Skip to content

docs(dispatch): correct dispatch-composer-evidence.md and headless-dispatch design.md #78

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
G-Hensley merged 1 commit into from
Sep 11, 2026
Merged

docs(dispatch): correct dispatch-composer-evidence.md and headless-dispatch design.md #78

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 4 additions & 3 deletions docs/design/headless-dispatch/design.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,9 +22,10 @@ worktree, `last_output`, program, and server; `session_cwd` and the final

`dispatch_task` (`src-tauri/src/mcp.rs:968`) wraps the brief
(`dispatch_prompt`, `mcp.rs:73`) and refuses it if it exceeds the current
target's own limit (`Shared::prompt_limit`, `lib.rs:631`; checked via
`oversize_refusal_at`, `mcp.rs:3360` — current as of the composer-fidelity
change, not `2907adf` above): 8192 bytes for a Linux Codex pane
target's own limit (`SessionManager::prompt_limit`, `lib.rs:631`; checked
via `oversize_refusal_at`, `mcp.rs:3360`, current as of the
composer-fidelity change, not `2907adf` above): 8192 bytes for a Linux
Codex pane
(`pane_input::CODEX_LINUX_MAX_BYTES`, `pane_input.rs:7`, evidence in
`docs/dispatch-composer-evidence.md`), 1023 bytes
(`pane_input::LEGACY_MAX_BYTES`, `pane_input.rs:6`) for every other
Expand Down
13 changes: 9 additions & 4 deletions docs/dispatch-composer-evidence.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,9 @@
# Dispatch composer evidence

Source: issue #51 (dispatch loses 1 KiB chunks) and `BACKLOG.md` line 209.
Source: issue #51 (dispatch loses 1 KiB chunks) and `BACKLOG.md`'s
["Dispatch loses whole 1 KiB chunks from the head of a long prompt"](../BACKLOG.md#dispatch-loses-whole-1-kib-chunks-from-the-head-of-a-long-prompt)
entry (line 243 as of this change; a heading link survives the line-number
drift a bare number would not).
Backs `pane_input::max_prompt_bytes` (`src-tauri/src/pane_input.rs:25`)
granting Linux Codex an 8192-byte prompt limit
(`CODEX_LINUX_MAX_BYTES`, `pane_input.rs:7`) instead of the legacy
Expand Down Expand Up @@ -34,7 +37,8 @@ unconditionally.
## Prerequisites

- Linux only: `src-tauri/tests/cli_composer.rs` is
`#![cfg(target_os = "linux")]` and will not compile elsewhere.
`#![cfg(target_os = "linux")]`, so the file compiles to an empty crate on
any other platform rather than failing to compile.
- A `codex` executable on `PATH`. This evidence was recorded against
codex-cli 0.154.0.
- An existing, already-trusted Codex checkout directory. The probe never
Expand Down Expand Up @@ -65,8 +69,9 @@ expected and received bytes as `<encoding>-<size>.expected` /
own comparison.

The probe never sends Enter (only Codex's external-editor key, `Ctrl-G`,
`\x07`) and passes `-s read-only -a never`, so no model request can be
made regardless of what is typed. Its own process group is killed on exit,
`\x07`); no model request is made because nothing the fixture does ever
submits the prompt, not because of the `-s read-only -a never` flags it
also passes. Its own process group is killed on exit,
bounded by a 60-second watchdog, and its scratch files live in a
`tempfile::tempdir()`.

Expand Down
Loading