aidev TIER 1 harness can1357

Oh My Pi

RELEASES30
BREAKING8
LATESTv17.4.0
SHIPPED6d
TRACKED SINCE2026-07-26
FULL TIMELINE → OFFICIAL SITE ↗DOCUMENTATION ↗REPOSITORY ↗RELEASES ↗ § MARKDOWN ⌁ FEED

RELEASE CADENCE · LAST 12 MONTHS

09
10
11
12
01
02
03
04
05
06
07
08
The numbers behind the bars
Releases per month, last twelve months
MonthReleasesShare
2025-09 0 0%
2025-10 0 0%
2025-11 0 0%
2025-12 0 0%
2026-01 0 0%
2026-02 0 0%
2026-03 0 0%
2026-04 0 0%
2026-05 0 0%
2026-06 0 0%
2026-07 8 27%
2026-08 22 73%
12-MONTH TOTAL 30—

BREAKING CHANGES

v17.4.0 Oh My Pi 17.4.0: model-scoped Tokenizer replaces global token counting (breaking) 6d v17.3.0 Oh My Pi v17.3.0: breaking config changes, new OpenAI models, per-agent advisors, Nix support 13d v17.2.12 Oh My Pi v17.2.12: memory leak fix, hashline breaking change, builtin consolidation 17d

RECENT

Density
6d
06:37
aidev oh-my-pi BREAKING

Oh My Pi 17.4.0: model-scoped Tokenizer replaces global token counting (breaking)

Oh My Pi 17.4.0 removes the global token-counting functions in @oh-my-pi/pi-agent-core and replaces them with immutable per-model Tokenizer instances, which several context-management functions now require explicitly. On the agent side it adds /cleanse, omp ps, speculative compaction, and an extendedContext setting controlling whether premium long-context windows are used.

breakingfeaturefixapi-changecontext-windowpricingmodel-support
6d
11:11
oh-my-pi Oh My Pi v17.3.8: compaction hardening, permission query handling, Qwen thinking effort fix v17.3.8 · SIG 2/5
7d
08:47
aidev oh-my-pi

Oh My Pi v17.3.7: grok-4.6 default, 1M context window on Codex, extension file fallback API

Default xAI model bumps to grok-4.6 for both API key and SuperGrok OAuth transports. GPT-5.6 context window on the Codex subscription transport is raised to 1M tokens (working around a stale Codex registry value). A new ExtensionAPI hook lets sandboxed hosts delegate file writes and deletes to a privileged channel. An omp/ User-Agent is now sent on xAI chat requests, and the stats dashboard binds to an explicit host address.

featurefixcontext-windowapi-change
9d
08:00
aidev oh-my-pi

Oh My Pi 17.3.5: xAI onboarded to Responses API, GLM-5.3 support, Grok-4.5 default, broad provider reliability overhaul

A substantial cross-package release. The paid xAI provider (XAI_API_KEY) moves from Chat Completions to the OpenAI Responses API, aligning with SuperGrok. Default model moves to grok-4.5 on both xai and xai-oauth. GLM-5.3 is now available with full reasoning-effort ladder and 1M context on z.AI. Dozens of fixes improve retry handling across Anthropic, DeepSeek, Kimi Code, Alibaba, and xAI providers, plus fixes to usage reporting, session recovery, replay of thinking content, and more.

featurefixmodel-supportapi-change
11d
13:14
aidev oh-my-pi

Oh My Pi v17.3.4: Codex V2 compaction, credential routing fixes, PDF backend swap, memory recall fix

Major patch release: fixes Codex V2 remote compaction, OAuth usage cache invalidation, GitHub Copilot credential routing (personal vs business endpoint), and OpenRouter DeepSeek effort clamping. Replaces MuPDF-WASM PDF backend with pdf-inspector, restores PDF page rendering via headless Chromium. Fixes Streamable HTTP MCP session invalidation, Ctrl+G on Windows, memory recall dropping global rows, DeviceCheck crashing on headless macOS, and terminal DA1 reply leaking into composer.

fixfeatureapi-change
12d
04:06
oh-my-pi Oh My Pi v17.3.3: Gemini thought-only recovery, malformed Markdown, memory channel fix v17.3.3 · SIG 2/5
12d
22:53
aidev oh-my-pi

Oh My Pi 17.3.2: Antigravity identity injection removed, DeepSeek V4 Pro preview and Gemini 3.7 Flash support

Second patch in the 17.3 line. Antigravity no longer injects a fake identity system prompt into Cloud Code Assist requests — user system prompts now pass through unmodified (still tagged as user role). New model support for DeepSeek V4 Pro preview and Gemini 3.7 Flash. Antigravity client-version discovery is now dynamic from the update manifest, so version-gated models appear without a code change. Several coding-agent fixes, including a TUI stall after subagent results and a read hashline path collision bug.

fixfeaturemodel-support
12d
19:12
oh-my-pi Oh My Pi 17.3.1: Antigravity Gemini discovery, Windows CRT fix, extension tool allowlisting v17.3.1 · SIG 2/5
13d
07:02
aidev oh-my-pi BREAKING

Oh My Pi v17.3.0: breaking config changes, new OpenAI models, per-agent advisors, Nix support

Breaking: the global advisor.subagents setting is removed (per-agent advisor config replaces it), withGeminiThinkingLoopGuard is renamed to withThinkingLoopGuard, and OpenAICompat.enableGeminiThinkingLoopGuard is removed. New: first-party OpenAI Daybreak Blue, Daybreak Red, and GPT-5.6 Cyber models; per-agent advisors via frontmatter or settings; redesigned /agents fullscreen hub; Nix support with reproducible builds, NixOS/Home Manager modules; Astral ty as a Python LSP fallback. Multiple fixes across Anthropic cache pricing, Ollama chat adapter, xAI usage reporting, and Grok thinking-loop guard. Legacy cost-estimation machinery for OpenCode Go is removed.

breakingfeaturefixmodel-supportcapability
14d
02:01
aidev oh-my-pi

Oh My Pi v17.2.15: omp compress, external thinking, tool discovery expansion

Several new features and significant fixes across the Oh My Pi monorepo. omp compress provides a two-tool agent loop for condensing text files. The --external-thinking flag and think tool are added for private scratchpad reasoning. Tool discovery in omp cleanse expands to Go, Python, JS/TS linters, and GitHub Actions. MCP Streamable HTTP transport hardens to protocol 2025-11-25 with fixes for connection drops, SSE resumption, and double-execution during auth refreshes. /handoff now copies local artifacts across the session boundary.

featurefixcapabilityapi-change
14d
19:29
oh-my-pi Oh My Pi v17.2.14: forceReasoningOff for OpenAI/Azure, externalThinking via think tool v17.2.14 · SIG 2/5
14d
14:41
aidev oh-my-pi

Oh My Pi v17.2.13: Bedrock role_arn credential resolution, Cursor double-execute fix, GLM-5.2 support

Significant fixes across transports and providers: Bedrock role chaining via STS is now supported for multi-account and EKS/IRSA setups; Nitro-based EC2 instances are also detected. Cursor double-execution of settled tools is fixed in both the dialect projector and inband tool stream paths. DeepSeek Responses thinking-mode continuation after compaction is repaired. Catalog gains image-capable metadata for Muse Spark 1.2 and GLM-5.2-specific thinking tiers. New features include Exa MCP search, Cloudflare AI Gateway, and an improved omp update migration path.

featurefixcapabilitymodel-supportapi-change
ALL 31 EVENTS →