my-pi

A collection of pi extensions for an enhanced coding experience.

Quick Start

Copy the prompt below and paste it into your pi editor, then press Enter:

Clone https://github.com/Loongphy/my-pi.git and copy all files to ~/.pi/agent/extensions/, then run /reload. If any filenames conflict, rename the incoming files before copying — I don't want to overwrite my existing extensions.

Option 2: Manual setup

Clone the repo and copy the files:

git clone https://github.com/Loongphy/my-pi.git /tmp/pi-extensions
cp -r /tmp/pi-extensions/*.ts ~/.pi/agent/extensions/
cp -r /tmp/pi-extensions/status/ ~/.pi/agent/extensions/status/

Warning

Check for filename conflicts. If you already have an extension with the same name in the ~/.pi/agent/extensions, rename the incoming files (e.g., collapse-tools.new.ts) rather than overwriting your existing ones.

Then reload pi:

/reload

Extensions

✏️ editor

Patches the DynamicBorder component to suppress redundant border lines in bash mode, keeping the editor area clean and clutter-free.

File: editor.ts

---

📝 request-logger

Logs every provider request to a file (~/.pi/agent/requests/<session>.request.log). Captures HTTP status, headers, token counts, model info, and sanitizes sensitive query parameters.

File: request-logger.ts

---

⌨️ shortcuts

Adds Ctrl+Shift+C to copy the current editor content to the system clipboard.

File: shortcuts.ts

---

📊 status

A comprehensive status bar suite with multiple modules:

Module	Description
index.ts	Main extension entry point, orchestrates all status modules
header.ts	Rich status header above the editor showing model, working directory + git branch, token statistics, context usage, generation speed, and TTFT
git.ts	Git status detection — branch name, ahead/behind counts, staged/modified/deleted/conflicted/untracked file counts
tps.ts	Token speed engine — real-time TPS estimation during streaming, accurate TPS after completion, TTFT measurement
title.ts	Animated terminal title with a braille spinner during agent activity
theme.ts	Cross-platform system dark/light mode detection and automatic pi theme switching
statusline.ts	/statusline command for interactive configuration of which items appear in the header

Files: status/index.ts, status/header.ts, status/git.ts, status/tps.ts, status/title.ts, status/theme.ts, status/statusline.ts

TPS (time window: first token → message_end, excluding TTFT):

• During streaming — estimated, via max(1, round(chars/4)), the same chars/4 heuristic pi's compaction module uses internally.
• After message_end — accurate, TPS = usage.output / (message_end - first_token_time).

TTFT (time to first token): TTFT = first_token_time - http_request_sent_time

---

429 Rate-Limit Workaround (OpenCode)

Temporary workaround for SDK 429 retry hang, integrated into request-logger.ts.

Related issues:

• pi#3671 — Copilot provider hangs on long Retry-Afters
• pi#4666 — 429 Retry-After waits ignore maxRetryDelayMs; Esc and /new do not recover

The problem (historical): When a provider returns HTTP 429 with a large retry-after header, the underlying SDK (OpenAI, Anthropic) used to sleep for that exact duration with no upper bound and no abort support — Esc could not cancel it, /new broke the session, and the only recovery was restarting pi. This has since been fixed upstream (pi now calls the SDKs with maxRetries: 0 plus capped, abort-aware delays), so a raw 429 no longer hangs.

Why it is still kept: formatting. The reset time lives in the retry-after header, which a raw 429 error message drops. Rewriting to a 400 (non-retryable, so the SDK throws immediately) with a human-readable body (Usage limit reached (FreeUsageLimitError): Resets in 2h 5m) makes the agent stop with a clear error that includes the reset time.

What it does: Intercepts fetch() on 429 responses from opencode.ai providers only. Parses the retry-after header, returns a 400 with a human-readable message (Usage limit reached (FreeUsageLimitError): Resets in 2h 5m), and avoids pi's retry-trigger keywords so the error displays once without entering a retry loop. The provider's error-type marker (e.g. FreeUsageLimitError / GoUsageLimitError) is preserved in the body so the SDK classifies it as a non-retryable limit and stops the agent with the reset time, rather than retrying.

Companion 429-retry.ts: auto-retries transient 429s (short retry-after), but fails fast on hard usage limits — detected via the error-type marker or a reset window beyond 10 minutes — so a multi-hour free-tier limit surfaces an error immediately instead of burning ~100 min on pointless retries.

Remove this workaround once the upstream issue is fixed.