docs: the assistant answering in Learn, as step 3 built it

Learn's AGENTS.md describes the field with the assistants answering in it
(ADR 0021, step 3): who answers, the scope, the conversation sheet and its
lookups, images, private studysets' warnings, kept conversations and the
tasks; and the testing trap of awaiting a finished subscription's cancel.
The shared packages' READMEs name providerTermsOf and onThisMachine.
Learn's README no longer says there is no end-to-end encryption, which
private studysets changed (ADR 0019).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>

packages/halkyon_app/README.md

@@ -83,7 +83,11 @@ Written as the apps are: Dart 3.13's primary constructors, Material from
   `AssistantGrantsCard` on `InstanceOperatorScreen`: grants with their
   limits, granted, changed and taken back, and everyone's use by month.
   `package:halkyon_app/assistant_models.dart` holds the models alone, in
-  pure Dart, for an app's own models.
+  pure Dart, for an app's own models. For the app's warnings before a
+  private studyset's content goes (ADR 0021, D9): `providerTermsOf` names a
+  provider's terms by its host (checked for Anthropic, OpenAI, Google and
+  OpenRouter, null for any other), and `OwnAssistant.onThisMachine` says
+  whether a person's own assistant runs on this machine.
 - **About** (`showAppAbout`): the app's name (its frame's), its version,
   icon and legalese where it gives them, where the source of what the
   instance runs is (`InstanceSourceCode`, from `AppDependencies.sourceCodeUrl`;

packages/halkyon_llm/CHANGELOG.md

@@ -3,4 +3,4 @@
 * One interface (`LlmClient`) over Anthropic's Messages API, OpenAI's chat completions
   (with each known host's fields, and the local servers') and OpenAI's Responses API,
   streamed over plain HTTP; failures by kind; `FakeLlmClient` for tests; the recording
-  tool.
+  tool; `LlmEndpoint.onThisMachine`.

packages/halkyon_llm/README.md

@@ -54,7 +54,9 @@ removed. Cancelling the subscription aborts the request.
 A base URL is `https`, but for a model on the same machine or network (`localhost`, a
 private or link-local address, a single-label name such as a container's, `.local`,
 `.internal`, `.home.arpa`), as Ollama and llama.cpp serve it; anything else is a
-`FormatException` when the endpoint is made.
+`FormatException` when the endpoint is made. `onThisMachine` says whether its requests stay on
+this machine (a loopback address, or `localhost` as RFC 6761 reserves it), not merely on
+the local network, which may be shared.
 
 - **Chat completions** go to OpenAI, Mistral, Gemini's compatibility endpoint, OpenRouter
   and local servers. OpenAI, Mistral and Gemini refuse fields they don't know and differ