Repository navigation
Peer link: prove a process app hears the system agent's lane live (ADR 0004 §6) - #197
Merged
Merged
Conversation
…R 0004's two lanes A real-kernel test for the process path (ADR 0004 §6): a process app opens its conversation over the peer link (`octos.session.open` without a client); the system agent sends the app's agent a message (`peer_send_input`); the turn reaches the process's link live as `conversation` frames with lane `system_agent` and its speaker: `turn/started` with the request, the user message without the kernel's marker, the answer (`assistant_delta`) and the end (`turn_terminal`), with no history request. The recorded frames are the fixtures of Makepad's client (makepad#77), which now reads them and queues a second message on a handle. The framing test also reads a conversation frame as Makepad writes it. ADR 0004: the two lanes' client items are done in Makepad; process apps get them once the runtime pin reaches it (the repin is pending). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…al-kernel test cleans up on failure Resolves ADR 0004 against main (origin/main's follow-up text kept; the two lanes' items ticked with makepad#77, in the runtime since the makepad e29a0eaa pin, which contains it). The process-app real-kernel test releases its link, broker, kernel, model and temporary files by a drop guard, pass or fail, and its doc says how CI runs it (phone.yml sets OCTOS_SHELL_TEST_KERNEL from the cached pinned octos build). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Contributor
Author
|
Local CI passed on 019aef0 ( GitHub would run for this PR: desktop.yml, phone.yml, apps.yml. Merged on this local pass (macOS runner queue); GitHub CI runs on the merge commit on main.
Notes:
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
ADR 0004's "two lanes, left open" for process apps. The Makepad side is makepad#77: the peer client now reads
conversationframes (both lanes, surfaced asPeerEvent::Conversation) and queues a second message on a handle (at most 8;queue_fullbeyond; closing cancels).The shell already sent the frames and needs no behaviour change: the client queues, and it sends the next message only after the running turn's reply, by which time the broker has freed the handle's turn slot. This PR:
real_kernel_a_process_app_hears_the_system_agents_lane_live(crates/shell/src/host_tools/real_kernel_tests.rs): on a real octos kernel, a process app's peer link (PeerLinks, the real broker) opens its conversation without aclient; the system agent'speer_send_inputturn on the app's agent reaches the process's link live asconversationframes with lanesystem_agentand its speaker:turn/startedwith the request, the user message'sdisplay_text, the answer (assistant_delta) and the end (turn_terminal), before and without anyoctos.session.history. The frames it prints are the fixtures of Makepad's client tests.conversationframe as Makepad's client writes it;wire.rs's doc names the client.OctosContextdirectly is still refused a second turn, and the repin is pending. Octoscript-Makepad #50, a repin of the same chain, is still open, so OctoSense stays on makepad a1c81312 until it lands.Tests run locally, from
phone/:OCTOS_SHELL_TEST_KERNEL=<octos fe08d8e6, built with tools/kernel-artifact.py --host> cargo test --locked --features mobile-apps -p octosense-shell real_kernel -- --test-threads=1: 4 passed (the new test and the three existing ones), on the base before Repin the runtime to makepad a1c81312 (terminal links and search) #193;cargo test --locked --features mobile-apps -p octosense-shell peer_link: 15 passed.🤖 Generated with Claude Code