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(nvidia): add the agent-3d embed as a second Inception product record
nirholas49670bcf0034
diff --git a/CHANGELOG.md b/CHANGELOG.mdindex cd10efe59..82a53451d 100644--- a/CHANGELOG.md+++ b/CHANGELOG.md@@ -15,8 +15,10 @@ Public history for [three.ws](https://three.ws), newest first. New pages come fr- **A batched order on the OKX Agent Identity Studio can no longer be paid at the price of one identity**: The Agent Identity Studio accepts several calls in a single request, and its price is the sum of every paid call in that request. The OKX X Layer payment option, the one an OKX buyer's tooling selects first, was quoting a flat one-identity price no matter how many identities the request actually ordered, so a multi-identity order could be paid at the price of one. Every payment option on that endpoint now quotes the same total, and a request with nothing chargeable in it still quotes the normal list price. Two smaller fixes ride along: a paid response now lets browser-based buyers read the settlement receipt header it has always sent, and the free catalog and health endpoints answer errors in the same shape as the paid endpoints beside them, with an Allow header naming the methods they take. (`/docs/okx-marketplace`) `[security, fix]`- **A bounty that does not exist now says so, instead of blaming pump.fun**: The bounty board at /bounties mirrors the public pump.fun GO board. When you opened a bounty link that had been deleted or mistyped, the page reported it the same way it reports pump.fun being down, so there was no way to tell a dead link from an outage. A bounty that genuinely is not there now answers as not found, and a real upstream problem keeps its own separate answer. A second, quieter case got fixed with it: when the submissions list for a bounty failed to load, the page showed the bounty with zero submissions and said nothing, which looks identical to a bounty nobody has entered. That failure is now recorded, so a pump.fun outage shows up as an outage rather than as a board that suddenly looks unpopular. (`/bounties`) `[fix]`- **A database blip during a scheduled job now reads as a database blip, not a healthy run**: Two of the scheduled jobs behind the trading fleet could fail without anything looking wrong. The job that reallocates budget across strategies talks to the database over a socket, and when that socket never opened the failure arrived as an object carrying no message at all. The job caught it, answered with an empty error string, and the scheduler recorded a success, so the health record stayed green while the loop did nothing. Failures of that kind are now recognised as what they are, a database that could not be reached, so the run is marked unhealthy and retried on the next tick instead of being logged as fine. Every page and endpoint that reads from the database inherits the same correction: this class of outage now returns a retryable service-unavailable response rather than an internal error. The fleet watchdog also gained a seventh check, over the calibration pass that measures whether a high-conviction coin actually wins as often as its score claims, so that loop can no longer go quiet unnoticed. `[fix, infra]`+- **A failed database migration can no longer freeze the schema in silence**: Every deploy carries database migrations that must run before the new code does. When one of them failed, the runner printed a single line with the database's own error and stopped, naming neither the file that failed nor the migrations queued behind it. Those stayed unapplied and unmentioned, so the site ended up serving code that reads columns the database never got, and the first visible symptom was an unrelated-looking error on a live page. The runner now names the failing file, how many ran before it, and every migration still waiting, and the deploy step that catches a database running behind the code is now part of the documented deploy path instead of something you had to remember to run. `[infra, fix]`- **A paid-call share link now previews on Slack, Telegram and LinkedIn**: Every settled x402 call has a permalink whose social preview image is drawn on demand. Slack, Telegram and LinkedIn ask an image URL for its type and size before they fetch it, and this one answered "method not allowed" to that probe, so the card silently never appeared on those platforms. It answers them now. Three more fixes to the same card: a receipt is captioned with the network it actually settled on rather than always reading mainnet; a link shared in the seconds before its receipt finishes saving, or during a cache outage, is retried a minute later instead of being pinned to the generic card for an hour; and a damaged record shows a plain "unknown" rather than a "NaNd ago" timestamp or an invented amount. (`/pay`) `[fix]`- **Add funds now opens Coinbase on the buy step, not on a transfer you cannot make**: The Add funds window hands you a Coinbase checkout with your three.ws wallet already filled in as the destination. That checkout was not told which step to open on, so Coinbase could land you on "transfer from your Coinbase balance", which is no help at all if you reached Add funds precisely because you hold nothing yet. It now always opens on the buy step for the asset you picked (USDC or SOL, both settling on Solana). The suggested dollar amount is also rounded to real dollars and cents before it reaches Coinbase, so an amount worked out from a shortfall can no longer show up as something like $25.777777 on the checkout screen or in the window itself. `[fix, improvement]`+- **AgenC task lists no longer hang when a Solana provider goes quiet**: Reading the AgenC task board scans the whole coordination program on Solana, the heaviest request we make of any provider. We already move to the next provider when one refuses us, but we had no answer for a provider that accepts the request and then never replies. Nothing was thrown, so nothing switched over, and the read sat there until the browser gave up. Each attempt now gets its own ten second budget, and a provider that runs past it is treated exactly like one that refused: parked for a cooldown, with the read retried on the next provider in the chain. If every provider stalls you get a clear try again shortly response instead of an endless wait. (`/docs/agenc`) `[fix, infra]`- **AgenC task lists no longer hang when a Solana provider goes quiet**: Browsing tasks on the AgenC coordination protocol reads directly from Solana, and the platform keeps a chain of providers so a bad one gets skipped. That skipping only worked when a provider answered with an error. A provider that accepted the request and then went silent held the page open for minutes with nothing to react to, which is what happened to the task list, the heaviest read of the set. Each provider now gets a fixed time to answer before the next one is tried, so a silent provider costs seconds instead of stalling the page. (`/agenc`) `[fix]`- **Agent share cards keep their portrait, and a sealed-drop card always paints a real theme**: The image an agent link unfurls with inlines the agent's avatar thumbnail. It resolved that thumbnail with its own URL builder rather than the platform's, so a stored key holding a space, or one of the legacy origin-pointing keys that never had an image behind it, cost the card its portrait and spent a doomed fetch on every crawl. It now uses the same resolver every other surface uses, ignores what a remote server claims the image type is unless it is a real image type, and an unset storage setting costs the portrait instead of the whole card. Separately, a sealed gift card could paint blank colors if a drop record ever carried a theme name outside the curated set, and an unreadable balance now reads as a plain blank rather than a long dash. `[fix]`- **Agent-runtime and agent-guards docs now match what the code actually returns**: Two developer packages shipped quick-start examples that described results the code does not produce: the runtime's guard example annotated the wrong decision and named a field on the ledger checker that does not exist, and the guards package documented only its hosted half while the local, no-network guard functions went unmentioned. Both are corrected against real executed output, with the local guards now documented end to end including a runnable example and the exact reason codes they return. `[docs, sdk]`@@ -62,6 +64,7 @@ Public history for [three.ws](https://three.ws), newest first. New pages come fr- **The contracts that hold agent earnings, reputation stakes and skill licenses are now audit-ready**: Before an outside firm can review the contracts behind agent royalties, reputation staking and wallet delegation, someone has to write down what those contracts promise and then prove it holds. That pack now exists and is public. Every value-bearing contract has numbered invariants, each backed by a test proving the property holds and a second test proving that an attacker trying to break it is rejected, alongside a threat model naming what is worth stealing, who could try, and what stops them. The Solidity contracts sit at 100% test coverage across 242 tests. The two Solana programs, which had no automated tests at all, now have 32 running against their real compiled on-chain bytecode rather than a stand-in, with the real token programs loaded. Two standard analyzers were run across everything, and every finding is either fixed or answered in writing. One of those fixes closes a real gap: a cross-chain permission request that was declined rather than failed outright would have credited a seller and charged a buyer for access that was never going to be granted. `[security, infra, docs]`- **The crypto knowledge base the agents read from only holds stories we could actually read**: The news reader recorded every URL it was handed into the durable corpus the 3D agents ground their crypto answers on, including ones no extraction attempt could read. That left headline-only rows with no article behind them, and because the reader is a public endpoint whose caller supplies the link, the headline, and the source name, a single request could plant a permanent entry with any headline someone chose. The reader now refuses anything that is not an ordinary web link, records only stories that yielded real article text, and the knowledge API filters the empty rows already on file out of every answer. (`/markets/news`) `[fix, security]`- **The crypto news archive keeps its place when its storage credentials briefly drop**: Every hour a scheduled job copies the latest crypto news into the permanent archive behind the markets pages, so that a story stays readable at its own link long after it has rotated out of the live feed. Before writing anything, the job asks Google Cloud for the credentials it needs. That request can come back empty for a moment, usually because the credential service itself is briefly unreachable rather than because anything is actually misconfigured, and the job treated that as a crash: it returned an unexplained internal error and raised an alert, which read as though the archive were broken. It now recognises a missing credential for what it is, records the hour as a skipped run with the reason stated plainly, and picks up again on the next pass. Nothing is lost when that happens, because each article is filed under a fingerprint of its own link, so the next successful run archives exactly the stories the skipped one would have and never files a duplicate. A genuine storage failure is still reported loudly. `[fix, infra]`+- **The example responses in four API guides now match what the API actually returns**: The docs for ghost-copy, Trader Passport, the wallet portfolio, and the airdrop checker each show a sample JSON response so you can see the shape before you write any code. Those samples had fallen behind the endpoints: they were missing whole fields that the live API returns, including the timestamp on an airdrop scan, the window start and per-trade detail on a ghost-copy replay, and the issuer note, verification instructions, and generation time on a passport. Anyone building against the samples would have concluded those fields did not exist. Every sample was re-checked field by field against a live call to the running endpoint and corrected, and the places where a sample is deliberately shortened are now marked as shortened instead of looking complete. (`/docs/trader-passport`) `[docs]`- **The free fact-check lane is callable from a browser again, and two paid endpoints answer to the price you set**: A site calling the fact checker from a browser got a confusing failure: the preflight check said the call was allowed, then the browser threw away the answer. The free daily lane was not sending the cross-origin headers the paid lane sends, so a perfectly good verdict never reached the page. Both lanes now answer identically, whether the check succeeds or the claim is rejected. Separately, the fact checker and the paid LLM proxy were the last two paid endpoints whose price was fixed in code: every other one can be repriced at deploy time, and now these two can too, with no code change. The IBM Granite health feed also stops reporting itself healthy when its own database read fails, so a dashboard cannot show green through an outage. Exploit-history and gas-oracle outages are now written to the server log rather than disappearing silently, which is the difference between an alert and a mystery. (`/docs/x402-endpoints`) `[fix, improvement]`- **The free object library can no longer be emptied by one bad upload**: The CC0 prop library at /objects reads a single index file from the CDN. If that file were ever uploaded truncated or in the wrong shape, the gallery treated it as a library with nothing in it and told the CDN to remember that answer for the next five minutes, so a bad upload kept the shelves bare long after it was replaced. A damaged index is now handled the same way as the CDN being briefly unreachable: the page stays quiet, nothing gets cached, and the props come back on the very next request. Two smaller fixes went with it: a request with a broken page cursor (a limit of "abc", a negative offset) is rejected outright instead of first fetching an index it was never going to use, and the developer documentation no longer claims the library API can be called from any website, which it cannot; it answers three.ws and its partner sites, and everything it points at is still yours to download and reuse anywhere. (`/objects`) `[fix, docs]`- **The internal money and platform health boards tell the truth again**: Three internal dashboards watch the platform for us: one for site and cron health, one for every money subsystem, one for payment outcomes. Two of them had gone quietly wrong. The platform board judged each scheduled job against a hand-written idea of how often it should run, and three jobs that run once a day were being judged against a window of minutes, so they reported as broken for roughly 23 hours out of every day and the board could never come back all-clear; every job is now measured against the schedule it genuinely runs on. One of its probes was also posting to a payments endpoint without declaring it was sending JSON, so it had been failing on its own mistake rather than on anything real. The money board was reading the wrong timestamp column on two payment logs and showing no activity at all for a payment ring that settles thousands of payments a day. It also accepted the credential our fund-moving scheduled jobs carry as an ops password; it now takes the same admin sign-in as the other boards, so an ops credential can never be escalated into triggering a payment job. `[fix, infra, security]`@@ -86,6 +89,7 @@ Public history for [three.ws](https://three.ws), newest first. New pages come fr- **The vanity grinders stop keeping any copy of a key you bought, and receipt checks now work from a plain link**: Two changes to the pay-per-call vanity endpoints and the free receipt verifier. First, security: when a paid call is retried, the server keeps a short-lived copy of the original response so you are never charged twice. On the provably-fair grinder that copy included the receipt's serverSeed, and the seed is enough to recompute the private key, so a retry cache that was only ever meant to hold public metadata was in fact holding key material. The stored copy is now stripped of the secret and the seed on both grinders, sealed deliveries included, and a retry returns the public metadata with a plain note that the key was delivered once. The key you receive in the live response is unchanged. Second, the free receipt verifier can now confirm a settlement transaction over a plain GET link: /api/x402/verify-receipt?hash=<tx>&network=solana, which works from a browser, a curl one-liner, or an uptime probe with no request body. Checking an attestation still uses POST because it needs the whole signed object. (`/docs/x402-dev-tools`) `[security, improvement]`- **Three paid three.ws services were invisible in the public x402 directory**: An hourly job keeps the platform's paid, agent-callable services listed on 402index, the public directory other agents search when they need a service to pay for. Before submitting a service the job confirmed the endpoint was really deployed, and it used a strict test: send a bare request and require a payment-required answer back. That test quietly disqualified any service whose paywall sits behind a parameter rather than on the bare address. Premium vanity addresses are exactly that shape: asking with no parameters is a free browse of what is in stock, and only naming a specific address asks for payment. The job cannot construct a valid paid request for an arbitrary service, so there was no hour on which such a service could ever have passed the test, and it was skipped in silence on every tick. Three live paid services were affected and had never been listed: premium vanity addresses, the pipeline endpoint, and pay-by-name. The check now asks the question it always meant to ask, which is whether the route exists at all. A missing route or a failing origin still keeps a service out of the directory, so the limited number of submissions allowed per hour is still protected, but a service that is genuinely deployed now gets listed whatever shape its paywall takes. `[fix]`- **Three paid x402 endpoints stop shortchanging the agents that call them**: The JSON schema checker at /api/x402/schema-check never read the request body, so every paid call charged the caller and answered that the requested feed was unsupported. It now reads the body it was always meant to, validates the changelog feed for real, and refuses an unknown target with a 400 before any payment settles. The per-call skill gate at /api/x402/skill-call answered browser preflight requests with a payment challenge and no cross-origin headers, which made it uncallable from any in-browser wallet; it now answers preflight correctly on every path, and a database outage returns a plain retry-shortly message instead of echoing internal configuration detail. The paid capacity probe at /api/x402/rate-limit-probe now only accepts metered routes as its target, so a paying caller cannot aim the server's own price lookup at an unrelated internal route. The remix endpoint's refusals (unknown source, empty instruction) are no longer recorded as payment leaks in the x402 ledger. (`/docs/x402-endpoints`) `[fix, security]`+- **Three pieces of docs text that were too faint to read comfortably are now legible**: An accessibility sweep of the documentation found three places where text sat below the readable-contrast standard against the surface behind it: the comment lines inside every code sample, which is where most of our samples put the explanation, the See what changed link on the page-freshness banner, and the Live label on runnable code blocks. All three came from the same habit of dimming text that was already dark, and all three are now above the standard in both the dark and light themes. The status page KPI labels got the same treatment. Nothing moved and no wording changed, the affected text is simply readable now. (`/docs/custody`) `[improvement, docs]`- **Token-to-3D calls return the real coin again, and receipts show the amount you actually paid**: Paying to turn a Solana token into a 3D mesh handed back an untitled, untextured cube for most of today's coins, $THREE included. The server only knew how to read a token's name, symbol and artwork from the older of the two places Solana stores it, and every launch since the Token-2022 change keeps that information inside the coin's own account instead. Those coins now come back with their real name, symbol and logo baked onto the mesh. The same read now refuses an address that is not a coin at all, rather than charging for a blank cube, and the batch call's published example uses a coin that actually works if you copy it. Separately, the receipts page could not scale amounts paid on X Layer or in $THREE, so a one-tenth-of-a-cent payment could read as a raw 1000; both now display correctly. (`/docs/x402-endpoints`) `[fix]`- **Tournament prize payouts now only pay pools the platform actually stands behind**: Anyone signed in can create a tournament and name a prize pool, and nothing ever asked them to deposit it, while the payout itself came from the platform's own prize wallet. That combination meant a made-up prize could be claimed as real money. Prize money is now released only for the platform's own Daily Arena, or for a hosted pool a deployment has explicitly agreed to sponsor up to a set ceiling. Everything else about a hosted competition is unchanged: it still runs, still ranks live, and still gets its result published on-chain. An unfunded prize is refused in the open, with the reason shown on the page, rather than quietly paid or quietly ignored. (`/arena`) `[security, fix]`- **Two chains that had quietly stopped feeding the cross-chain agent directory are indexing again**: The agent directory lists agents registered on twenty-two EVM networks, kept current by a job that walks each network's registry in windows of blocks every fifteen minutes. That window grows while a network is behind and shrinks when a node refuses the request, and the two rules deadlocked each other: a refusal recorded the smaller window but left the backlog wide, so the very next run doubled straight back to the size that had just been refused. Two networks sat in that loop indefinitely, reading zero blocks per run while their backlogs grew, and nothing looked wrong because every run still reported success. A refused window is now retried at half the width immediately, inside the same run, so a run that overshoots still delivers blocks instead of nothing. A second fault sat underneath it on one test network, whose configured nodes decline every log query outright, at any width, so no amount of shrinking could ever have helped; that refusal was being misread as a complaint about size. The job now tells the two apart, and that network has been pointed at a node that serves the query. Confirmed over consecutive runs: one network caught fully up, the other began clearing its backlog. `[fix, infra]`@@ -93,6 +97,7 @@ Public history for [three.ws](https://three.ws), newest first. New pages come fr- **Withdrawing a reputation stake now stops counting as a vote of confidence**: You can back an agent by staking SOL behind it, and the agent's reputation card shows how much conviction is escrowed and who put it there. Until now that figure only ever went up: withdrawing your stake returned your money but left your endorsement standing forever, so an agent could show several SOL of backing that nobody was actually behind any more. The card now counts what is still staked. The moment a withdrawal settles, that position drops out of the total and its owner drops out of the list of backers. Nothing is hidden by the change: the card also reports everything ever staked and everything settlement gave back, so a long track record of people staking and withdrawing still reads as one. Only the market's own escrow can retire a position, so nobody can make an agent look abandoned by publishing a note about somebody else's stake, and the total can never be pushed below zero. The staking market page also now says why staking is closed when a network has no escrow configured, instead of showing a greyed-out button with no explanation. (`/reputation/market`) `[fix, feature]`- **Withdrawing your agent token's earnings is confirmed against the real payout, and cannot be triggered from another site**: Two fixes to the withdraw flow on your agent's pump.fun token. The confirmation step used to accept any successful Solana transaction that simply mentioned your token and its authority wallet, which an ordinary transfer or memo does, and report it back as a completed payout; it now requires that the payments program actually ran, the same check the launch flow already made. Both withdraw steps also now reject requests that arrive from a site other than three.ws while riding your signed-in session, so a page you happen to have open elsewhere cannot start a withdrawal on your behalf. Alongside that, an unreadable token-program value in a withdraw request is now refused up front with a clear message instead of being quietly ignored and producing a transaction that only failed after you had signed it. `[security, fix]`- **Your agent now leads with the GitHub work you seeded it from**: Seeding an agent from your GitHub profile and repositories stored the facts correctly, but filed all of them in the tier the agent only consults when a message happens to match one. So the seed worked and yet asking your agent an open question like 'what do you know about my work?' could come back with nothing, because nothing in the question named a repository. The five top-ranked facts from a seed now sit in the context the agent always carries, so they ground an answer without being prompted for by name; the rest stay searchable exactly as before and still surface when you ask about that specific project. Facts also keep the order they were distilled in, which matters because each reply carries only the ten strongest memories, and until now that cut was made arbitrarily among facts that all scored the same. Nothing about what is read or stored changes: the seed still reads only the profile, repositories, and READMEs you ticked, and disconnecting GitHub still deletes every fact it wrote across all your agents. (`/docs/github-memory-seeding`) `[fix, improvement]`+- **Your agent's memory graph was always empty. It is not any more**: Every memory your agent stores is scanned for the things it mentions: tokens, tickers, wallets, people, and the trading rules you told it to follow. Those become the nodes of a knowledge graph you can browse, so you can ask what your agent remembers about one specific token and read back every memory that touched it. The scan ran correctly and then failed at the last step, when it tried to save what it found, because of a type mismatch on one database column. The failure was recorded quietly rather than shown, so nothing ever looked broken: memories saved fine, search worked fine, and the graph simply came back empty for every agent, forever. The save now works, memories that were skipped get scanned the next time you open the graph, and a test covers the exact shape that broke. (`/docs/memory`) `[fix]`- **Your agent's memory graph was never being built, and now it is**: Memory Studio promises a knowledge graph: every fact your agent stores is scanned for the mints, tickers, wallets, people, and strategies it mentions, and those become nodes you can click through to ask what your agent remembers about any one of them. That scan was failing on a type mismatch the moment it tried to save a node, and because the failure was caught quietly, nothing ever looked broken: the graph simply came back empty forever, for every agent, and the memories were never marked as scanned. The save is fixed, so the graph now fills in on your next visit, including for everything your agents have stored up to now. Recall and search were never affected. (`/memory`) `[fix]`- **Your referrals page now shows how many people actually opened your link**: Every three.ws referral link already recorded a visit when someone opened it, but nothing ever showed you those numbers. The referrals dashboard now has a Share funnel: link visits, how many of those visits became signups, and how many of those signups went on to make their first creation, with the conversion rate between each step over the last 30 days. Visits are counted once per person per day, so refreshing a link does not inflate the number, and the page only shows the funnel once you have at least one visit. The same data is available on the referrals API as a funnel block, with an optional lookback window. (`/dashboard/referrals`) `[feature, improvement]`@@ -1152,7 +1157,7 @@ Public history for [three.ws](https://three.ws), newest first. New pages come fr- **AR Forge** (`/ar`): Type a prompt, get a real 3D model, and place it in your room: free text-to-3D with one-tap AR on any phone (WebXR, Scene Viewer, Quick Look) and a QR handoff from desktop.- **Docs · 3D Studio MCP (free)** (`/docs/mcp-studio`): The free, no-auth MCP server behind the ChatGPT 3D Studio app: text and image to 3D, rigged avatars, conversational refinement, and living agent personas.- **Docs · Sketchfab Showcase** (`/docs/sketchfab`): How the best community forge models get published to the official three.ws Sketchfab account: what gets picked, AI disclosure, and the backlinks that make referrals measurable.-- **End-User License Agreement** (`/legal/eula`): License terms for the three.ws Seeker/Saga app.+- **End-User License Agreement** (`/legal/eula`): License terms for the three.ws Seeker/Saga mobile app: what the license covers, camera and selfie data, updates, and termination.- **"Deploy onchain" now opens the real deploy flow with your model loaded**: After forging a model, the "Deploy onchain" shortcut used to drop you on Launchpad Studio (the white-label site builder), which ignored your model entirely. It now opens the on-chain deploy wizard with your model and its name already filled in, so you go straight from forge to a permanent on-chain agent identity. The search palette's "Launch a coin" action was also mispointed and now opens the pump.fun coin launch flow. (`/deploy`) `[fix]`- **3D models now render inside ChatGPT**: Fixed the free 3D Studio app for ChatGPT so generated models actually display in the conversation. The inline viewer's security allowlist was missing the storage domain models are served from, so ChatGPT blocked every model from loading; it is now included automatically. Living agent personas (create, reload, speak) also gained their own inline view in ChatGPT, so a saved agent body idles, lip-syncs, and emotes right in the chat. `[fix, improvement]`- **3D previews stop erroring on the scanner and every plain page footer**: The walking robot in the site footer failed to appear on plain pages like /hydrate and /login because its compressed 3D model needed a decoder the page never loaded; the decoder is now wired in before the viewer starts, so the robot renders everywhere. The 3D scanner also downloaded a 748KB placeholder avatar the moment the page opened, even for visitors who never started a scan, and logged a fetch error on every visit to the old /scan address; the placeholder now loads only when the build step is actually on screen, and if its download ever fails the progress view stays clean instead of showing a dead black panel. (`/create/selfie`) `[fix]`@@ -2191,6 +2196,7 @@ Public history for [three.ws](https://three.ws), newest first. New pages come fr## 2026-06-21- **Arm your agent: automate Oracle conviction** (`/oracle/arm`): Configure your 3D agent to trade Oracle conviction automatically: set the minimum score, position size, daily and open-position caps, narrative filters and Telegram alerts, then arm it in simulate or live mode. Every action is graded against the outcome.+- **Content Policy** (`/legal/content-policy`): What you may publish, mint, and monetise on three.ws: prohibited content, agent behaviour, marketplace listings, and how we enforce.- **Credits** (`/credits`): Top up with SOL or $THREE and spend credits on 3D generation, agents, and compute. Hold $THREE for up to 30% off every spend.- **Docs · Walk Avatar** (`/docs/walk`): Developer and user documentation for the three.ws walking-avatar system: the /walk page, iframe embeds, the JavaScript SDK, the postMessage event contract, companion mode, analytics, and the Chrome extension.- **Feed** (`/feed`): Live avatars, agents, 3D models, worlds, and launches across three.ws: filter to everyone or just the people you follow.@@ -3193,7 +3199,7 @@ Public history for [three.ws](https://three.ws), newest first. New pages come fr- **Docs · Skills** (`/docs/skills`): Write, package, and monetize agent skills.- **OpenAPI** (`/openapi.json`): Machine-readable HTTP API spec.- **Privacy Policy** (`/legal/privacy`): What data three.ws collects, why we hold it, who we share it with, how long we keep it, and the controls you have over it.-- **Terms of Use** (`/legal/tos`): Terms of service for three.ws.+- **Terms of Service** (`/legal/tos`): The agreement between you and three.ws: your account, acceptable use, payments and on-chain activity, liability, and DMCA takedowns.- **Tutorial · Build a custom skill** (`/tutorials/custom-skill`): A complete walkthrough of building a weather skill an agent can invoke: bundle format, LLM tool calling, context API.- **Tutorial · Build a personal AI website** (`/tutorials/personal-ai-site`): A 3D version of you, on your domain, ready to answer any question about your work: with a portfolio, contact form, and resume.- **Tutorial · Build your first agent (from scratch)** (`/tutorials/first-agent`): A complete walkthrough from a blank HTML file to a live, talking 3D character you wrote yourself: no framework, no build tools.diff --git a/data/changelog.json b/data/changelog.jsonindex c31b3963f..03a627029 100644--- a/data/changelog.json+++ b/data/changelog.json@@ -1,6 +1,36 @@{"$comment": "Curated changelog entries for three.ws, the editorial layer of the public changelog. New PAGE launches are tracked automatically via the `added` field in data/pages.json; this file is for everything else holders care about: improvements, fixes, SDK releases, trading/payments capabilities, security work. scripts/build-page-index.mjs merges both sources into CHANGELOG.md, public/changelog.json, and public/changelog.xml. Newest first. Allowed tags: feature, improvement, fix, sdk, infra, docs, security. `link` is optional and must be a live three.ws path.","entries": [+ {+ "date": "2026-08-14",+ "title": "The choreography docs no longer teach an example that quietly does nothing",+ "summary": "A routine step can pin one specific animation clip instead of letting the gesture slot pick one, and the choreography reference showed how, with a shareable link that pinned the dance slot to the rumba clip. That example never pinned anything. Rumba is already the platform default for the dance slot, and the runtime deliberately drops a clip that matches the slot's own default, so that a shared routine still adapts to an agent that has remapped its gestures. Anyone copying the example got a step that resolved normally instead of the pinned motion they asked for, with nothing to indicate why. The example now pins a clip that is not the default, and the field reference states the drop rule outright. The selfie-to-avatar page also had a summary line quoting an older, lower head-coverage number than the measurement reported further down the same page; both now read the measured figure.",+ "tags": [+ "docs",+ "fix"+ ],+ "link": "/docs/choreography"+ },+ {+ "date": "2026-08-14",+ "title": "The marketplace, the directory and the home page open with a lot less work behind them",+ "summary": "A Lighthouse sweep of eight pages against the live site found four costs that every visitor was paying for content they could not see. The marketplace was laying out its entire catalogue on first paint, a grid measured at over 31,000 pixels tall of which one screenful is visible, and the directory did the same at smaller scale; both now skip the offscreen rows until you scroll near them. The analytics library was loading in the page head on every route, which cost around a second of browser time before anything was interactive, and now waits for the first idle moment instead, with nothing dropped in between. The embeddable avatar bundle on our CDN was shipping unminified because of a build-tool rule meant for packages that get re-bundled, which does not apply to a file browsers download directly; it is now 2.77 MB instead of 4.24 MB, and 792 KB instead of 1.03 MB over the wire. And the thumbnails in both grids now decode off the critical path. The rules behind all four, and how to measure before changing any of them, are written up so the next change does not quietly undo one.",+ "tags": [+ "improvement",+ "infra"+ ],+ "link": "/marketplace"+ },+ {+ "date": "2026-08-14",+ "title": "Three pieces of docs text that were too faint to read comfortably are now legible",+ "summary": "An accessibility sweep of the documentation found three places where text sat below the readable-contrast standard against the surface behind it: the comment lines inside every code sample, which is where most of our samples put the explanation, the See what changed link on the page-freshness banner, and the Live label on runnable code blocks. All three came from the same habit of dimming text that was already dark, and all three are now above the standard in both the dark and light themes. The status page KPI labels got the same treatment. Nothing moved and no wording changed, the affected text is simply readable now.",+ "tags": [+ "improvement",+ "docs"+ ],+ "link": "/docs/custody"+ },{"date": "2026-08-14","title": "The example responses in four API guides now match what the API actually returns",@@ -10,6 +40,34 @@],"link": "/docs/trader-passport"},+ {+ "date": "2026-08-14",+ "title": "Your agent's memory graph was always empty. It is not any more",+ "summary": "Every memory your agent stores is scanned for the things it mentions: tokens, tickers, wallets, people, and the trading rules you told it to follow. Those become the nodes of a knowledge graph you can browse, so you can ask what your agent remembers about one specific token and read back every memory that touched it. The scan ran correctly and then failed at the last step, when it tried to save what it found, because of a type mismatch on one database column. The failure was recorded quietly rather than shown, so nothing ever looked broken: memories saved fine, search worked fine, and the graph simply came back empty for every agent, forever. The save now works, memories that were skipped get scanned the next time you open the graph, and a test covers the exact shape that broke.",+ "tags": [+ "fix"+ ],+ "link": "/docs/memory"+ },+ {+ "date": "2026-08-14",+ "title": "AgenC task lists no longer hang when a Solana provider goes quiet",+ "summary": "Reading the AgenC task board scans the whole coordination program on Solana, the heaviest request we make of any provider. We already move to the next provider when one refuses us, but we had no answer for a provider that accepts the request and then never replies. Nothing was thrown, so nothing switched over, and the read sat there until the browser gave up. Each attempt now gets its own ten second budget, and a provider that runs past it is treated exactly like one that refused: parked for a cooldown, with the read retried on the next provider in the chain. If every provider stalls you get a clear try again shortly response instead of an endless wait.",+ "tags": [+ "fix",+ "infra"+ ],+ "link": "/docs/agenc"+ },+ {+ "date": "2026-08-14",+ "title": "A failed database migration can no longer freeze the schema in silence",+ "summary": "Every deploy carries database migrations that must run before the new code does. When one of them failed, the runner printed a single line with the database's own error and stopped, naming neither the file that failed nor the migrations queued behind it. Those stayed unapplied and unmentioned, so the site ended up serving code that reads columns the database never got, and the first visible symptom was an unrelated-looking error on a live page. The runner now names the failing file, how many ran before it, and every migration still waiting, and the deploy step that catches a database running behind the code is now part of the documented deploy path instead of something you had to remember to run.",+ "tags": [+ "infra",+ "fix"+ ]+ },{"date": "2026-08-14","title": "Your agent's memory graph was never being built, and now it is",diff --git a/docs/nvidia-apps-catalog-listing.md b/docs/nvidia-apps-catalog-listing.mdindex 86d9a8f97..c5da2c996 100644--- a/docs/nvidia-apps-catalog-listing.md+++ b/docs/nvidia-apps-catalog-listing.md@@ -109,6 +109,43 @@ All in the repo, web-ready:- Full press kit: `public/brand/three-ws-press-kit.zip`- Screenshots: capture live from https://three.ws/forge (generation in progress plus result), https://three.ws/agora (3D world), and https://three.ws/character-library (rigged character grid) at 1920x1080+## Second product record: the `<agent-3d>` digital-human embed++A genuinely distinct shipping surface with its own NVIDIA story, filed as a separate portal product AFTER the main record is corrected. It targets the Digital Humans / conversational AI workload filter, which the main (Generative AI content creation) entry does not.++**Product name:** three.ws Agent Embed (`<agent-3d>`)++**Stage:** Shipping++**Acceleration:** GPU Accelerated++**Product URL:** https://three.ws/docs/embedding++### Short description (about 130 characters)++> One script tag puts a talking, lip-synced 3D AI agent on any website. Voice by NVIDIA Riva, facial animation by Audio2Face-3D.++### Medium description (about 500 characters)++> The `<agent-3d>` web component embeds a live, conversational 3D avatar on any site with one script tag: no plugins, no framework requirements. The avatar listens (NVIDIA Riva ASR), reasons (Nemotron via NIM APIs), speaks (Riva Magpie multilingual TTS), and moves its face in real time from NVIDIA Audio2Face-3D ARKit blendshape tracks, mapped at runtime onto whatever morph convention the visitor's avatar ships (ARKit-52, VRM vowels, Oculus visemes). Every reply is gated by NemoGuard content safety.++### Technologies Used (this record)++- Riva (ASR + Magpie TTS over NVCF gRPC)+- Audio2Face (A2F-3D blendshape streams driving browser playback)+- NIM APIs (Nemotron brains, NemoGuard safety)+- CUDA Toolkit / cuDNN (the avatar bodies it renders come from the L4 fleet)++Nothing else checked. TensorRT/Triton/Omniverse stay unchecked here too.++### Links (this record)++- Docs: https://three.ws/docs/embedding+- Live A2F demo: https://three.ws/demos/audio2face+- Voice stack source: https://github.com/nirholas/three.ws/tree/main/src/voice++---+## Submission checklist1. Log in to the Inception portal, go to Profile, then Add product.@@ -117,3 +154,6 @@ All in the repo, web-ready:4. Check every NVIDIA technology the form offers that appears in the list above; check nothing that does not.5. Upload the logo mark and OG image; add screenshots if the form accepts them.6. Submit. Listings are reviewed on a rolling basis; note the submission date in this doc when done.+7. Correct the existing record per the Status section (move Riva/Audio2Face/NIM to Used, drop DeepVariant NIM) BEFORE adding anything new.+8. Add the `<agent-3d>` embed as a second product using its section above. Do not file more records than these two: duplicates read as noise to the curator.+9. Send the inclusion-request email in [nvidia-apps-catalog-request.md](./nvidia-apps-catalog-request.md) to [email protected] and record the send date there.diff --git a/public/changelog.json b/public/changelog.jsonindex 4d4fc3cf3..d4e0c99b3 100644--- a/public/changelog.json+++ b/public/changelog.json@@ -1,5 +1,5 @@{- "generated_at": "2026-08-14T18:22:30.219Z",+ "generated_at": "2026-08-14T19:37:07.682Z","generated_by": "scripts/build-page-index.mjs from data/pages.json + data/changelog.json","site": {"name": "three.ws",@@ -99,6 +99,17 @@"infra"]},+ {+ "date": "2026-08-14",+ "type": "update",+ "title": "A failed database migration can no longer freeze the schema in silence",+ "summary": "Every deploy carries database migrations that must run before the new code does. When one of them failed, the runner printed a single line with the database's own error and stopped, naming neither the file that failed nor the migrations queued behind it. Those stayed unapplied and unmentioned, so the site ended up serving code that reads columns the database never got, and the first visible symptom was an unrelated-looking error on a live page. The runner now names the failing file, how many ran before it, and every migration still waiting, and the deploy step that catches a database running behind the code is now part of the documented deploy path instead of something you had to remember to run.",+ "link": null,+ "tags": [+ "infra",+ "fix"+ ]+ },{"date": "2026-08-14","type": "update",@@ -120,6 +131,17 @@"improvement"]},+ {+ "date": "2026-08-14",+ "type": "update",+ "title": "AgenC task lists no longer hang when a Solana provider goes quiet",+ "summary": "Reading the AgenC task board scans the whole coordination program on Solana, the heaviest request we make of any provider. We already move to the next provider when one refuses us, but we had no answer for a provider that accepts the request and then never replies. Nothing was thrown, so nothing switched over, and the read sat there until the browser gave up. Each attempt now gets its own ten second budget, and a provider that runs past it is treated exactly like one that refused: parked for a cooldown, with the read retried on the next provider in the chain. If every provider stalls you get a clear try again shortly response instead of an endless wait.",+ "link": "/docs/agenc",+ "tags": [+ "fix",+ "infra"+ ]+ },{"date": "2026-08-14","type": "update",@@ -608,6 +630,16 @@"infra"]},+ {+ "date": "2026-08-14",+ "type": "update",+ "title": "The example responses in four API guides now match what the API actually returns",+ "summary": "The docs for ghost-copy, Trader Passport, the wallet portfolio, and the airdrop checker each show a sample JSON response so you can see the shape before you write any code. Those samples had fallen behind the endpoints: they were missing whole fields that the live API returns, including the timestamp on an airdrop scan, the window start and per-trade detail on a ghost-copy replay, and the issuer note, verification instructions, and generation time on a passport. Anyone building against the samples would have concluded those fields did not exist. Every sample was re-checked field by field against a live call to the running endpoint and corrected, and the places where a sample is deliberately shortened are now marked as shortened instead of looking complete.",+ "link": "/docs/trader-passport",+ "tags": [+ "docs"+ ]+ },{"date": "2026-08-14","type": "update",@@ -873,6 +905,17 @@"security"]},+ {+ "date": "2026-08-14",+ "type": "update",+ "title": "Three pieces of docs text that were too faint to read comfortably are now legible",+ "summary": "An accessibility sweep of the documentation found three places where text sat below the readable-contrast standard against the surface behind it: the comment lines inside every code sample, which is where most of our samples put the explanation, the See what changed link on the page-freshness banner, and the Live label on runnable code blocks. All three came from the same habit of dimming text that was already dark, and all three are now above the standard in both the dark and light themes. The status page KPI labels got the same treatment. Nothing moved and no wording changed, the affected text is simply readable now.",+ "link": "/docs/custody",+ "tags": [+ "improvement",+ "docs"+ ]+ },{"date": "2026-08-14","type": "update",@@ -949,6 +992,16 @@"improvement"]},+ {+ "date": "2026-08-14",+ "type": "update",+ "title": "Your agent's memory graph was always empty. It is not any more",+ "summary": "Every memory your agent stores is scanned for the things it mentions: tokens, tickers, wallets, people, and the trading rules you told it to follow. Those become the nodes of a knowledge graph you can browse, so you can ask what your agent remembers about one specific token and read back every memory that touched it. The scan ran correctly and then failed at the last step, when it tried to save what it found, because of a type mismatch on one database column. The failure was recorded quietly rather than shown, so nothing ever looked broken: memories saved fine, search worked fine, and the graph simply came back empty for every agent, forever. The save now works, memories that were skipped get scanned the next time you open the graph, and a test covers the exact shape that broke.",+ "link": "/docs/memory",+ "tags": [+ "fix"+ ]+ },{"date": "2026-08-14","type": "update",@@ -11149,7 +11202,7 @@"date": "2026-07-14","type": "launch","title": "End-User License Agreement",- "summary": "License terms for the three.ws Seeker/Saga app.",+ "summary": "License terms for the three.ws Seeker/Saga mobile app: what the license covers, camera and selfie data, updates, and termination.","link": "/legal/eula","tags": ["launch"@@ -21271,6 +21324,16 @@"launch"]},+ {+ "date": "2026-06-21",+ "type": "launch",+ "title": "Content Policy",+ "summary": "What you may publish, mint, and monetise on three.ws: prohibited content, agent behaviour, marketplace listings, and how we enforce.",+ "link": "/legal/content-policy",+ "tags": [+ "launch"+ ]+ },{"date": "2026-06-21","type": "launch",@@ -30165,8 +30228,8 @@{"date": "2026-04-27","type": "launch",- "title": "Terms of Use",- "summary": "Terms of service for three.ws.",+ "title": "Terms of Service",+ "summary": "The agreement between you and three.ws: your account, acceptable use, payments and on-chain activity, liability, and DMCA takedowns.","link": "/legal/tos","tags": ["launch"diff --git a/public/changelog.xml b/public/changelog.xmlindex 0f4420396..ce49ec651 100644--- a/public/changelog.xml+++ b/public/changelog.xml@@ -78,6 +78,14 @@<category>fix</category><description>Two of the scheduled jobs behind the trading fleet could fail without anything looking wrong. The job that reallocates budget across strategies talks to the database over a socket, and when that socket never opened the failure arrived as an object carrying no message at all. The job caught it, answered with an empty error string, and the scheduler recorded a success, so the health record stayed green while the loop did nothing. Failures of that kind are now recognised as what they are, a database that could not be reached, so the run is marked unhealthy and retried on the next tick instead of being logged as fine. Every page and endpoint that reads from the database inherits the same correction: this class of outage now returns a retryable service-unavailable response rather than an internal error. The fleet watchdog also gained a seventh check, over the calibration pass that measures whether a high-conviction coin actually wins as often as its score claims, so that loop can no longer go quiet unnoticed.</description></item>+ <item>+ <title>A failed database migration can no longer freeze the schema in silence</title>+ <link>https://three.ws/changelog</link>+ <guid isPermaLink="false">2026-08-14:A failed database migration can no longer freeze the schema in silence</guid>+ <pubDate>Fri, 14 Aug 2026 12:00:00 GMT</pubDate>+ <category>infra</category>+ <description>Every deploy carries database migrations that must run before the new code does. When one of them failed, the runner printed a single line with the database's own error and stopped, naming neither the file that failed nor the migrations queued behind it. Those stayed unapplied and unmentioned, so the site ended up serving code that reads columns the database never got, and the first visible symptom was an unrelated-looking error on a live page. The runner now names the failing file, how many ran before it, and every migration still waiting, and the deploy step that catches a database running behind the code is now part of the documented deploy path instead of something you had to remember to run.</description>+ </item><item><title>A paid-call share link now previews on Slack, Telegram and LinkedIn</title><link>https://three.ws/pay</link>@@ -94,6 +102,14 @@<category>fix</category><description>The Add funds window hands you a Coinbase checkout with your three.ws wallet already filled in as the destination. That checkout was not told which step to open on, so Coinbase could land you on "transfer from your Coinbase balance", which is no help at all if you reached Add funds precisely because you hold nothing yet. It now always opens on the buy step for the asset you picked (USDC or SOL, both settling on Solana). The suggested dollar amount is also rounded to real dollars and cents before it reaches Coinbase, so an amount worked out from a shortfall can no longer show up as something like $25.777777 on the checkout screen or in the window itself.</description></item>+ <item>+ <title>AgenC task lists no longer hang when a Solana provider goes quiet</title>+ <link>https://three.ws/docs/agenc</link>+ <guid isPermaLink="false">2026-08-14:AgenC task lists no longer hang when a Solana provider goes quiet</guid>+ <pubDate>Fri, 14 Aug 2026 12:00:00 GMT</pubDate>+ <category>fix</category>+ <description>Reading the AgenC task board scans the whole coordination program on Solana, the heaviest request we make of any provider. We already move to the next provider when one refuses us, but we had no answer for a provider that accepts the request and then never replies. Nothing was thrown, so nothing switched over, and the read sat there until the browser gave up. Each attempt now gets its own ten second budget, and a provider that runs past it is treated exactly like one that refused: parked for a cooldown, with the read retried on the next provider in the chain. If every provider stalls you get a clear try again shortly response instead of an endless wait.</description>+ </item><item><title>AgenC task lists no longer hang when a Solana provider goes quiet</title><link>https://three.ws/agenc</link>@@ -454,6 +470,14 @@<category>fix</category><description>Every hour a scheduled job copies the latest crypto news into the permanent archive behind the markets pages, so that a story stays readable at its own link long after it has rotated out of the live feed. Before writing anything, the job asks Google Cloud for the credentials it needs. That request can come back empty for a moment, usually because the credential service itself is briefly unreachable rather than because anything is actually misconfigured, and the job treated that as a crash: it returned an unexplained internal error and raised an alert, which read as though the archive were broken. It now recognises a missing credential for what it is, records the hour as a skipped run with the reason stated plainly, and picks up again on the next pass. Nothing is lost when that happens, because each article is filed under a fingerprint of its own link, so the next successful run archives exactly the stories the skipped one would have and never files a duplicate. A genuine storage failure is still reported loudly.</description></item>+ <item>+ <title>The example responses in four API guides now match what the API actually returns</title>+ <link>https://three.ws/docs/trader-passport</link>+ <guid isPermaLink="false">2026-08-14:The example responses in four API guides now match what the API actually returns</guid>+ <pubDate>Fri, 14 Aug 2026 12:00:00 GMT</pubDate>+ <category>docs</category>+ <description>The docs for ghost-copy, Trader Passport, the wallet portfolio, and the airdrop checker each show a sample JSON response so you can see the shape before you write any code. Those samples had fallen behind the endpoints: they were missing whole fields that the live API returns, including the timestamp on an airdrop scan, the window start and per-trade detail on a ghost-copy replay, and the issuer note, verification instructions, and generation time on a passport. Anyone building against the samples would have concluded those fields did not exist. Every sample was re-checked field by field against a live call to the running endpoint and corrected, and the places where a sample is deliberately shortened are now marked as shortened instead of looking complete.</description>+ </item><item><title>The free fact-check lane is callable from a browser again, and two paid endpoints answer to the price you set</title><link>https://three.ws/docs/x402-endpoints</link>@@ -646,6 +670,14 @@<category>fix</category><description>The JSON schema checker at /api/x402/schema-check never read the request body, so every paid call charged the caller and answered that the requested feed was unsupported. It now reads the body it was always meant to, validates the changelog feed for real, and refuses an unknown target with a 400 before any payment settles. The per-call skill gate at /api/x402/skill-call answered browser preflight requests with a payment challenge and no cross-origin headers, which made it uncallable from any in-browser wallet; it now answers preflight correctly on every path, and a database outage returns a plain retry-shortly message instead of echoing internal configuration detail. The paid capacity probe at /api/x402/rate-limit-probe now only accepts metered routes as its target, so a paying caller cannot aim the server's own price lookup at an unrelated internal route. The remix endpoint's refusals (unknown source, empty instruction) are no longer recorded as payment leaks in the x402 ledger.</description></item>+ <item>+ <title>Three pieces of docs text that were too faint to read comfortably are now legible</title>+ <link>https://three.ws/docs/custody</link>+ <guid isPermaLink="false">2026-08-14:Three pieces of docs text that were too faint to read comfortably are now legible</guid>+ <pubDate>Fri, 14 Aug 2026 12:00:00 GMT</pubDate>+ <category>improvement</category>+ <description>An accessibility sweep of the documentation found three places where text sat below the readable-contrast standard against the surface behind it: the comment lines inside every code sample, which is where most of our samples put the explanation, the See what changed link on the page-freshness banner, and the Live label on runnable code blocks. All three came from the same habit of dimming text that was already dark, and all three are now above the standard in both the dark and light themes. The status page KPI labels got the same treatment. Nothing moved and no wording changed, the affected text is simply readable now.</description>+ </item><item><title>Token-to-3D calls return the real coin again, and receipts show the amount you actually paid</title><link>https://three.ws/docs/x402-endpoints</link>@@ -702,6 +734,14 @@<category>fix</category><description>Seeding an agent from your GitHub profile and repositories stored the facts correctly, but filed all of them in the tier the agent only consults when a message happens to match one. So the seed worked and yet asking your agent an open question like 'what do you know about my work?' could come back with nothing, because nothing in the question named a repository. The five top-ranked facts from a seed now sit in the context the agent always carries, so they ground an answer without being prompted for by name; the rest stay searchable exactly as before and still surface when you ask about that specific project. Facts also keep the order they were distilled in, which matters because each reply carries only the ten strongest memories, and until now that cut was made arbitrarily among facts that all scored the same. Nothing about what is read or stored changes: the seed still reads only the profile, repositories, and READMEs you ticked, and disconnecting GitHub still deletes every fact it wrote across all your agents.</description></item>+ <item>+ <title>Your agent's memory graph was always empty. It is not any more</title>+ <link>https://three.ws/docs/memory</link>+ <guid isPermaLink="false">2026-08-14:Your agent's memory graph was always empty. It is not any more</guid>+ <pubDate>Fri, 14 Aug 2026 12:00:00 GMT</pubDate>+ <category>fix</category>+ <description>Every memory your agent stores is scanned for the things it mentions: tokens, tickers, wallets, people, and the trading rules you told it to follow. Those become the nodes of a knowledge graph you can browse, so you can ask what your agent remembers about one specific token and read back every memory that touched it. The scan ran correctly and then failed at the last step, when it tried to save what it found, because of a type mismatch on one database column. The failure was recorded quietly rather than shown, so nothing ever looked broken: memories saved fine, search worked fine, and the graph simply came back empty for every agent, forever. The save now works, memories that were skipped get scanned the next time you open the graph, and a test covers the exact shape that broke.</description>+ </item><item><title>Your agent's memory graph was never being built, and now it is</title><link>https://three.ws/memory</link>@@ -8460,7 +8500,7 @@<guid isPermaLink="false">2026-07-14:End-User License Agreement</guid><pubDate>Tue, 14 Jul 2026 12:00:00 GMT</pubDate><category>launch</category>- <description>License terms for the three.ws Seeker/Saga app.</description>+ <description>License terms for the three.ws Seeker/Saga mobile app: what the license covers, camera and selfie data, updates, and termination.</description></item><item><title>"Deploy onchain" now opens the real deploy flow with your model loaded</title>@@ -16214,6 +16254,14 @@<category>launch</category><description>Configure your 3D agent to trade Oracle conviction automatically: set the minimum score, position size, daily and open-position caps, narrative filters and Telegram alerts, then arm it in simulate or live mode. Every action is graded against the outcome.</description></item>+ <item>+ <title>Content Policy</title>+ <link>https://three.ws/legal/content-policy</link>+ <guid isPermaLink="false">2026-06-21:Content Policy</guid>+ <pubDate>Sun, 21 Jun 2026 12:00:00 GMT</pubDate>+ <category>launch</category>+ <description>What you may publish, mint, and monetise on three.ws: prohibited content, agent behaviour, marketplace listings, and how we enforce.</description>+ </item><item><title>Credits</title><link>https://three.ws/credits</link>@@ -23127,12 +23175,12 @@<description>What data three.ws collects, why we hold it, who we share it with, how long we keep it, and the controls you have over it.</description></item><item>- <title>Terms of Use</title>+ <title>Terms of Service</title><link>https://three.ws/legal/tos</link>- <guid isPermaLink="false">2026-04-27:Terms of Use</guid>+ <guid isPermaLink="false">2026-04-27:Terms of Service</guid><pubDate>Mon, 27 Apr 2026 12:00:00 GMT</pubDate><category>launch</category>- <description>Terms of service for three.ws.</description>+ <description>The agreement between you and three.ws: your account, acceptable use, payments and on-chain activity, liability, and DMCA takedowns.</description></item><item><title>Tutorial · Build a custom skill</title>diff --git a/public/features.json b/public/features.jsonindex f04e657ff..51c2e2d9c 100644--- a/public/features.json+++ b/public/features.json@@ -1,5 +1,5 @@{- "generated_at": "2026-08-14T18:22:30.133Z",+ "generated_at": "2026-08-14T19:37:07.244Z","generated_by": "scripts/build-page-index.mjs from data/pages.json","site": {"name": "three.ws",@@ -6984,7 +6984,7 @@{"path": "/legal/eula","title": "End-User License Agreement",- "description": "License terms for the three.ws Seeker/Saga app.",+ "description": "License terms for the three.ws Seeker/Saga mobile app: what the license covers, camera and selfie data, updates, and termination.","added": "2026-07-14","auth": null,"indexable": true,@@ -6993,8 +6993,8 @@},{"path": "/legal/tos",- "title": "Terms of Use",- "description": "Terms of service for three.ws.",+ "title": "Terms of Service",+ "description": "The agreement between you and three.ws: your account, acceptable use, payments and on-chain activity, liability, and DMCA takedowns.","added": "2026-04-27","auth": null,"indexable": true,@@ -7014,8 +7014,8 @@{"path": "/legal/content-policy","title": "Content Policy",- "description": "What you may publish, mint, and monetise on the platform.",- "added": null,+ "description": "What you may publish, mint, and monetise on three.ws: prohibited content, agent behaviour, marketplace listings, and how we enforce.",+ "added": "2026-06-21","auth": null,"indexable": true,"tags": [],