Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.
JavaScript 3d-generationai-agentsanimationavatardigital-humangenerative-aiglbgltfimage-to-3dllmmcpmcp-servermodel-context-protocolsolanatext-to-3dthreejsweb-componentswebglwebxrx402
fix(render): accept rigged avatars up to 25 MB so X avatar replies can attach their poster
nirholas11914b4c84f6
diff --git a/api/render/glb.js b/api/render/glb.jsindex 7b1202255..24d0e9d41 100644--- a/api/render/glb.js+++ b/api/render/glb.js@@ -30,7 +30,7 @@// - `background` must parse as a CSS color. It is interpolated into the// render page, and that page has container network egress, so markup in// this field would be script execution with an internal-network view.-// - GLB HEAD-fetched first to enforce a 10 MB cap before chromium boots.+// - GLB HEAD-fetched first to enforce a 25 MB cap before chromium boots.// - Distributed IP rate limit (60 renders / 10 min / IP, shared with// /api/render/avatar-clip) to keep chromium costs bounded under abuse.@@ -44,7 +44,7 @@ export const maxDuration = 30;const MAX_DIM = 2048;const MIN_DIM = 64;-const MAX_GLB_BYTES = 10 * 1024 * 1024;+const MAX_GLB_BYTES = 25 * 1024 * 1024;export default wrap(async function handler(req, res) {if (cors(req, res, { methods: 'GET,POST,OPTIONS' })) return;diff --git a/prompts/finish/_context/x-grok-PROGRESS.md b/prompts/finish/_context/x-grok-PROGRESS.mdindex 9c8fa79c6..73de94f7a 100644--- a/prompts/finish/_context/x-grok-PROGRESS.md+++ b/prompts/finish/_context/x-grok-PROGRESS.md@@ -181,3 +181,9 @@ Added `agent-get-started`, `daily-3d-brief`, `asset-pack`, `avatar-from-photo` t- Shipped `api/_lib/x-mention-make.js`: studio moderation, then the free gpt-forge text-to-3D lane, a bounded wait (`X_MAKE_BUDGET_MS`, default 150s), and one recorded end state per mention: reply (rendered PNG plus `/m/<creation>` link), pending (job id kept on the row, answered once by `finishPendingMakes`, with `X_MAKE_GIVE_UP_MS` default 30 min), the designed failure reply with a prefilled `/forge?prompt=` link, or a fixed refusal. Creations belong to the system bot account (`users.service_account`), are `unlisted`, and carry `forge_creations.x_author_id` (migration `20261009201500_forge_creations_x_author.sql`, applied) for order 058. Live delivery goes through an injected `deliver`; without one a live row is held `paused`.- Evidence: `prompts/x-grok/_generated/make-dry-run.json` (script `scripts/x-mention-make-dry-run.mjs`): 5 prompts, 6 real production generations, all reply rows, every poster PNG and `/m/` link answered 200. Three outlived the wait budget and were answered by the follow-up tick. `tests/x-mention-make.test.js` 15 tests (success, timeout then follow-up incl. overlapping ticks, failure, unsafe) on the real migrations via PGlite.- Code landed in sweep commits `9997cdcd4` and `607229b22` by a concurrent agent; the close-out commit carries this log and the order deletion. Observed: the free lane took 4 to 7 minutes per model today, so the 150s default usually ends pending; the follow-up tick must run (order 050's cron should call `finishPendingMakes`).++## 2026-10-09: order 054 (avatar from profile) built, one done line waiting on a deploy++- Shipped `api/_lib/x-mention-avatar.js` (own profile image only, `_normal` upgraded to 400x400, vision moderation, image-to-3D then auto-rig, bounded wait plus `finishPendingAvatars` follow-up, `default_avatar` reply with the `/create` link), wired into `x-mention-poll.js`. Tests: `tests/x-mention-avatar.test.js` 14 passing, including own-image-only and the named-other-account case. Doc: `docs/x-mention-avatar.md`. Code landed in sweep commit `9997cdcd4`.+- Evidence: `prompts/x-grok/_generated/avatar-dry-run.json`: 3 reply rows from real public profile images (profile image fetch 200, pose link 200) plus a default-image row.+- Remaining line: media status codes. Production `/api/render/glb` caps input at 10 MB and the rigged avatars are 15 to 16 MB, so the poster answers 413 there. This commit raises the cap to 25 MB (the renderer's own default); against a local server the same rigged GLB renders 200 image/png. After the next deploy, rerun `node scripts/x-mention-avatar-dry-run.mjs` to record production media 200s, then delete the order file. Owner: whoever runs the deploy (gated).