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
docs(orders): record a payment-free route to the OKX reply lane, and the measured i18n throughput
nirholasbb63ec98f7cb
diff --git a/prompts/finish/311-home-17-a11y-i18n-mobile.md b/prompts/finish/311-home-17-a11y-i18n-mobile.mdindex 48a70fd60..354352d6e 100644--- a/prompts/finish/311-home-17-a11y-i18n-mobile.md+++ b/prompts/finish/311-home-17-a11y-i18n-mobile.md@@ -50,7 +50,17 @@ one payment.Until it clears, the free lanes make real progress and should be left grinding: the run isresumable, idempotent, and an incomplete locale is held out of `manifest.json` rather than shipped-half-translated, so stopping mid-way is safe. When it lands: confirm `npm run i18n:lint` is clean,+half-translated, so stopping mid-way is safe.++**Measured throughput, so nobody re-derives it.** Turning off the reasoning model's chain of+thought is worth more than any amount of parallelism and was the single biggest win+(`d14497d11`): on a five-string chunk, 40 output tokens in 1.1s with thinking off against+hundreds of tokens and tens of seconds with it on. Even so, NVIDIA's free tier will not absorb a+bulk run. Six processes collapsed into `429 Too Many Requests`; three at `--concurrency=4` still+drew `429` plus `503 Service temporarily overloaded`, and a chunk that fails takes its whole+locale's pass with it, so an over-parallel run can spend an hour and finish nothing. One process+at `--concurrency=2` is the sustainable shape. Budget on the order of a day of wall clock for the+remaining catalog on free lanes, against roughly one pass on Vertex. When it lands: confirm `npm run i18n:lint` is clean,delete the `publishLocale` helper in `tests/e2e/home-a11y.spec.js` (the RTL and never-translatetests can read the real manifest once a locale is complete), re-run the suite, and retire thisfile per the section at the bottom.diff --git a/prompts/finish/907-backlog-08-okx-chat-bot-always-on.md b/prompts/finish/907-backlog-08-okx-chat-bot-always-on.mdindex 8623ccdce..3d5506630 100644--- a/prompts/finish/907-backlog-08-okx-chat-bot-always-on.md+++ b/prompts/finish/907-backlog-08-okx-chat-bot-always-on.md@@ -94,6 +94,37 @@ Order does not matter; whichever lands second is picked up by the next 15-minuteAn email OTP is **not** currently needed: the session has been `loggedIn: true` since2026-09-05.+## A second route to the reply lane, found 2026-09-10 (owner's call, not taken)++Owner action 1 (clear the GCP billing hold) is still the clean fix, and it now unblocks four+orders rather than two, so it is worth doing on its own merits. But it is no longer the ONLY+route to a working reply lane, and the alternative needs no payment:++`providerLanes()` in [config.js](../../workers/okx-chat-bot/config.js) already carries an+`anthropic-gateway` lane that accepts any Anthropic-wire-format endpoint via+`ANTHROPIC_BASE_URL` + `ANTHROPIC_AUTH_TOKEN`. **three.ws serves exactly that shape itself** at+`/api/llm/anthropic`, the we-pay proxy the avatar embeds already use, and its free NVIDIA rungs+were repaired on 2026-09-10 (`4dbe02cc7`); before that they were retired ids answering `410 Gone`,+so this route would not have worked even if someone had tried it. Pointing the bot at our own+proxy bills no third party and needs no Google lane.++Three things gate it, and none is a code change:++1. **The API deploy carrying `4dbe02cc7` has to land first**, or the proxy's NVIDIA rungs are+ still the dead ids.+2. **The proxy meters per agent.** It answers `400 agent query param required`, so this needs an+ agent id owned by the platform user, and the i18n lane's own docs warn never to point it at+ someone else's agent. Which agent to meter is an owner decision. Its embed-policy default is+ 10 requests/min, which is ample for chat but is a real ceiling.+3. **Electing a lane respawns the daemon.** That is the documented mechanism (an env overlay plus+ a respawn), and this bot's identity, the wallet keyring plus the XMTP client database, is a+ single-writer state object whose corruption costs a human email OTP to recover. That risk is+ why this was left for the owner rather than applied from a session.++Not verified end to end from here, deliberately: proving it would mean metering a real agent and+respawning the live daemon. What IS verified is that the gateway lane exists, that+`/api/llm/anthropic` speaks the Anthropic wire format, and that its free chain answers again.+## Definition of done- [ ] **Chat delivery verified end to end with a real inbound message.** The inbound half