<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
  <id>https://kinherd.dev/changelog/</id>
  <title>Kinherd changelog</title>
  <subtitle>What shipped, newest first.</subtitle>
  <link rel="self" type="application/atom+xml" href="https://kinherd.dev/changelog.xml"/>
  <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/"/>
  <updated>2026-11-20T00:00:00Z</updated>
  <author><name>Kinherd</name></author>
  <entry>
    <id>https://kinherd.dev/changelog/#node-v0.2.10</id>
    <title>node-v0.2.10: A node in three commands</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#node-v0.2.10"/>
    <updated>2026-11-20T00:00:00Z</updated>
    <category term="node" label="Linux node"/>
    <category term="android" label="Android app"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;aigw-node pair, install and up take a machine from nothing to a serving node: pair with the hub, install the service (systemd on Linux, launchd on macOS), then add the chat models found on the machine and wait until the hub has the node online. Each command is safe to run again.&lt;/li&gt;&lt;li&gt;A &amp;quot;tools&amp;quot; setting on a chat model starts llama.cpp with its chat template, so tool calls work on the Anthropic, Responses and chat completions APIs.&lt;/li&gt;&lt;li&gt;The Android app starts serving when it is opened, also on TVs whose firmware blocks apps from starting themselves after an update.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#2026-11-20-anthropic-messages-and-openai-responses-apis-tool-calls</id>
    <title>Anthropic Messages and OpenAI Responses APIs, tool calls</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#2026-11-20-anthropic-messages-and-openai-responses-apis-tool-calls"/>
    <updated>2026-11-20T00:00:00Z</updated>
    <category term="hub" label="Hub"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;Chat models on your nodes now also answer the Anthropic Messages API (/anthropic/v1/messages, streaming and tool use included), so clients built for it can point at the hub.&lt;/li&gt;&lt;li&gt;The OpenAI Responses API (/openai/v1/responses) works on the same models, stateless: every request carries its whole conversation.&lt;/li&gt;&lt;li&gt;Tool calls pass through to llama.cpp on every API, with limits on the number and size of tools, and streamed tool calls are indexed the way OpenAI SDKs expect.&lt;/li&gt;&lt;li&gt;The same keys, scopes, local-only rule, quotas and audit apply as on chat completions.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#node-v0.2.9</id>
    <title>node-v0.2.9: Nodes stay on the LAN address that answers</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#node-v0.2.9"/>
    <updated>2026-11-20T00:00:00Z</updated>
    <category term="hub" label="Hub"/>
    <category term="node" label="Linux node"/>
    <category term="android" label="Android app"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;The hub advertises only the LAN address of the interface that carries its default route (lan.allInterfaces restores all interfaces; an optional CIDR allow list narrows what is advertised), so nodes no longer spend a dial on an address they cannot reach.&lt;/li&gt;&lt;li&gt;When a node has fallen back to the public URL, its check for the LAN path backs off: the interval doubles after each round without an answer, from 2 up to 30 minutes, on the Android app and the Linux node.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#2026-11-20-free-with-no-paid-tiers</id>
    <title>Free, with no paid tiers</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#2026-11-20-free-with-no-paid-tiers"/>
    <updated>2026-11-20T00:00:00Z</updated>
    <category term="website" label="Website"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;The pricing page now says what is true today: the hub and the nodes are free to run, and there are no paid tiers.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#2026-11-20-website-live-served-by-the-hub</id>
    <title>Website live, served by the hub</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#2026-11-20-website-live-served-by-the-hub"/>
    <updated>2026-11-20T00:00:00Z</updated>
    <category term="website" label="Website"/>
    <category term="hub" label="Hub"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;The website, documentation, compatibility registry and in-browser demo are public. The hub serves them itself for every path no API route claims.&lt;/li&gt;&lt;li&gt;Strict Content Security Policy with hashed inline scripts, precompressed files and atomic deploys with rollback.&lt;/li&gt;&lt;li&gt;Download links always point at the current stable release through /downloads/latest/.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#node-v0.2.8</id>
    <title>node-v0.2.8: End-to-end sealing, chat models and signed checksums</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#node-v0.2.8"/>
    <updated>2026-11-19T00:00:00Z</updated>
    <category term="node" label="Linux node"/>
    <category term="android" label="Android app"/>
    <category term="security" label="Security"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;Nodes open end-to-end sealed requests (HPKE, per-node job keys); the hub forwards a sealed body without being able to read it.&lt;/li&gt;&lt;li&gt;The Linux node runs chat models through a managed llama-server and streams the answer back through the hub.&lt;/li&gt;&lt;li&gt;Release checksums are signed (SHA256SUMS.sig) and verified by the installer and the node before an update.&lt;/li&gt;&lt;li&gt;Automatic benchmarks wait until the device owner’s protections allow them (nothing loads while the node is paused).&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#2026-11-19-end-to-end-sealing</id>
    <title>End-to-end sealing</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#2026-11-19-end-to-end-sealing"/>
    <updated>2026-11-19T00:00:00Z</updated>
    <category term="hub" label="Hub"/>
    <category term="sdk" label="SDKs"/>
    <category term="security" label="Security"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;Clients can seal a request to the node that runs it. The hub routes, meters and audits the job but cannot read the prompt or the answer.&lt;/li&gt;&lt;li&gt;Client helpers for Go, Python and the Android library, with shared test vectors; they refuse to send without a pinned node key.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#2026-11-19-openai-compatible-chat-on-your-own-devices</id>
    <title>OpenAI-compatible chat on your own devices</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#2026-11-19-openai-compatible-chat-on-your-own-devices"/>
    <updated>2026-11-19T00:00:00Z</updated>
    <category term="hub" label="Hub"/>
    <category term="node" label="Linux node"/>
    <category term="sdk" label="SDKs"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;POST /openai/v1/chat/completions runs on chat models served by your nodes, streamed or not, under the same keys, scopes and quotas as every other call.&lt;/li&gt;&lt;li&gt;Thin Python and JavaScript helpers on top of the official OpenAI clients.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#2026-11-19-trusted-groups</id>
    <title>Trusted groups</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#2026-11-19-trusted-groups"/>
    <updated>2026-11-19T00:00:00Z</updated>
    <category term="hub" label="Hub"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;Accounts can share their devices with family and friends in a trusted group; your own devices are always tried first.&lt;/li&gt;&lt;li&gt;Invitations are capped per inviter, so one account cannot flood someone else’s invitations.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#2026-11-19-self-service-accounts-with-passkeys</id>
    <title>Self-service accounts with passkeys</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#2026-11-19-self-service-accounts-with-passkeys"/>
    <updated>2026-11-19T00:00:00Z</updated>
    <category term="hub" label="Hub"/>
    <category term="security" label="Security"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;Sign in at /account with a passkey, with recovery codes for a lost device. No passwords.&lt;/li&gt;&lt;li&gt;Owner dashboard: create client keys, pair devices and see their status.&lt;/li&gt;&lt;li&gt;Operator controls for accounts and invitations at /admin/accounts.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#2026-11-19-signed-releases-and-safer-installs</id>
    <title>Signed releases and safer installs</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#2026-11-19-signed-releases-and-safer-installs"/>
    <updated>2026-11-19T00:00:00Z</updated>
    <category term="hub" label="Hub"/>
    <category term="node" label="Linux node"/>
    <category term="security" label="Security"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;Every release ships a signed SHA256SUMS; the hub checks the signature when a release is uploaded, and install.sh verifies it again on the device.&lt;/li&gt;&lt;li&gt;The installer refuses downgrades and stops when a signature, checksum or version cannot be verified.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#2026-11-19-zero-config-lan-and-a-trusted-public-registry</id>
    <title>Zero-config LAN and a trusted public registry</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#2026-11-19-zero-config-lan-and-a-trusted-public-registry"/>
    <updated>2026-11-19T00:00:00Z</updated>
    <category term="hub" label="Hub"/>
    <category term="node" label="Linux node"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;The hub listens on the LAN by default and announces itself over mDNS, so a Linux node can find it for pairing without typing an address. Discovered hubs are never trusted automatically: pairing checks the certificate fingerprint.&lt;/li&gt;&lt;li&gt;The public registry decides a device by its hardware probe fingerprint, and the operator marks which owners’ results count as trusted.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#node-v0.2.7</id>
    <title>node-v0.2.7: Android start-up fix</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#node-v0.2.7"/>
    <updated>2026-11-19T00:00:00Z</updated>
    <category term="android" label="Android app"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;Release builds keep their local databases intact, which fixes a crash at start on phones.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#node-v0.2.6</id>
    <title>node-v0.2.6: Check for updates on demand</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#node-v0.2.6"/>
    <updated>2026-11-19T00:00:00Z</updated>
    <category term="android" label="Android app"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;Settings &amp;gt; Version can check for an app update right away instead of waiting for the next scheduled check.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#node-v0.2.5</id>
    <title>node-v0.2.5: In-app update test release</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#node-v0.2.5"/>
    <updated>2026-11-19T00:00:00Z</updated>
    <category term="android" label="Android app"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;A release with no code changes, published to exercise the in-app update path end to end.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#node-v0.2.4</id>
    <title>node-v0.2.4: Publisher keys survive restarts</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#node-v0.2.4"/>
    <updated>2026-11-19T00:00:00Z</updated>
    <category term="android" label="Android app"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;Publisher keys the hub offers to a device stay pending across restarts until the owner answers.&lt;/li&gt;&lt;li&gt;Offers lost on devices paired with 0.2.3 are recovered from the hub.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#node-v0.2.3</id>
    <title>node-v0.2.3: TV home screen fix</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#node-v0.2.3"/>
    <updated>2026-11-19T00:00:00Z</updated>
    <category term="android" label="Android app"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;The TV home screen shows the recent jobs card again; the release build had crashed on the TV.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#node-v0.2.2</id>
    <title>node-v0.2.2: Same-device calls across signing keys</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#node-v0.2.2"/>
    <updated>2026-11-19T00:00:00Z</updated>
    <category term="android" label="Android app"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;Apps on the same device can call the gateway over IPC when they are signed by a known key, not only by the gateway’s own key.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#node-v0.2.1</id>
    <title>node-v0.2.1: Permanent release signing key</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#node-v0.2.1"/>
    <updated>2026-11-19T00:00:00Z</updated>
    <category term="android" label="Android app"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;Release APKs are signed with the permanent release key, so later versions install as updates.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#node-v0.2.0</id>
    <title>node-v0.2.0: Self-updating nodes</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#node-v0.2.0"/>
    <updated>2026-11-19T00:00:00Z</updated>
    <category term="node" label="Linux node"/>
    <category term="android" label="Android app"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;First release with self-update: nodes fetch new versions from their own hub, try them and roll back if the new version does not come up.&lt;/li&gt;&lt;li&gt;The hub serves installs and updates itself (one-line install for Linux nodes).&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#2026-11-18-public-registry-waitlist-and-packaging</id>
    <title>Public registry, waitlist and packaging</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#2026-11-18-public-registry-waitlist-and-packaging"/>
    <updated>2026-11-18T00:00:00Z</updated>
    <category term="hub" label="Hub"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;Public read-only registry (GET /public/v1/registry) and a waitlist endpoint, with a CORS allow-list and per-IP limits.&lt;/li&gt;&lt;li&gt;GET /v1/nodes is redacted for non-admin keys; /metrics on the public port is admin-only.&lt;/li&gt;&lt;li&gt;Distroless hub image, Docker Swarm stack and a deploy script that verifies the live commit.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#2026-11-18-server-hub-and-outbound-nodes</id>
    <title>Server hub and outbound nodes</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#2026-11-18-server-hub-and-outbound-nodes"/>
    <updated>2026-11-18T00:00:00Z</updated>
    <category term="hub" label="Hub"/>
    <category term="node" label="Linux node"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;Go server edition of the hub: SQLite store, virtual keys (scopes, rate limits, daily quotas, budgets), audit and usage, admin dashboard.&lt;/li&gt;&lt;li&gt;OpenAI-compatible API: /openai/v1/models, /embeddings (image data URLs) and /chat/completions passthrough to a configured provider.&lt;/li&gt;&lt;li&gt;Nodes connect over an outbound WebSocket: no inbound port, works behind NAT and on mobile data.&lt;/li&gt;&lt;li&gt;Router: loaded models first, then the lowest measured latency, then the fewest in flight; failover on transport errors.&lt;/li&gt;&lt;li&gt;Compatibility registry: nodes file their benchmark results (works / works with tweaks / fails) automatically.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#2026-11-18-multi-model-gateway-tv-support-and-the-client-library</id>
    <title>Multi-model gateway, TV support and the client library</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#2026-11-18-multi-model-gateway-tv-support-and-the-client-library"/>
    <updated>2026-11-18T00:00:00Z</updated>
    <category term="android" label="Android app"/>
    <category term="sdk" label="SDKs"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;Model packages (manifest schema 1, SHA-256 verified), per-model scheduler with bounded queues, memory budget, lazy load and idle unload.&lt;/li&gt;&lt;li&gt;Two runtime builds: LiteRT 2 for phones (GPU / NPU / CPU) and LiteRT 1.4 + NNAPI for 32-bit TVs.&lt;/li&gt;&lt;li&gt;Fixed 9 s NNAPI runs on the TV; EfficientNet-Lite4 now runs in about 22 ms on its NPU. Per-request timing breakdown in every answer.&lt;/li&gt;&lt;li&gt;Same-device Binder IPC with a signature permission, and the Android client library (IPC, then hub, then app fallback).&lt;/li&gt;&lt;li&gt;mDNS discovery of other gateways, never trusted automatically.&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
  <entry>
    <id>https://kinherd.dev/changelog/#2026-11-18-first-prototype</id>
    <title>First prototype</title>
    <link rel="alternate" type="text/html" href="https://kinherd.dev/changelog/#2026-11-18-first-prototype"/>
    <updated>2026-11-18T00:00:00Z</updated>
    <category term="android" label="Android app"/>
    <content type="html">&lt;ul&gt;&lt;li&gt;Single-model gateway on a Pixel 7 Pro: HTTPS with per-client tokens, LAN-only binding, rate limits.&lt;/li&gt;&lt;li&gt;CLIP ViT-L/14-336 (fp16 weights, 582 MB) on the GPU: about 1.5 s per photo, embeddings identical to the fp32 original (cosine 1.000).&lt;/li&gt;&lt;/ul&gt;</content>
  </entry>
</feed>
