + "support_input_label": "Policy inputs", + "support_scope_label": "Operating scope", + "support_guide": "Configuration guide", + "support_codex_tag": "ACCOUNT / CODEX CLI", + "support_codex_title": "GPT via Codex", + "support_codex_config": "Use a recent official Codex CLI, sign in with your ChatGPT account, and choose a model ID available to that account. ManiLoop delegates authentication to the CLI.", + "support_codex_input": "Task instruction, camera RGB, robot proprioception and permitted action feedback.", + "support_codex_scope": "Structured TCP actions in an installed simulation backend. Run text, image and action-format diagnostics for the selected model before controlling the robot.", + "support_api_tag": "HOSTED / SELF-HOSTED", + "support_api_title": "Compatible vision API", + "support_api_config": "Set the service base URL, model ID and API key. The endpoint must support Responses, image input and JSON Schema structured output.", + "support_api_input": "Task instruction, camera RGB, robot proprioception and permitted action feedback through the Responses adapter.", + "support_api_scope": "Hosted or self-hosted vision services implementing that protocol. Select an available model and verify all three capabilities in the local diagnostics.", + "support_smol_tag": "LOCAL / LANGUAGE-CONDITIONED", + "support_smol_title": "SmolVLA", + "support_smol_config": "Install or reuse the smolvla-libero preset and its pinned checkpoint in the separate .venv-vla inference environment.", + "support_smol_input": "Language instruction, two 256 × 256 RGB views and the 8D robot proprioceptive state.", + "support_smol_scope": "LIBERO Panda tasks with the preset observation processor and action protocol. Use the scene’s original English task instruction.", + "support_act_tag": "LOCAL / VISUAL IMITATION", + "support_act_title": "ACT", + "support_act_config": "Install or reuse the act-libero preset and its pinned community checkpoint in .venv-vla.", + "support_act_input": "Two 256 × 256 RGB views and the 8D robot proprioceptive state. This checkpoint does not consume language.", + "support_act_scope": "LIBERO only, with the checkpoint’s saved preprocessing and action chunks. Editing the text instruction does not change its task conditioning.", + "support_dp_tag": "LOCAL / VISUAL DIFFUSION", + "support_dp_title": "Diffusion Policy", + "support_dp_config": "Install or reuse the diffusion-libero preset and its pinned community checkpoint in .venv-vla.", + "support_dp_input": "Two 256 × 256 RGB views and the 8D robot proprioceptive state, updated each control step. This checkpoint does not consume language.", + "support_dp_scope": "LIBERO only, preserving the checkpoint’s observation history, action chunks and sampling configuration.", + "support_jev_tag": "TYPESAFE / TEXT PRIMITIVES", + "support_jev_title": "Jev", + "support_jev_config": "Provide a separate TypeSafe API key in the local workspace or TYPESAFE_API_KEY. No Jev weights or TypeSafe SDK installation is required.", + "support_jev_input": "Explicit text instructions and public robot proprioception. Images are not sent to Jev.", + "support_jev_scope": "Selects one bounded motion primitive per instruction in controlled timing. Local code handles units, coordinates and execution; this mode does not locate objects visually.", + "support_local_title": "Local preset requirements", + "support_local_note": "SmolVLA, ACT and Diffusion Policy share the isolated .venv-vla runtime and support CPU, Apple MPS or NVIDIA CUDA as available. Install the LIBERO backend separately. The adapter accepts these three catalog presets, not arbitrary checkpoint imports." }, "zh": { "skip": "跳至正文", @@ -339,20 +360,8 @@ "result_default": "任务是否完成由仿真器独立评分判定。", "single_episode_note": "单次演示不代表基准成功率。", "results_link": "查看实际结果", - "models_eyebrow": "03 / 模型示例", - "models_title": "各模型实际完成了什么。", - "comparison_scope": "官方 LIBERO 任务的单次运行示例。每行记录一轮运行及其控制器和推理配置。", - "comparison_aria": "模型运行结果,可横向滚动查看所有列", - "comparison_caption": "各模型实际运行及独立任务结果", - "col_model": "模型", - "col_result": "官方结果", - "col_policy": "策略调用数", - "col_inference": "推理调用数", - "col_steps": "控制步数", - "col_setup": "配置与适用范围", - "comparison_loading": "正在载入模型运行示例", - "comparison_caveat": "这些是单次示例,不能估计成功率。预算、动作块和控制方式不同,不能据步数或耗时进行速度排名。", - "comparison_results": "核对模型配置与结果", + "models_eyebrow": "03 / 模型接入", + "models_title": "支持的模型、配置与适用范围", "start_eyebrow": "04 / 开始使用", "start_title": "在你的电脑上运行闭环。", "start_intro": "先启动本地工作台和手动控制,再选择可选后端与策略。各仿真运行环境与 VLA 推理保持隔离。", @@ -444,10 +453,6 @@ "copy_done": "命令已复制到剪贴板。", "copy_manually": "请手动复制", "copy_fallback": "命令已选中,请使用系统复制快捷键。", - "comparison_unavailable": "模型示例暂时无法载入,请阅读下方结果文档。", - "comparison_sim_time": "仿真时间:{seconds} 秒", - "comparison_wall_time": "墙钟时间:{seconds} 秒", - "comparison_wall_time_unknown": "墙钟时间:未记录", "environments_title": "四类环境,明确各自的范围。", "environments_intro": "在本地工作台选择场景。可选后端分别维护运行环境、控制器与独立任务评分。", "env_builtin": "内置 MuJoCo", @@ -480,7 +485,44 @@ "env_builtin_tag": "01 / 桌面", "env_libero_tag": "02 / 基准任务", "env_robosuite_tag": "03 / 操作任务", - "env_robocasa_tag": "04 / 厨房" + "env_robocasa_tag": "04 / 厨房", + "support_intro": "接入账号支持的 GPT 模型、兼容视觉 API、固定的本地 LeRobot 预设,或 Jev 文本原语。每条路径分别配置认证、观测与适用范围。", + "support_config_label": "连接与配置", + "support_input_label": "策略输入", + "support_scope_label": "适用范围", + "support_guide": "配置说明", + "support_codex_tag": "账号 / CODEX CLI", + "support_codex_title": "GPT / Codex 登录", + "support_codex_config": "使用近期官方 Codex CLI 登录 ChatGPT 账号,选择该账号当前支持的 model ID。认证由 CLI 处理。", + "support_codex_input": "任务指令、相机 RGB、机器人本体状态与允许的动作反馈。", + "support_codex_scope": "通过结构化 TCP 动作连接已安装的仿真后端。先对所选模型完成文字、图像与动作格式诊断,再执行控制。", + "support_api_tag": "托管 / 自托管服务", + "support_api_title": "兼容视觉 API", + "support_api_config": "填写服务基础地址、model ID 和 API Key。接口须同时支持 Responses、图像输入和 JSON Schema 结构化输出。", + "support_api_input": "通过 Responses 适配器提供任务指令、相机 RGB、机器人本体状态与允许的动作反馈。", + "support_api_scope": "适用于实现上述协议的托管或自托管视觉服务。选择服务可用模型,并通过本地诊断核对三项能力。", + "support_smol_tag": "本地 / 语言条件策略", + "support_smol_title": "SmolVLA", + "support_smol_config": "在独立 .venv-vla 推理环境中安装或复用 smolvla-libero 预设及其固定检查点。", + "support_smol_input": "语言指令、两路 256 × 256 RGB 图像与 8 维机器人本体状态。", + "support_smol_scope": "用于采用预设观测处理器与动作协议的 LIBERO Panda 任务。使用场景原有的英文任务指令。", + "support_act_tag": "本地 / 视觉模仿策略", + "support_act_title": "ACT", + "support_act_config": "在 .venv-vla 中安装或复用 act-libero 预设及其固定的社区检查点。", + "support_act_input": "两路 256 × 256 RGB 图像与 8 维机器人本体状态。该检查点不读取语言。", + "support_act_scope": "仅适用于 LIBERO,沿用检查点保存的预处理和动作块。修改文本指令不会改变它的任务条件。", + "support_dp_tag": "本地 / 视觉扩散策略", + "support_dp_title": "Diffusion Policy", + "support_dp_config": "在 .venv-vla 中安装或复用 diffusion-libero 预设及其固定的社区检查点。", + "support_dp_input": "每个控制步更新的两路 256 × 256 RGB 图像与 8 维机器人本体状态。该检查点不读取语言。", + "support_dp_scope": "仅适用于 LIBERO,保留检查点的观测历史、动作块和采样配置。", + "support_jev_tag": "TYPESAFE / 文本原语", + "support_jev_title": "Jev", + "support_jev_config": "在本地工作台或 TYPESAFE_API_KEY 中提供独立的 TypeSafe API Key,无需下载 Jev 权重或安装 TypeSafe SDK。", + "support_jev_input": "明确的文字指令与公开机器人本体状态,不向 Jev 发送图像。", + "support_jev_scope": "在受控时序下为每条指令选择一个有界动作原语。单位、坐标与执行由本地代码处理;此模式不进行视觉物体定位。", + "support_local_title": "本地预设要求", + "support_local_note": "SmolVLA、ACT 与 Diffusion Policy 使用隔离的 .venv-vla 运行环境,按硬件选择 CPU、Apple MPS 或 NVIDIA CUDA;LIBERO 后端另行安装。当前适配器接受模型目录中的这三个预设,不提供任意检查点导入。" } }; const byId = (id) => document.getElementById(id); @@ -492,9 +534,7 @@ const videoPath = /^assets\/episodes\/[a-zA-Z0-9_/-]+\.mp4$/; let language = initialLanguage(); let recording = null; - let comparison = null; let recordingFailed = false; - let comparisonFailed = false; let videoFailed = false; let frames = []; let index = 0; @@ -546,8 +586,6 @@ updatePlaybackLabels(); if (recording) { renderRecordingMetadata(recording); renderFrame(index); } if (recordingFailed) renderRecordingError(); - if (comparison) renderComparison(comparison); - if (comparisonFailed) renderComparisonError(); document.dispatchEvent(new CustomEvent("maniloop:language", { detail: { language } })); if (persist) { try { localStorage.setItem("maniloop-language", language); } catch { /* Switching still works without storage. */ } @@ -722,55 +760,6 @@ } catch { recordingFailed = true; renderRecordingError(); } } - function cell(row, text, className) { - const element = document.createElement("td"); - if (className) element.className = className; - element.textContent = text; - row.appendChild(element); - return element; - } - - function renderComparison(data) { - byId("comparison-scope").textContent = localized(data.scope); - byId("comparison-caveat").textContent = localized(data.caveat); - const body = byId("comparison-rows"); body.replaceChildren(); - for (const item of data.rows) { - const row = document.createElement("tr"); - cell(row, item.model, "comparison-model"); - const result = cell(row, localized(item.status), "comparison-result"); - result.dataset.result = item.success === true ? "success" : item.success === false ? "failure" : "unknown"; - cell(row, metric(item.decisions), "comparison-number"); - cell(row, metric(item.inference_calls), "comparison-number"); - cell(row, metric(item.control_steps), "comparison-number"); - const details = cell(row, "", "comparison-details"); - for (const [text, className] of [[localized(item.conditions), "comparison-conditions"], [localized(item.notes), "comparison-notes"]]) { - const paragraph = document.createElement("p"); paragraph.className = className; paragraph.textContent = text; details.appendChild(paragraph); - } - const timings = document.createElement("p"); timings.className = "comparison-timings"; - timings.textContent = t("comparison_sim_time", { seconds: metric(item.simulation_seconds, 2) }) + " · " + (nonnegativeNumber(item.wall_seconds) ? t("comparison_wall_time", { seconds: metric(item.wall_seconds, 2) }) : t("comparison_wall_time_unknown")); - details.appendChild(timings); body.appendChild(row); - } - } - - function renderComparisonError() { - const body = byId("comparison-rows"); body.replaceChildren(); - const row = document.createElement("tr"); const text = cell(row, t("comparison_unavailable")); text.colSpan = 6; body.appendChild(row); - } - - async function loadComparison() { - try { - const response = await fetch("assets/model-comparison.json?v=bilingual-workspace-1", { credentials: "omit" }); - if (!response.ok) throw new Error("Comparison unavailable"); - const data = await response.json(); - if (!Array.isArray(data.rows) || !data.rows.length || data.rows.length > 30 || - !localizedText(data.scope) || !localizedText(data.caveat) || !data.rows.every((row) => - row && typeof row.model === "string" && localizedText(row.status) && localizedText(row.conditions) && localizedText(row.notes))) { - throw new Error("Invalid comparison"); - } - comparison = data; renderComparison(data); - } catch { comparisonFailed = true; renderComparisonError(); } - } - previous.addEventListener("click", () => selectFrame(index - 1)); next.addEventListener("click", () => selectFrame(index + 1)); playButton.addEventListener("click", () => setPlaying(timer === null && frames.length > 1)); @@ -801,5 +790,4 @@ } setLanguage(language, ["en", "zh"].includes(new URLSearchParams(window.location.search).get("lang"))); loadRecording(); - loadComparison(); })(); diff --git a/site/index.html b/site/index.html index 238d9fd..51a6fba 100644 --- a/site/index.html +++ b/site/index.html @@ -11,8 +11,8 @@ ManiLoop — See. Act. Observe again. - - + + @@ -124,10 +124,40 @@

See.

Loading run conditions
Task completion is determined by the independent simulator evaluator. One episode does not establish a benchmark success rate.

Read the results
-

03 / MODEL EXAMPLES

What each model actually did.

Recorded examples on an official LIBERO task. Each row describes one run, with its own controller and inference setup.

-
Actual model runs with independently scored task results
ModelOfficial resultPolicy callsInference callsControl stepsConfiguration and scope
Loading recorded model examples
-

These are individual examples, not success-rate estimates. Budgets, action chunks and control methods differ; step counts and timings do not form a speed ranking.

- Check configurations and results +

03 / MODEL CONNECTIONS

Supported models and configurations.

Connect an account-backed GPT model, a compatible vision API, a fixed local LeRobot preset, or Jev text primitives. Each route has its own authentication, observations and operating scope.

+
+

ACCOUNT / CODEX CLI

GPT via Codex

+
Connection & configuration
Use a recent official Codex CLI, sign in with your ChatGPT account, and choose a model ID available to that account. ManiLoop delegates authentication to the CLI.
+
Policy inputs
Task instruction, camera RGB, robot proprioception and permitted action feedback.
+
Operating scope
Structured TCP actions in an installed simulation backend. Run text, image and action-format diagnostics for the selected model before controlling the robot.
+
Configuration guide
+

HOSTED / SELF-HOSTED

Compatible vision API

+
Connection & configuration
Set the service base URL, model ID and API key. The endpoint must support Responses, image input and JSON Schema structured output.
+
Policy inputs
Task instruction, camera RGB, robot proprioception and permitted action feedback through the Responses adapter.
+
Operating scope
Hosted or self-hosted vision services implementing that protocol. Select an available model and verify all three capabilities in the local diagnostics.
+
Configuration guide
+

LOCAL / LANGUAGE-CONDITIONED

SmolVLA

+
Connection & configuration
Install or reuse the smolvla-libero preset and its pinned checkpoint in the separate .venv-vla inference environment.
+
Policy inputs
Language instruction, two 256 × 256 RGB views and the 8D robot proprioceptive state.
+
Operating scope
LIBERO Panda tasks with the preset observation processor and action protocol. Use the scene’s original English task instruction.
+
Configuration guide
+

LOCAL / VISUAL IMITATION

ACT

+
Connection & configuration
Install or reuse the act-libero preset and its pinned community checkpoint in .venv-vla.
+
Policy inputs
Two 256 × 256 RGB views and the 8D robot proprioceptive state. This checkpoint does not consume language.
+
Operating scope
LIBERO only, with the checkpoint’s saved preprocessing and action chunks. Editing the text instruction does not change its task conditioning.
+
Configuration guide
+

LOCAL / VISUAL DIFFUSION

Diffusion Policy

+
Connection & configuration
Install or reuse the diffusion-libero preset and its pinned community checkpoint in .venv-vla.
+
Policy inputs
Two 256 × 256 RGB views and the 8D robot proprioceptive state, updated each control step. This checkpoint does not consume language.
+
Operating scope
LIBERO only, preserving the checkpoint’s observation history, action chunks and sampling configuration.
+
Configuration guide
+

TYPESAFE / TEXT PRIMITIVES

Jev

+
Connection & configuration
Provide a separate TypeSafe API key in the local workspace or TYPESAFE_API_KEY. No Jev weights or TypeSafe SDK installation is required.
+
Policy inputs
Explicit text instructions and public robot proprioception. Images are not sent to Jev.
+
Operating scope
Selects one bounded motion primitive per instruction in controlled timing. Local code handles units, coordinates and execution; this mode does not locate objects visually.
+
Configuration guide
+
+

Local preset requirements

SmolVLA, ACT and Diffusion Policy share the isolated .venv-vla runtime and support CPU, Apple MPS or NVIDIA CUDA as available. Install the LIBERO backend separately. The adapter accepts these three catalog presets, not arbitrary checkpoint imports.

04 / GET STARTED

Run the loop on your machine.

Start with the local workspace and manual control, then choose an optional backend and policy. Simulation runtimes and VLA inference stay isolated.

diff --git a/site/styles.css b/site/styles.css index ea9b3ec..e2ccdf1 100644 --- a/site/styles.css +++ b/site/styles.css @@ -205,22 +205,17 @@ button[disabled] { cursor: default; opacity: .35; } .result-context .text-link { flex: 0 0 auto; color: #e9c09d; font-size: 11px; } .result-context .text-link:hover { color: white; } .models-section { border-bottom: 0; } -.comparison-scroll { width: 100%; overflow-x: auto; border: 1px solid var(--line); scrollbar-width: thin; } -.comparison-table { width: 100%; min-width: 860px; border-collapse: collapse; text-align: left; font-size: 12px; } -.comparison-table th { padding: 15px 17px; background: #e8ebe0; color: #5c6454; font-size: 10px; font-weight: 600; letter-spacing: .02em; } -.comparison-table td { vertical-align: top; padding: 23px 17px; border-top: 1px solid var(--line); } -.comparison-table tbody tr:nth-child(even) { background: #eef0e6; } -.comparison-model { min-width: 130px; font-size: 13px; font-weight: 650; } -.comparison-result { min-width: 118px; font-size: 11px; } -.comparison-result[data-result="success"] { color: #3b6334; } -.comparison-result[data-result="failure"] { color: #9c4b32; } -.comparison-number { font: 14px/1.7 var(--mono); font-variant-numeric: tabular-nums; } -.comparison-details { min-width: 330px; max-width: 500px; font-size: 10px; line-height: 1.9; } -.comparison-conditions { color: #3b4334; } -.comparison-notes { color: var(--muted); margin-top: 8px; } -.comparison-timings { margin-top: 10px; color: #626d56; font: 9px/1.8 var(--mono); } -.comparison-caveat { max-width: 1040px; color: var(--muted); font-size: 11px; line-height: 1.95; margin-top: 20px; } -.comparison-results { margin-top: 15px; } +.model-support-grid { display: grid; grid-template-columns: repeat(3, minmax(0, 1fr)); gap: 22px; } +.model-support-card { display: flex; flex-direction: column; min-width: 0; padding: 27px 25px; border: 1px solid var(--line); background: #eeeee5; } +.model-support-card h3 { margin-top: 15px; font-size: 23px; line-height: 1.4; font-weight: 580; letter-spacing: -.025em; } +.model-support-card dl { margin: 23px 0 24px; } +.model-support-card dl > div + div { margin-top: 20px; padding-top: 17px; border-top: 1px solid var(--line); } +.model-support-card dt { color: var(--accent); font-size: 10px; line-height: 1.7; font-weight: 550; } +.model-support-card dd { margin: 7px 0 0; color: var(--muted); font-size: 12px; line-height: 1.9; overflow-wrap: anywhere; } +.model-support-card .text-link { align-self: start; margin-top: auto; font-size: 11px; } +.model-support-note { margin-top: 24px; padding: 20px 24px; border-left: 2px solid #b6beaa; background: #ebede3; } +.model-support-note h3 { font-size: 15px; line-height: 1.6; font-weight: 580; } +.model-support-note p { margin-top: 8px; color: var(--muted); font-size: 11px; line-height: 1.9; } .start-section { border-top: 1px solid var(--line); } .install-grid { display: grid; grid-template-columns: repeat(2, minmax(0, 1fr)); gap: 25px; } .install-card { border: 1px solid var(--line); padding: 25px; min-width: 0; } @@ -422,8 +417,6 @@ button[disabled] { cursor: default; opacity: .35; } .run-metrics .metric-result strong { font-size: 26px; } .run-metrics span { font-size: 9px; } .result-context p { font-size: 10px; } - .comparison-caveat { font-size: 10px; } - .comparison-results { font-size: 11px; } .install-card { padding: 21px 18px; } .install-heading { gap: 10px; } .install-heading h3 { font-size: 17px; } @@ -540,3 +533,13 @@ html[lang="en"] .run-metrics .metric-result strong { font-size: clamp(20px, 2.2v .optional-backends { display: block; padding: 21px 18px; } .optional-backend-links { margin-top: 16px; } } + +@media (max-width: 1050px) { + .model-support-grid { grid-template-columns: repeat(2, minmax(0, 1fr)); gap: 18px; } +} +@media (max-width: 560px) { + .model-support-grid { grid-template-columns: 1fr; gap: 17px; } + .model-support-card { padding: 24px 22px; } + .model-support-card h3 { font-size: 22px; } + .model-support-note { padding: 19px 20px; } +}