ProgressBoard

The ProgressBoard is TeXRA's live view in VS Code for watching running tasks, inspecting tool calls and reasoning, reviewing generated files, and resuming or forking past runs.

CLI

The ProgressBoard is the VS Code extension's live view. The CLI shows the same streaming reasoning, tool calls, and diffs in its texra chat terminal UI. Past tasks are shared across surfaces. Browse them with texra history list in the terminal, or with the Tasks button of the TeXRA view in VS Code.

Background service

On macOS and Linux, a VS Code window runs its tasks in the background TeXRA service, the same one texra chat uses. A task keeps running when its window closes, and another window, or /tasks in a terminal, can follow it. The window's editor still serves those tasks: diagnostics, inline criticism, opening a PDF and the editable diff of a proposed edit. If the service cannot start, the window runs its tasks itself and says so once.

texra history

$texra history list -n 3

9f3a6c81d24e2025-11-04T14:31:08.412Zpolishcompleteddraft.tex

c4e19b07a52d2025-11-04T11:02:44.917Zresearchresumablesections/intro.tex

e2c70a94b18f2025-11-03T17:45:13.208Zcorrecterrorappendices.tex

$texra resume c4e19b07a52d

reopens the stored session in the chat TUI, on the saved agent and model

The board's tasks, from a terminal: one tab-separated row each. texra resume picks a stored task back up.

Opening the ProgressBoard

The ProgressBoard shares the TeXRA view with the New task screen. Select the TeXRA icon in the Secondary Side Bar; the Tasks button at the top left of the panel lists every task, and + starts a new task.

• Automatic: It usually opens when you execute an agent.
• Manual: Open it from the Command Palette (Ctrl+Shift+P or Cmd+Shift+P) with TeXRA: Show Tasks, or press Ctrl+Alt+P (Cmd+Option+P on macOS). From the New task screen this opens the newest task.
• Editor tab: Run TeXRA: Open Tasks in Editor, or pick Open tasks in editor from the panel's ⋯ menu, to open the ProgressBoard as a full editor tab.

entanglement-paper — texra

New task Progress

orchestratorGoal4 turns, 18 tool calls

09:02:11

Survey steady-state entanglement, derive the concurrence, and formalize the entropy bound.

I'll split this across three specialists and assemble their results.

agent — lean0:24

Agentlean(anthropic/claude-opus-5-5)

InstructionFormalize subadditivity of the von Neumann entropy.

FilesEntropy.lean(input)

% 3 references found · grounded in arXiv

@article{wolff2021steady,

title = {Steady-state entanglement in driven spin chains},

author = {Wolff, S. and Kollath, C.},

journal = {Phys. Rev. A}, year = {2021},

doi = {10.1103/PhysRevA.103.022210}

}

Every entry comes from a real lookup — no fabricated citations. ✓ 3/3 verified

(* concurrence of the steady state, N = 2 *)

rho = SteadyState[lindblad];

C = Concurrence @ rho // FullSimplify

(γ² Sin[2θ]) / (γ² + 4Δ²)

Cross-checked against exact diagonalization at N = 8. ✓ agrees to 1e-9

theorem entropy_subadditive (ρ : State H) :

S (ptrace ρ) ≤ S ρ := by

have h := strong_subadditivity ρ

linarith

⟳ checking proof state · Loogle found 1 lemma

The ProgressBoard: stream header and live log on the left, the run's output files on the right.

Layout overview

The ProgressBoard is split into two main sections (usually side by side, but configurable):

1. Stream Tabs: A list on the side (often right) showing different agent runs (streams).
2. Content Area: The main area (often left) displaying the header and log details for the selected stream.

Stream tabs section

This section lists the stored tasks for the workspace, including tasks started earlier and from other TeXRA hosts; their conversations load from the workspace's SQLite database.

• Switching streams: Select a stream name (e.g., polish: paper.tex) to view its logs and status in the Content Area.
• Removing a stream: Each tab has an button that removes that stream and its logs from the ProgressBoard view.
• Metadata: Tabs display the model and when the stream was last active on a second line. Icons indicate the agent type and whether multiple output files were generated.

Content area

This area shows the details for the stream selected in the Stream Tabs section.

Header

The header provides a summary and actions for the selected stream:

• Stream name: Displays the identifier of the current run. Tabs show the agent name (with #executionId when parallel runs would otherwise collide). Input files stay on the files panel, not on the tab chip. The model appears on the second line of the tab.
• Status indicator: A colored circle shows the current status. The four states read at a glance:

RunningAgent is actively processing

WaitingPaused for your input or a follow-up

CompletedFinished (or cancelled) — grey

FailedThe run hit an error

The status dot: green while running, blue while waiting for input, gray once finished, red on error.

• Token and cost summary: Displays the combined input and output token counts from all completed revisions (e.g., r0, r1, r2, …) along with the estimated cost.
• Header row: Shows the Tasks button, the task's title, status and elapsed time (plus Pass 2 of 3 for multi-pass document tasks), Stop while running (aborts the active request immediately for providers supporting AbortController, or stops after the current API call finishes), New task, and the More menu (⋯).
• More menu (⋯):
  • All tasks:
    • Rename: Change the task title.
    • Open task folder: Reveal the run folder in run storage (executions/<run-id>/) to inspect generated files, compile logs, mirrored dependencies, and intermediate artifacts.
    • Export: Save the transcript as Markdown, HTML, or PDF.
    • Copy diagnostics: Copy the task's title, agent, model, status, task ID, output paths, and compile failures as plain text for bug reports or follow-up tasks.
    • Delete task: Remove the task once stopped.
  • Document tasks:
    • Run again…: Pre-fill the New task form with the same agent, model, files, and instruction to start a fresh run.
    • latexdiff: Compare the original input .tex file(s) with the generated output .tex file(s) using latexdiff (read the LaTeX Diff guide).
    • Archive outputs: Move the run's output files and log into the History folder (read the file management guide).
    • Delete output files: Remove the run folder's generated files.
  • Chat tasks: Fork (branch the conversation into a new task), Compact (summarize earlier turns), and links to any forks created from the conversation.
  • Global: Open tasks in editor (or Back to sidebar) and LaTeXDiffs.
• After a task ends: The composer area is replaced by a completion bar offering Fork (starts a new task with the conversation up to its last finished turn, leaving the original untouched) and, for interrupted tasks or workflows with saved outputs, Resume (continues from saved outputs instead of starting over). Hovering over any of your messages after the first also offers Fork from here, which branches the conversation before that message and places the message in the composer.

Reviewed outputs are accepted per file: each row under Generated Files has an Accept action that copies the edited version into your workspace.

Checking an agent's launch

When an agent asks to start another agent, its card shows the agent, the model, the instruction and the files it would run. Edit (e) opens the New task form filled with that proposal. Change the files, the switches, the instruction, the agent or the model, then press Approve to start it with your changes. It still runs as the asking agent's child, and its result goes back to that agent. Back to … returns to the asking task and leaves the proposal waiting.

Auto-approval

An approval card's ▾ menu (Approve all … in this run) is where you let a task approve a kind of request on its own: edits, commands or agent work. Later requests of that kind in the task are approved without asking. Edits and commands are independent. Agent work also covers the other two, and turning it off returns all three to asking.

While a grant is on, the task's header shows an amber chip for it (Auto: edits, Auto: commands, Auto: agent work); its × turns the grant off. In the narrow sidebar one Auto chip stands for all of them, and its × turns them all off. The composer's Approval chip sets the policy a new task starts with. The CLI shows the same grants as AUTO-EDIT, AUTO-BASH and AUTO-TASK badges.

Context utilization

A small percentage next to the token count shows how full the model's context window is. When it climbs toward 100%, the conversation may get compacted automatically, or you may want to start a fresh session.

After a workflow run

To follow up on a finished document task in chat, use Copy diagnostics in the header menu to put the task's output paths and compile failures on the clipboard, then start a chat from the New view and paste that text into the instruction box.

When a run recorded a compile failure, Fix compile errors still appears under Generated Files and starts a repair chat from those logs.

Memory

Agents can remember things between sessions. When memory is enabled (toggle in the Settings Memory tab), agents save useful notes about your project. You can browse, pin, and delete these notes from the Memory tab in Settings, reached with TeXRA: Open Settings. Read the memory guide for a full walkthrough.

Log content

This scrollable area displays the detailed, timestamped logs for the selected agent run.

• Structure: Logs are organized into expandable groups (e.g., Initialization, Revision 1, Model Operation). Response cycles are logged within the corresponding revision group. Select the arrow next to a group name to expand or collapse it.
• Log levels: Messages are prefixed with levels like INFO, DEBUG, WARN, ERROR to indicate severity. Verbose debug messages (DEBUG) are only shown when texra.logger.debugMode is set to true in <project>/.texra/config.json or ~/.texra/v1/global-storage/config.json; TeXRA does not read VS Code settings.
• Agent thinking: The log highlights model reasoning in purple Thinking blocks. These sections are flagged internally with a thinking type so you can spot when the model is exploring ideas.
• Errors: Errors are highlighted and often show what went wrong.

ProgressBoardlogSearch log…

All Info Warn Error

• 14:02:11INFORun r0 started · spectral-gap.tex#a3f1
• 14:03:48OKWrote r0/output.xml — 4 files
• 14:05:02WARNApproaching context window — 182k/200k tokens
• 14:05:39ERRORlatexdiff exited 1 — see build logr0/build/diff.log:84 Undefined control sequence \DIFadd

The log: each row is color-keyed by severity (green for info/success, yellow for warnings, red for errors), with expandable nested detail and per-task IDs.

The log content is the main source for diagnosing problems and seeing how TeXRA and the models process your requests. Read the troubleshooting guide for more on using logs.