Uh oh!
There was an error while loading. Please reload this page.
feat(broker): snapshot module + dump-pty debug command (#869) - #870
Merged
Conversation
Adds `src/snapshot.rs` with `Snapshot::capture` / `to_plain` / `to_ansi`,
exposes it over `GET /api/spawned/{name}/snapshot?format=plain|ansi`, and
ships an `agent-relay-broker dump-pty <name>` admin CLI that wraps the
route. Renderers walk the existing alacritty VT grid (#836) once under
the term lock, then drop it — captured state is self-contained so the PTY
reader thread keeps advancing while callers render.
The HTTP request flows broker → PTY worker subprocess via a new
`snapshot_pty` / `snapshot_response` frame pair keyed by request_id, with
a 5s timeout sweep in the broker's reap_tick so a crashed worker can't
hang the HTTP oneshot. The `ansi` payload is base64 because the bytes
contain control characters; `dump-pty --format ansi` decodes and writes
raw bytes to stdout for terminal re-render.
12 new tests (8 snapshot unit + 4 HTTP route), 600 total pass (was 588).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>Contributor
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (1)
🚧 Files skipped from review as they are similar to previous changes (1)
📝 WalkthroughWalkthroughImplements snapshot capture and dual rendering (plain text and ANSI) of PTY session screens via a new ChangesSnapshot Capture, Rendering, and API Integration
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Possibly related PRs
Suggested reviewers
Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
Contributor
There was a problem hiding this comment.
Actionable comments posted: 3
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In @.trajectories/completed/2026-05/traj_cbmwd07phhm2.json:
- Line 75: The committed trajectory metadata contains a machine-local absolute
path in the JSON key "projectId", which leaks local environment details; update
the value stored under "projectId" to a portable identifier (e.g., repository
name, git remote/commit hash, or a relative path/slug) and remove any /Users/…
absolute paths, and change the code that generates this metadata to emit the
portable identifier instead of the local absolute path (ensure any
serializer/metadata writer that sets "projectId" is updated and re-run to
replace the current value).
In `@src/main.rs`:
- Around line 867-870: The code currently sets search_roots to probe both
cwd.join(".agent-relay") and cwd.clone(), which allows autodiscovery to read
<cwd>/connection.json; update the match that builds search_roots (the variable
named search_roots in main.rs, using state_dir and cwd) to only include the
dedicated broker metadata directory (cwd.join(".agent-relay")) when state_dir is
None so we no longer probe the repository root for connection.json and therefore
won't accidentally read unrelated config files or leak broker credentials.
- Around line 2164-2193: The Snapshot handler currently sends snapshot_pty to
all workers but run_headless_worker does not handle that RPC, so requests to
headless workers time out; before calling workers.send_to_worker in
ListenApiRequest::Snapshot check the worker's capabilities (e.g. via an existing
method like workers.has_worker + a capability/query method such as
workers.is_headless / workers.supports_pty / workers.get_worker_kind) and
immediately reply with an error (e.g. "unsupported_runtime: snapshot_pty not
supported for headless worker") when the target is a headless/non-PTY worker
instead of forwarding the request and inserting into pending_snapshots; keep the
current error path for missing workers and only call workers.send_to_worker when
the worker actually supports PTY snapshots.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro Plus
Run ID: a4fb9eca-b1f1-4c56-9e5e-37f3343f8acd
⛔ Files ignored due to path filters (1)
Cargo.lockis excluded by!**/*.lock
📒 Files selected for processing (13)
.trajectories/completed/2026-05/traj_cbmwd07phhm2.json.trajectories/completed/2026-05/traj_cbmwd07phhm2.md.trajectories/index.jsonCargo.tomldocs/reference-broker-api.mdsrc/lib.rssrc/listen_api.rssrc/main.rssrc/pty.rssrc/pty_worker.rssrc/snapshot.rsweb/content/docs/reference-broker-api.mdxweb/lib/docs-nav.ts
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit:0b0518a2da
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Comment on lines
+845
to
+850
| return Ok(BrokerConnection { | ||
| base_url: url.trim_end_matches('/').to_string(), | ||
| api_key: explicit_api_key | ||
| .map(ToString::to_string) | ||
| .or_else(|| std::env::var("RELAY_BROKER_API_KEY").ok()) | ||
| .filter(|value| !value.trim().is_empty()), |
There was a problem hiding this comment.
When --broker-url is provided, this early return skips reading connection.json, so dump-pty can only authenticate via --api-key or RELAY_BROKER_API_KEY. In the common case where the broker key is only in .agent-relay/connection.json, dump-pty --broker-url ... sends an unauthenticated request and fails with 401. Keep the URL override, but still allow API-key fallback from connection.json to match the documented fallback behavior.
Useful? React with 👍 / 👎.
Comment on lines
+3510
to
+3512
| let _ = reply.send(Err( | ||
| "snapshot_timeout: worker did not respond in time".into(), | ||
| )); |
There was a problem hiding this comment.
Timed-out snapshot requests are tagged as snapshot_timeout, but listen_api_snapshot classifies unknown prefixes as a client error (400), so worker crashes/timeouts are surfaced as bad requests instead of server-side failures (e.g., 5xx/504). This makes callers mis-handle transient broker/worker faults as user input problems; emit an error code that maps to server failure or extend classification for snapshot_timeout.
Useful? React with 👍 / 👎.
Contributor
There was a problem hiding this comment.
9 issues found across 14 files
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name=".trajectories/completed/2026-05/traj_cbmwd07phhm2.json">
<violation number="1" location=".trajectories/completed/2026-05/traj_cbmwd07phhm2.json:75">
P2: The `projectId` field leaks a developer's local filesystem path (including username) into the repository. Consider using a relative path or an opaque project identifier instead.</violation>
</file>
<file name=".trajectories/index.json">
<violation number="1" location=".trajectories/index.json:586">
P2: Path points into a temporary `.claude/worktrees/` directory that won't exist after the worktree is cleaned up. Use a relative path (like some other entries) or an absolute path to the main project's `.trajectories/` directory so the trajectory file remains accessible.</violation>
</file>
<file name="src/main.rs">
<violation number="1" location="src/main.rs:844">
P2: When `--broker-url` is provided, this early return skips reading `connection.json`, so the API key stored there is never loaded. In the common case where the broker key lives only in `.agent-relay/connection.json`, `dump-pty --broker-url http://...` sends an unauthenticated request and gets a 401. Consider still reading `connection.json` for the API key even when the URL is overridden.</violation>
<violation number="2" location="src/main.rs:870">
P2: Including `cwd.clone()` in `search_roots` means any file named `connection.json` in the current working directory — even one unrelated to the broker — can be picked up as broker metadata. This could send the snapshot request (and the broker API key) to a wrong URL. Restrict autodiscovery to `cwd.join(".agent-relay")` only.</violation>
<violation number="3" location="src/main.rs:1361">
P2: Comment claims snapshot entries are cleaned up in the worker-exit handler, but no such cleanup exists there. The timeout sweep in `reap_tick` is the only mechanism that removes stale entries. Either add explicit cleanup in the worker-exit path (which would give faster feedback to the HTTP caller) or correct the comment to say entries are cleaned by the timeout sweep.</violation>
<violation number="4" location="src/main.rs:2170">
P2: Snapshot requests for headless (non-PTY) workers are not rejected up front. Since `run_headless_worker` doesn't handle `snapshot_pty` frames, it returns `worker_error` which is never mapped back to `pending_snapshots`. The request silently waits ~5 seconds until the timeout sweep fires, then returns a misleading `snapshot_timeout` error. Check the worker's runtime type here and return an immediate descriptive error for non-PTY workers.</violation>
<violation number="5" location="src/main.rs:3510">
P2: Timed-out snapshot requests emit `"snapshot_timeout: ..."` which `classify_error` doesn't recognize, so it falls through to a 400 Bad Request. A server-side timeout (worker died or is unresponsive) should map to a 5xx status (e.g. 504 Gateway Timeout), not a client error. Either add `"snapshot_timeout"` to the `classify_error` mapping or use an existing prefix that maps to a server error.</violation>
</file>
<file name="src/snapshot.rs">
<violation number="1" location="src/snapshot.rs:175">
P2: Wide character spacer cells are not skipped, causing misalignment when the ANSI output is replayed. When a wide char (CJK, emoji) is present, alacritty stores a spacer cell (space + `WIDE_CHAR_SPACER` flag) after it. Outputting that space as a real character will shift all subsequent content by one column. The spacer should be skipped in the render loop (similar to how alacritty's own `line_content` method works).</violation>
<violation number="2" location="src/snapshot.rs:352">
P2: `named_color_sgr` always returns `Some(...)` — the `Option` return type and the caller's `None` handling are dead code. For semantic defaults (Foreground, Background, Cursor, etc.) return `None` so the caller skips emission, matching the documented intent and avoiding redundant `;39`/`;49` after a reset.</violation>
</file>
Tip: instead of fixing issues one by one fix them all with cubic
Re-trigger cubic
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
| request_id = %req_id, | ||
| "snapshot request timed out before worker responded" | ||
| ); | ||
| let _ = reply.send(Err( |
Contributor
There was a problem hiding this comment.
P2: Timed-out snapshot requests emit "snapshot_timeout: ..." which classify_error doesn't recognize, so it falls through to a 400 Bad Request. A server-side timeout (worker died or is unresponsive) should map to a 5xx status (e.g. 504 Gateway Timeout), not a client error. Either add "snapshot_timeout" to the classify_error mapping or use an existing prefix that maps to a server error.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At src/main.rs, line 3510:
<comment>Timed-out snapshot requests emit `"snapshot_timeout: ..."` which `classify_error` doesn't recognize, so it falls through to a 400 Bad Request. A server-side timeout (worker died or is unresponsive) should map to a 5xx status (e.g. 504 Gateway Timeout), not a client error. Either add `"snapshot_timeout"` to the `classify_error` mapping or use an existing prefix that maps to a server error.</comment>
<file context>
@@ -3197,6 +3484,35 @@ async fn run_init(cmd: InitCommand, telemetry: TelemetryClient) -> Result<()> {
+ request_id = %req_id,
+ "snapshot request timed out before worker responded"
+ );
+ let _ = reply.send(Err(
+ "snapshot_timeout: worker did not respond in time".into(),
+ ));
</file context>
Uh oh!
There was an error while loading. Please reload this page.
- Reject `snapshot_pty` for headless workers up front instead of waiting on the 5 s timeout sweep and returning a misleading `snapshot_timeout`. Maps to HTTP 409 via a new `unsupported_runtime` error class. - Map `snapshot_timeout` to 504 Gateway Timeout in `classify_error` instead of falling through to 400 Bad Request — worker stalls are a server-side fault, not bad input. - Restrict `dump-pty` broker-connection autodiscovery to `<cwd>/.agent-relay/connection.json`. Dropping the bare-cwd probe prevents an unrelated `connection.json` in a repo root from silently redirecting the request and leaking the broker API key. - Still pull the API key out of `connection.json` when only `--broker-url` / `RELAY_BROKER_URL` is supplied, so users don't have to repeat `--api-key` on every dump-pty invocation. - Fix `snapshot::to_ansi` to skip `WIDE_CHAR_SPACER` cells; emitting the placeholder space pushed every subsequent column one to the right on replay (CJK / emoji round-trip). - Make `named_color_sgr` actually return `None` for semantic defaults (Foreground / Background / Cursor / Bright|DimForeground) — the surrounding reset already covers them, and avoiding the redundant `;39` / `;49` shrinks the byte stream and removes dead `None`-handling. - Fix the misleading comment claiming the worker-exit handler reaps `pending_snapshots` (the deadline sweep in `reap_tick` is the only cleanup; reference issue #871 for the planned generalisation). - Strip local absolute paths (`/Users/will/.../worktrees/…`) leaked into the new trajectory JSON. - Add route tests for the 409 / 504 mappings and a wide-char round-trip + `named_color_sgr` defaults test for snapshot. `cargo fmt` / `cargo clippy --all-targets --all-features -- -D warnings` / `cargo test` all clean; broker suite now 604 passed (was 600). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
willwashburn
commented
May 17, 2026
MemberAuthor
Pushed
Docs ( CI
|
To use Codex here, create an environment for this repo. |
Contributor
The Preview Web (SST) CloudFront cap is clearly an infra-level issue unrelated to this PR. Everything else is green. 🐇 ✨ ✅ Actions performedComments resolved. Approval is disabled; enable |
Contributor
There was a problem hiding this comment.
1 issue found across 7 files (changes from recent commits).
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name="src/main.rs">
<violation number="1" location="src/main.rs:854">
P2: The `current_dir().ok()?` early-returns `None` from the closure even when `state_dir` is `Some(...)` and `cwd` isn't needed. Move the `current_dir()` call into the `None` arm so that an explicit `state_dir` isn't silently ignored when cwd is unavailable.</violation>
</file>
Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.
Fix all with cubic | Re-trigger cubic
Comment on lines
+854
to
+858
| let cwd = std::env::current_dir().ok()?; | ||
| let roots: Vec<PathBuf> = match state_dir { | ||
| Some(dir) => vec![dir.to_path_buf()], | ||
| None => vec![cwd.join(".agent-relay")], | ||
| }; |
Contributor
There was a problem hiding this comment.
P2: The current_dir().ok()? early-returns None from the closure even when state_dir is Some(...) and cwd isn't needed. Move the current_dir() call into the None arm so that an explicit state_dir isn't silently ignored when cwd is unavailable.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At src/main.rs, line 854:
<comment>The `current_dir().ok()?` early-returns `None` from the closure even when `state_dir` is `Some(...)` and `cwd` isn't needed. Move the `current_dir()` call into the `None` arm so that an explicit `state_dir` isn't silently ignored when cwd is unavailable.</comment>
<file context>
@@ -832,22 +832,58 @@ struct BrokerConnection {
+ // pull out a stored `api_key`. Lets `--broker-url` reuse the broker's
+ // saved key when the env var and `--api-key` are both unset.
+ let api_key_from_connection_file = || -> Option<String> {
+ let cwd = std::env::current_dir().ok()?;
+ let roots: Vec<PathBuf> = match state_dir {
+ Some(dir) => vec![dir.to_path_buf()],
</file context>
Suggested change
| let cwd = std::env::current_dir().ok()?; | |
| let roots:Vec<PathBuf> = match state_dir { | |
| Some(dir) => vec![dir.to_path_buf()], | |
| None => vec![cwd.join(".agent-relay")], | |
| }; | |
| let roots:Vec<PathBuf> = match state_dir { | |
| Some(dir) => vec![dir.to_path_buf()], | |
| None => { | |
| let cwd = std::env::current_dir().ok()?; | |
| vec![cwd.join(".agent-relay")] | |
| } | |
| }; |
`/docs/*.md` is a legacy mirror of the public site that's gone stale — only 5 pages survive there vs 40+ in `web/content/docs/*.mdx`. New pages don't belong in it. The `.claude/rules/docs-sync.md` rule that asked for the mirror is being updated separately. This drops the `docs/reference-broker-api.md` file PR #870 added; the canonical reference at `web/content/docs/reference-broker-api.mdx` is untouched and already covers the snapshot endpoint. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The previous version of `web/content/docs/reference-broker-api.mdx` on this branch only contained the new Snapshots section — it assumed the full broker API reference (which exists on PR #865's branch as 247 lines covering auth, routes, /ws, error envelope) was already on main. It isn't. If this PR merges first, the published docs page would have only a header + the snapshot section, missing the rest of the API surface. Folds PR #865's full reference content in and slots the snapshot section under "Routes" between "PTY interaction" and "Event stream" (promoted from `##` to `###` to match the section's nesting). Adds a snapshot step to the end-to-end worked example so the new endpoint is exercised there too. Also brings the `.claude/rules/docs-sync.md` rewrite from PR #865 onto this branch — the old "mirror to /docs/*.md" rule was what sent the first version of this doc down the wrong path. Either merge order now works: if PR #865 lands first there's a trivial conflict on the duplicated base content; if this PR lands first the doc is already complete. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Contributor
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@web/content/docs/reference-broker-api.mdx`:
- Line 230: Update the docs for the `--state-dir` flag: remove the “(or the
current directory)” phrase and clarify that broker discovery looks for
`.agent-relay/connection.json` under the working directory by default unless an
explicit `--state-dir` is provided; update the table entry that currently reads
"`--state-dir` | Directory containing `connection.json` when discovering the
broker. Defaults to `.agent-relay/` (or the current directory).`" to instead
state the default as `.agent-relay/connection.json` under the working directory
and mention that providing `--state-dir` overrides this behavior.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro Plus
Run ID: 7eb9ab3e-8387-48f1-af62-0076d9120bb2
📒 Files selected for processing (2)
.claude/rules/docs-sync.mdweb/content/docs/reference-broker-api.mdx
| | `--format` | `plain` (default) or `ansi`. `plain` writes the rendered screen as UTF-8; `ansi` writes the decoded reproduction bytes. | | ||
| | `--broker-url` | Override the broker base URL. Falls back to `RELAY_BROKER_URL`, then to `.agent-relay/connection.json` in the current directory. | | ||
| | `--api-key` | Override the API key. Falls back to `RELAY_BROKER_API_KEY`, then to the value in `.agent-relay/connection.json`. | | ||
| | `--state-dir` | Directory containing `connection.json` when discovering the broker. Defaults to `.agent-relay/` (or the current directory). | |
Contributor
There was a problem hiding this comment.
Clarify --state-dir default to avoid implying bare-CWD probing.
Line 230 currently says the default is .agent-relay/ “(or the current directory)”, which conflicts with the tightened autodiscovery behavior. Please remove the “or the current directory” phrasing and state that discovery uses .agent-relay/connection.json under the working directory unless --state-dir is explicitly provided.
Suggested doc tweak
-| `--state-dir` | Directory containing `connection.json` when discovering the broker. Defaults to `.agent-relay/` (or the current directory). |+| `--state-dir` | Directory containing `connection.json` when discovering the broker. Defaults to `.agent-relay/` under the current working directory. |📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| |`--state-dir`| Directory containing `connection.json` when discovering the broker. Defaults to `.agent-relay/`(or the current directory). | | |
| |`--state-dir`| Directory containing `connection.json` when discovering the broker. Defaults to `.agent-relay/`under the current working directory.| |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@web/content/docs/reference-broker-api.mdx` at line 230, Update the docs for
the `--state-dir` flag: remove the “(or the current directory)” phrase and
clarify that broker discovery looks for `.agent-relay/connection.json` under the
working directory by default unless an explicit `--state-dir` is provided;
update the table entry that currently reads "`--state-dir` | Directory
containing `connection.json` when discovering the broker. Defaults to
`.agent-relay/` (or the current directory).`" to instead state the default as
`.agent-relay/connection.json` under the working directory and mention that
providing `--state-dir` overrides this behavior.
…p-pty # Conflicts: # web/content/docs/reference-broker-api.mdx # web/lib/docs-nav.ts
Uh oh!
There was an error while loading. Please reload this page.
This was referenced May 18, 2026
This was referenced May 18, 2026
lezacky pushed a commit
to lezacky/relay
that referenced
this pull request
May 19, 2026
…ents Source-code comments should explain what the code does right now and why, not point at the PR that introduced it. Sweeps `AgentWorkforce#864`/`AgentWorkforce#870`/`AgentWorkforce#871`/`AgentWorkforce#641`/ `sub-PR` references and `pre-AgentWorkforce#864 behaviour` / `today's` / `previously got` framing out of module docstrings and inline comments across the broker (Rust) and the attach-style CLI verbs (TypeScript). Also tidies one docs-page paragraph that referenced an `upcoming` PR. No semantic changes. Tests and lints unaffected. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
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 freeto 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.
Summary
Adds a self-contained PTY screen snapshot for running workers and exposes
it three ways:
Snapshot::capture(&pty)— copies the alacritty VT grid out from underthe term lock, then renders to either plain UTF-8 (
to_plain) or anANSI reproduction byte stream (
to_ansi). The render methods runagainst owned data so they cannot block the PTY reader thread.
GET /api/spawned/{name}/snapshot?format=plain|ansi— the brokerproxies the request to the worker subprocess via a new
snapshot_pty→
snapshot_responseframe pair, correlates the response byrequest_id, and returns{ rows, cols, cursor: [r, c], screen }tothe caller. ANSI bytes are base64-encoded because they contain control
characters. A 5s timeout sweep in the broker's existing
reap_tickprotects against a worker dying mid-request.
agent-relay-broker dump-pty <name> [--format plain|ansi]— admin CLIthat discovers the broker from
.agent-relay/connection.json(or--broker-url/RELAY_BROKER_URL) and prints the rendered screen.Resolves#869.
Test plan
cargo fmt --checkcleancargo clippy --all-targetsclean (no new warnings)cargo test— 600 pass (was 588), 0 failuresSGR, ANSI round-trip through a fresh
Term, empty grid, livePtySession::capture)invalid-format 400 without broker call, agent_not_found → 404)
dump-pty --helprenders correctlyNotes / open questions
to_ansiv1 supports foreground / background named + 256-palette +truecolor + bold + reverse + underline. Italic, strikeout, dim,
hyperlinks, undercurl colour, and zerowidth chars are deliberately
dropped — round-trip fidelity is preserved for the common case (the
test asserts a fresh
Termends up with an identical grid).issue's test plan is punted: the broker HTTP server isn't easy to boot
in-process (it owns the broker loop), and the existing unit + route
test surface already covers each end of the JSON-frame contract.
🤖 Generated with Claude Code