@gavin-hu/my-pi-agent 0.14.0 → 0.15.1

package/extensions/bytefly/update/suppress.ts

@@ -0,0 +1,66 @@
+/**
+ * Pi suppresses its own package-update block while bytefly owns the surface.
+ *
+ * Pi draws `Package Updates Available` from `InteractiveMode.run()` after the
+ * session starts, with no setting or extension event to turn it off. The one
+ * public seam is the package manager it calls: replacing
+ * `checkForAvailableUpdates` with an empty result suppresses that block, while
+ * the captured original still serves bytefly's own query. The patch is
+ * process-wide, so it is installed at `session_start` and restored at
+ * `session_shutdown`.
+ *
+ * A future host that changes the method shape makes the patch a no-op rather
+ * than an error; the update surface then falls back to Pi's block.
+ */
+
+import { DefaultPackageManager } from "@earendil-works/pi-coding-agent";
+import type { PackageUpdate, PackageUpdateCheck } from "./types.ts";
+
+/** A patchable class prototype, kept narrow so a test needs no host object. */
+export interface PackageManagerPrototype {
+	checkForAvailableUpdates: PackageUpdateCheck;
+}
+
+/** The installed suppression: the real check, and how to restore Pi's method. */
+export interface UpdateSuppression {
+	/** The original method, bound at call time so it still sees its manager. */
+	readonly check: PackageUpdateCheck;
+	/** Put Pi's method back; safe to call more than once. */
+	restore(): void;
+}
+
+/** Marker for the installed patch, so a second registration reuses it. */
+const SUPPRESSION = Symbol.for("@gavin-hu/my-pi-agent/pi-update-suppression");
+
+interface MarkedPrototype extends PackageManagerPrototype {
+	[SUPPRESSION]?: UpdateSuppression;
+}
+
+/**
+ * Replace Pi's package-update check with one that reports nothing.
+ *
+ * `prototype` is a seam for tests; production passes Pi's
+ * `DefaultPackageManager.prototype`. When the method is missing, the returned
+ * suppression is inert and the original read stays `[]`.
+ */
+export function suppressPiPackageUpdates(
+	prototype: PackageManagerPrototype = DefaultPackageManager.prototype as PackageManagerPrototype,
+): UpdateSuppression {
+	const marked = prototype as MarkedPrototype;
+	const existing = marked[SUPPRESSION];
+	if (existing) return existing;
+
+	const original = prototype.checkForAvailableUpdates;
+	if (typeof original !== "function") return { check: async () => [], restore: () => {} };
+	const originalCheck = original as unknown as (this: DefaultPackageManager) => Promise<PackageUpdate[]>;
+	const check: PackageUpdateCheck = (manager) => originalCheck.call(manager);
+	const patched: PackageUpdateCheck = async () => [];
+	const restore = (): void => {
+		if (prototype.checkForAvailableUpdates === patched) prototype.checkForAvailableUpdates = original;
+		delete marked[SUPPRESSION];
+	};
+	const suppression: UpdateSuppression = { check, restore };
+	prototype.checkForAvailableUpdates = patched;
+	marked[SUPPRESSION] = suppression;
+	return suppression;
+}

package/extensions/bytefly/update/types.ts

@@ -0,0 +1,15 @@
+/**
+ * The package-update shape Pi's manager returns.
+ *
+ * Pi exports the manager but not its `PackageUpdate` interface, so the type is
+ * derived from the public method. That keeps the extension on the same contract
+ * without reaching into a dist path or restating the fields.
+ */
+
+import type { DefaultPackageManager } from "@earendil-works/pi-coding-agent";
+
+/** One package Pi reports as updateable. */
+export type PackageUpdate = Awaited<ReturnType<DefaultPackageManager["checkForAvailableUpdates"]>>[number];
+
+/** Pi's check, captured before suppression and called with a manager. */
+export type PackageUpdateCheck = (manager: DefaultPackageManager) => Promise<PackageUpdate[]>;

package/extensions/check/README.md

@@ -0,0 +1,253 @@
+# check — cross-stack verification runs
+
+Run the project's test, lint, or typecheck command from Pi, watch it live, and
+turn its output into a navigable `file:line` list. The same run is available to
+the model as `run_check`, with structured failures instead of a wall of log
+text.
+
+`check` is deliberately generic: it detects the project's command from its
+manifest or takes `check.command`, and it does not assume Node, Bun, a test
+framework, or a particular CI system.
+
+```
+pi --extension ./extensions/check   # load just this extension
+pi -e .                             # load the whole @gavin-hu/my-pi-agent package
+pi install ./                       # install the package
+```
+
+## Screens
+
+<img src="../../assets/check.png" alt="The /check screen: a failed run with failure rows and the focused failure's message" width="100%">
+
+<img src="../../assets/check-detail.png" alt="The /check failure detail: the location and message over a source excerpt with the failing line marked by >" width="100%">
+
+## Tool rendering
+
+The call line is `run_check <target>`. While the command runs, the result is one
+line — `Running bun run check… · 12s` — and it repaints once a second, so a
+silent compiler still shows progress. The settled block keeps the transcript to
+a capped view of the run: a status summary, at most five failure rows, and the
+bounded tail when nothing parsed. `ctrl+o` expands it to the model's 20-failure
+cap; `/check` remains the full navigable list.
+
+<img src="../../assets/tools/run_check.png" alt="The run_check transcript block: a status summary and two failure rows under a rail" width="100%">
+
+A failed command whose output parsed no location names the exit status in the
+summary, prints `no parsed failure; last output:`, and shows the last five tail
+lines instead of an exit-status row. The transcript never prints the model
+report; the full failure list and the tail stay in `details`.
+
+## What it does
+
+- **Detects a command** from `package.json`, `Cargo.toml`, `go.mod`, a pytest
+  configuration, a `justfile`, or a `Makefile`, or uses `check.command`.
+- **Runs it in a supervised process** with a wall-clock timeout and a bounded
+  log; `Esc` cancels the process tree.
+- **Shows a live tail** while the command runs, then a failure list with a
+  detail view for the focused failure: a source excerpt around `path:line` when
+  the file is readable, otherwise the bounded tail.
+- **Returns structured failures** to the model with `path`, `line`, `column`,
+  `test`, `message`, and `command`.
+- **Falls back honestly**: a failed command with no recognizable failure becomes
+  one “Command exited with status N” entry, with the bounded tail in its detail.
+- **One line on success**: `Check passed: bun run check · 12.4s`.
+
+## Tool
+
+`run_check` takes optional `target`:
+
+| Field | Value |
+|---|---|
+| `target` | `test`, `lint`, `typecheck`, or `all`; defaults to `all`. |
+| `exposure` | `direct`, active by default. |
+| `executionMode` | `sequential` — one run owns the process and the live buffer. |
+| `annotations` | `readOnlyHint: false`, `destructiveHint: false`, `openWorldHint: false`. |
+| `outputSchema` | `{ target, status, commands, exitCode, durationMs, truncated, failures[] }`, mirrored by `structuredContent`. |
+
+`content` is a short summary plus up to 20 failures. `details` carries the full
+bounded failure list and the tail, which the transcript block reads. A failing
+check is a normal result (`isError: false`) so the model can fix it; a check
+that cannot run — no detected command, a spawn failure, or another run already
+active — is an error with the corrective action.
+
+## Command
+
+| Command | What it does |
+|---|---|
+| `/check` | Run every detected command (or `check.command`). |
+| `/check test` | Run the detected test command. |
+| `/check lint` | Run the detected lint command. |
+| `/check typecheck` | Run the detected typecheck command. |
+| `/check all` | Same as `/check`. |
+
+`getArgumentCompletions` offers the four targets. The command refuses while plan
+mode is on: relying on tests is still writing to the project (caches, snapshots,
+generated files), so the plan must be approved first.
+
+## Behaviour by mode
+
+| Mode | Behaviour |
+|---|---|
+| Interactive TUI | A live run screen with a scrolling tail and `Esc cancel`; then a failure list with `Enter` detail, `r` rerun, and `Esc` close. A pass posts one notice. |
+| RPC (`hasUI`) | Runs headless and posts the summary plus the failure list; no custom screen. |
+| `print` / `json` | Runs headless and posts the same report. |
+| `run_check` | No UI of its own. Returns the summary in `content`, the structured result, and a capped transcript block drawn from `details`. |
+
+## Configuration
+
+The `check` section of the shared `bytefly.json` (global then project, project
+wins). All keys are JSON-only; none has a `/bytefly settings` row.
+
+| Key | Default | Bounds | Meaning |
+|---|---|---|---|
+| `command` | `""` | | When non-empty, run this shell command for every target instead of detection. |
+| `timeoutMs` | `600000` | `0`–`3600000` | Wall-clock limit per command; `0` disables it. |
+| `maxLogBytes` | `1048576` | `65536`–`16777216` | Bytes of tail retained per command. |
+
+`command` is the escape hatch for any stack without a detector. It is a shell
+command and runs with the project root as its working directory.
+
+## Detection
+
+Detection order stops at the first family that declares a command:
+
+| Manifest | `test` | `lint` | `typecheck` | `all` |
+|---|---|---|---|---|
+| `package.json` | `scripts.test` | `scripts.lint` | `scripts.typecheck` | `scripts.check`, else the detected targets |
+| `Cargo.toml` | `cargo test` | `cargo clippy -- -D warnings` | `cargo check` | the detected targets |
+| `go.mod` | `go test ./...` | `go vet ./...` | `go build ./...` | the detected targets |
+| pytest config or `tests/` | `pytest` | `ruff check .` with `[tool.ruff]` | `mypy .` with `[tool.mypy]` | the detected targets |
+| `justfile` | `just test` if declared | `just lint` if declared | `just typecheck` if declared | `just check` if declared, else the detected targets |
+| `Makefile` | `make test` if declared | `make lint` if declared | — | `make check` if declared, else the detected targets |
+| any other stack | — | — | — | `check.command` |
+
+Node package managers come from the lockfile (`bun`, `pnpm`, `yarn`, `npm`;
+default `npm`). Scripts run as `<pm> run <name>`; Make and Just recipes only by
+their declared names. Nothing is guessed beyond those manifests.
+
+A detector is one function returning `DetectedCommands`; adding a family is one
+module entry, not a change to the runtime, screens, or tool.
+
+## Failure parsing
+
+Known runner lines are recognized first; the generic compiler-style location is
+the fallback for everything else:
+
+| Output shape | Parser |
+|---|---|
+| `path:line(:col): error …` | generic, including a Windows `C:\…` path |
+| `FAIL path` and `● test` | Vitest/Jest |
+| `● test` and `at … (path:line:col)` | Jest/Vitest, Java, and bare stack frames |
+| `File "path", line N, in test` | Python tracebacks |
+| `FAILED path::test - message` | pytest |
+| `--- FAIL: Test` and `path:line:` | Go |
+| `panicked at path:line` / `thread 'test'` | Cargo |
+| anything else on a non-zero exit | one fallback entry with the bounded tail |
+
+Failures are deduped and capped at 200; the model report shows 20. A miss is
+never turned into an invented location.
+
+## Security
+
+`check` executes project scripts. That is the same trust boundary as running the
+command yourself or letting the agent call `bash`: a project with a hostile
+`test` script can do anything your user can do. The extension does not add a
+sandbox. `/check` is user-invoked and refuses in plan mode; `run_check` declares
+`readOnlyHint: false`, so a permission policy can gate it, and plan mode blocks
+it.
+
+The extension adds no network calls of its own. The command's environment is
+made non-interactive (`CI=1`, `NO_COLOR=1`, `FORCE_COLOR=0`, `TERM=dumb`,
+`GIT_PAGER=cat`, `PAGER=cat`).
+
+## Limitations
+
+- Detection covers the families above. Other stacks need `check.command`.
+- Parsing is best-effort. A custom reporter that prints neither a location nor a
+  known shape falls back to the tail, not to a guessed file.
+- The failure list and detail are TUI-only; other modes print the report.
+- A source excerpt is read only from a regular file inside the project root, at
+  most 512 KiB, and only when the reported line exists. A missing, binary,
+  oversized, or external file falls back to the bounded tail with the reason.
+- There is no run history, retry policy, or watch mode. Each invocation is one
+  run.
+- A test runner that opens a full-screen UI is not supported: stdout is a pipe,
+  stdin is closed, and `TERM=dumb`.
+- The tail is bounded; earlier lines are dropped and marked truncated.
+
+## Non-goals
+
+- Auto-fixing or editing code in response to a failure.
+- A remote CI client, a scheduler, or a durable run queue.
+- Watching files and rerunning on change.
+- Parsing every reporter in existence; the generic location and the bounded tail
+  are the contract.
+- Replacing `job` for long-lived processes.
+
+## Pi integration
+
+| Aspect | Contract |
+|---|---|
+| Registration | `check(pi)` guards on `isExtensionEnabled("check")` and calls `registerCheck(pi)`, which registers `run_check`, `/check`, and one `session_shutdown` hook. The factory only registers. |
+| Tool | `run_check`, `exposure: "direct"`, `executionMode: "sequential"`, annotations `{ readOnlyHint: false, destructiveHint: false, openWorldHint: false }`, TypeBox `parameters` and `outputSchema`, `structuredContent` mirror. |
+| Command | `/check [target]` with `getArgumentCompletions` for the four targets; reports through `ctx.ui.notify`. |
+| State | None persisted. The runtime closure owns the single active child, its controller, and the live line buffer; `session_shutdown` aborts it. |
+| Lifecycle | Each run spawns one process per command and kills the tree on timeout, cancel, or shutdown. No `session_start` work. |
+| Config | `check` section of `bytefly.json`, read once per run. |
+
+## Design notes
+
+- **The excerpt is guarded like a path-taking tool.** The path comes from command
+  output, so `source.ts` resolves it, requires both string and real-path
+  containment under the project root, rejects non-files, binaries, and files over
+  512 KiB, and checks the line before reading. A miss shows the tail, never a
+  guessed source line.
+- **Why `check`.** It runs the project's *check* command, matching `bun run check`
+  and the scripts it detects. `/verify` describes a verdict rather than the
+  action, and `run_verify` reads worse than `run_check`; the extension,
+  command, tool, and config section all use one word.
+- **Generic by design.** Only `detect.ts` knows manifests, and only `failures.ts`
+  knows runner output. The runtime, screens, and tool speak in `CommandSpec` and
+  `Failure`.
+- **One-shot, not a job.** `job` owns reattachment, stdin, detached lifetime, and
+  a registry; `check` owns a foreground verdict. Reusing `job` would make both
+  surfaces worse, so the runner is local and smaller.
+- **A bounded tail is the honest fallback.** A parser miss must not invent a
+  location; the exit status and the tail are useful on their own.
+- **The transcript is a capped view.** `render.ts` draws from the same `details`
+  the model report reads but caps the rows and never prints the report; `/check`
+  is still the full navigable list.
+- **The failure list is a `DockListScreen`.** The frame, the id-anchored cursor,
+  the navigation keys, the range row, and the detail pane come from
+  `lib/tui/dock-screen.ts`, the base `/jobs`, `/watch`, and `/agents` share, so
+  this module holds only the check vocabulary. A failure has no id, so its index
+  in the finished run's fixed array is the cursor key.
+- **One run at a time.** The runtime refuses a second start instead of
+  interleaving outputs or processes.
+
+## Files
+
+| File | Responsibility |
+|---|---|
+| `index.ts` | Registrar: guard, runtime, tool, command, shutdown. |
+| `types.ts` | `CheckTarget`, `CommandSpec`, `Failure`, `CheckResult`, `RunCheckDetails`, `ActiveCheck`. |
+| `config.ts` | `check` config defaults, clamps, and loader. |
+| `detect.ts` | The detector registry and `check.command` planning. |
+| `runner.ts` | One-shot streaming process, bounded log, timeout, cancel, tree kill. |
+| `failures.ts` | Known-runner and generic failure parsers, stack frames, and the tail fallback. |
+| `source.ts` | Guarded source excerpt around a failure's `path:line`. |
+| `runtime.ts` | Planning, sequencing, aggregation, live lines, cancel/shutdown. |
+| `format.ts` | Summary, failure rows, source-excerpt/detail text, model report. |
+| `tui.ts` | `CheckRunComponent` (a live tail) and `CheckResultComponent` (a `DockListScreen` over the failures). |
+| `tools.ts` | `run_check` schema, registration, the running partial result, and the call/result renderers. |
+| `render.ts` | The transcript result block: status summary, capped failure rows, tail fallback. |
+| `commands.ts` | `/check`, target parsing, completions, TUI/headless flow. |
+
+## Testing
+
+Co-located tests cover config clamps, every detector family, the runner's
+streaming/timeout/cancel/log cap, the parsers, command planning, aggregation,
+the tool's output contract and transcript block, the command's guards and
+headless path, and both screens' keys and width bounds.
+`extensions/check/check.test.ts` runs a real Node script through the extension,
+so the process path is exercised without depending on Bun or a test framework.

package/extensions/check/commands.ts

@@ -0,0 +1,128 @@
+/**
+ * `/check` — run the project's verification command with a live tail and a
+ * navigable failure list.
+ *
+ * The TUI gets the two screens; RPC, print, and JSON run headless and report
+ * the summary plus the failure list. The command refuses while plan mode is on
+ * because running tests is a write to the project (caches, logs, generated
+ * files), and the tool is separately blocked by its annotation.
+ */
+
+import type { ExtensionAPI, ExtensionCommandContext } from "@earendil-works/pi-coding-agent";
+import { resolveEffectiveCwd } from "../../lib/env.ts";
+import { isPlanModeEnabled } from "../../lib/plan-state.ts";
+import { withRailsSuppressed } from "../../lib/rails.ts";
+import { formatCheckReport, formatCheckSummary } from "./format.ts";
+import { CheckError, type CheckRuntime } from "./runtime.ts";
+import { CheckResultComponent, CheckRunComponent } from "./tui.ts";
+import type { CheckResult, CheckTarget } from "./types.ts";
+
+/** The accepted targets and their completion descriptions. */
+const TARGETS: ReadonlyArray<{ value: CheckTarget; description: string }> = [
+	{ value: "test", description: "Run the test command" },
+	{ value: "lint", description: "Run the lint command" },
+	{ value: "typecheck", description: "Run the typecheck command" },
+	{ value: "all", description: "Run every detected command (default)" },
+];
+
+/** Parse one target, or report usage and return undefined. */
+function parseTarget(raw: string, ctx: ExtensionCommandContext): CheckTarget | undefined {
+	const token = raw.trim().split(/\s+/)[0]?.toLowerCase() ?? "";
+	if (token === "") return "all";
+	if (TARGETS.some((target) => target.value === token)) return token as CheckTarget;
+	ctx.ui.notify(`Unknown check target "${token}". Use test, lint, typecheck, or all.`, "warning");
+	return undefined;
+}
+
+/** The one-line command summary shown while the process runs. */
+function runLabel(commands: readonly string[]): string {
+	return commands.length > 0 ? commands.join(" + ") : "check";
+}
+
+/** The TUI loop: run, then list/detail, with rerun. */
+async function runTui(runtime: CheckRuntime, ctx: ExtensionCommandContext, target: CheckTarget): Promise<void> {
+	while (true) {
+		let active: ReturnType<CheckRuntime["start"]>;
+		try {
+			active = runtime.start(target, ctx);
+		} catch (error) {
+			ctx.ui.notify(error instanceof CheckError ? error.message : String(error), "warning");
+			return;
+		}
+
+		const result = await ctx.ui.custom<CheckResult | undefined>((tui, theme, _keybindings, done) => {
+			const component = new CheckRunComponent({
+				command: runLabel(active.commands),
+				lines: active.lines,
+				elapsedMs: active.elapsedMs,
+				theme,
+				onCancel: () => active.cancel(),
+				requestRender: () => tui.requestRender(),
+				viewportRows: () => tui.terminal?.rows,
+			});
+			active.promise.then(done);
+			return component;
+		});
+		if (!result) return;
+
+		if (result.status === "passed") {
+			ctx.ui.notify(formatCheckSummary(result), "info");
+			return;
+		}
+		if (result.status === "cancelled") {
+			ctx.ui.notify("Check cancelled.", "info");
+			return;
+		}
+		if (result.status === "error") {
+			ctx.ui.notify(formatCheckReport(result), "error");
+			return;
+		}
+
+		const choice = await ctx.ui.custom<"rerun" | undefined>((tui, theme, _keybindings, done) => {
+			const component = new CheckResultComponent({
+				result,
+				cwd: resolveEffectiveCwd(ctx.cwd),
+				theme,
+				onRerun: () => done("rerun"),
+				onClose: () => done(undefined),
+				requestRender: () => tui.requestRender(),
+				viewportRows: () => tui.terminal?.rows,
+			});
+			return component;
+		});
+		if (choice !== "rerun") return;
+	}
+}
+
+export function registerCheckCommand(pi: ExtensionAPI, runtime: CheckRuntime): void {
+	pi.registerCommand("check", {
+		description: "Run the project's test, lint, or typecheck command and show failures",
+		getArgumentCompletions: (prefix) => {
+			const trimmed = prefix.trim();
+			if (trimmed.includes(" ")) return null;
+			const items = TARGETS.filter((target) => target.value.startsWith(trimmed));
+			return items.length > 0
+				? items.map((target) => ({ label: target.value, value: `${target.value} `, description: target.description }))
+				: null;
+		},
+		handler: async (args, ctx) => {
+			if (isPlanModeEnabled(ctx.sessionManager.getBranch())) {
+				ctx.ui.notify("Plan mode is read-only. Approve the plan before running checks.", "warning");
+				return;
+			}
+			const target = parseTarget(args, ctx);
+			if (!target) return;
+
+			if (ctx.mode !== "tui") {
+				const result = await runtime.run(target, ctx);
+				ctx.ui.notify(
+					result.status === "passed" ? formatCheckSummary(result) : formatCheckReport(result),
+					result.status === "error" ? "error" : "info",
+				);
+				return;
+			}
+
+			await withRailsSuppressed(pi, () => runTui(runtime, ctx, target));
+		},
+	});
+}

package/extensions/check/config.ts

@@ -0,0 +1,39 @@
+/**
+ * Configuration for the check extension.
+ *
+ * The `check` section of the shared `bytefly.json` (global then project, project
+ * wins). Missing or malformed documents are ignored, and every value is clamped
+ * to a safe range.
+ */
+
+import { clampInteger, cleanString, loadConfigSection } from "../../lib/config.ts";
+
+export interface CheckConfig {
+	/** Shell command used instead of detection for every target. Empty = detect. */
+	command: string;
+	/** Wall-clock limit for one command; 0 disables it. */
+	timeoutMs: number;
+	/** Bytes of output retained per command. */
+	maxLogBytes: number;
+}
+
+export const DEFAULT_CONFIG: CheckConfig = {
+	command: "",
+	timeoutMs: 600_000,
+	maxLogBytes: 1_048_576,
+};
+
+/** Normalize one merged config object over the running base. */
+export function normalizeConfig(raw: Record<string, unknown> | undefined, base: CheckConfig): CheckConfig {
+	if (!raw) return base;
+	return {
+		command: cleanString(raw.command, base.command),
+		timeoutMs: clampInteger(raw.timeoutMs, base.timeoutMs, 0, 3_600_000),
+		maxLogBytes: clampInteger(raw.maxLogBytes, base.maxLogBytes, 65_536, 16_777_216),
+	};
+}
+
+/** Load the effective config for a working directory. */
+export function loadConfig(cwd: string): CheckConfig {
+	return loadConfigSection(cwd, "check", DEFAULT_CONFIG, normalizeConfig);
+}