\", \"\\n\")\n@@ -58,15 +64,24 @@ function parseEventTag(tag) {\n \treturn { attributes };\n }\n \n-function findValidNestedEvent(raw, bodyStart, close) {\n-\tlet candidate = raw.indexOf(\"
wo';\n \tconst partial = drainEvents(`noise${partialEnvelope}`);\n@@ -358,12 +366,15 @@ async function testParsing() {\n \t\t},\n \t]);\n \n-\tconst truncatedA =\n-\t\t'
\", \"\\n\")\n@@ -58,15 +64,24 @@ function parseEventTag(tag) {\n \treturn { attributes };\n }\n \n-function findValidNestedEvent(raw, bodyStart, close) {\n-\tlet candidate = raw.indexOf(\"
wo';\n \tconst partial = drainEvents(`noise${partialEnvelope}`);\n@@ -358,12 +366,15 @@ async function testParsing() {\n \t\t},\n \t]);\n \n-\tconst truncatedA =\n-\t\t'
\", \"\\n\")\n\t\t.replaceAll(\"
\", \"\\n\")\n\t\t.replaceAll(\"<\", \"<\")\n\t\t.replaceAll(\">\", \">\")\n\t\t.replaceAll(\""\", '\"')\n\t\t.replaceAll(\"&\", \"&\");\n}\n\nfunction protocolError(message) {\n\tconst error = new Error(`invalid spt EVENT stream: ${message}`);\n\terror.code = \"SPT_PROTOCOL_ERROR\";\n\treturn error;\n}\n\nfunction parseEventTag(tag) {\n\tif (!tag.startsWith(\"
\", \"\\n\")\r\n\t\t.replaceAll(\"
\", \"\\n\")\r\n\t\t.replaceAll(\"<\", \"<\")\r\n\t\t.replaceAll(\">\", \">\")\r\n\t\t.replaceAll(\""\", '\"')\r\n\t\t.replaceAll(\"&\", \"&\");\r\n}\r\n\r\nfunction protocolError(message) {\r\n\tconst error = new Error(`invalid spt EVENT stream: ${message}`);\r\n\terror.code = \"SPT_PROTOCOL_ERROR\";\r\n\treturn error;\r\n}\r\n\r\nfunction parseEventTag(tag) {\r\n\tif (!tag.startsWith(\"
\", \"\\n\")\r\n\t\t.replaceAll(\"
\", \"\\n\")\r\n\t\t.replaceAll(\"<\", \"<\")\r\n\t\t.replaceAll(\">\", \">\")\r\n\t\t.replaceAll(\""\", '\"')\r\n\t\t.replaceAll(\"&\", \"&\");\r\n}\r\n\r\nfunction protocolError(message) {\r\n\tconst error = new Error(`invalid spt EVENT stream: ${message}`);\r\n\terror.code = \"SPT_PROTOCOL_ERROR\";\r\n\treturn error;\r\n}\r\n\r\nfunction parseEventTag(tag) {\r\n\tif (!tag.startsWith(\"
\n ![\"omp\"](\"https://github.com/can1357/oh-my-pi/blob/main/assets/hero.png?raw=true\")
\n
\n A coding agent with the IDE wired in.\n omp.sh\n
\n\n\n ![\"npm](\"https://img.shields.io/npm/v/@oh-my-pi/pi-coding-agent?style=flat&colorA=222222&colorB=CB3837\")
![\"Changelog\"](\"https://img.shields.io/badge/changelog-keep-E05735?style=flat&colorA=222222\"){kind=icon}
![\"CI\"](\"https://img.shields.io/github/actions/workflow/status/can1357/oh-my-pi/ci.yml?style=flat&colorA=222222&colorB=3FB950\")
![\"License\"](\"https://img.shields.io/github/license/can1357/oh-my-pi?style=flat&colorA=222222&colorB=58A6FF\")
![\"TypeScript\"](\"https://img.shields.io/badge/TypeScript-3178C6?style=flat&colorA=222222&logo=typescript&logoColor=white\"){kind=icon}
![\"Rust\"](\"https://img.shields.io/badge/Rust-DEA584?style=flat&colorA=222222&logo=rust&logoColor=white\"){kind=icon}
![\"Bun\"](\"https://img.shields.io/badge/runtime-Bun-f472b6?style=flat&colorA=222222\"){kind=icon}
![\"Discord\"](\"https://img.shields.io/badge/Discord-5865F2?style=flat&colorA=222222&logo=discord&logoColor=white\"){kind=icon}
\n Fork of Pi by @mariozechner \n
\n\nThe most capable agent surface that ships. Continuously tuned by real-world use — complete out of the box, open all the way down.\n\n**40+** providers · **32** built-in tools · **14** lsp ops · **28** dap ops · **~55k** lines of Rust core.\n\n## Install\n\n**macOS · Linux**\n\n```sh\ncurl -fsSL https://omp.sh/install | sh\n```\n\n**Homebrew**\n\n```sh\nbrew install can1357/tap/omp\n```\n\n**Bun (recommended)**\n\n```sh\nbun install -g @oh-my-pi/pi-coding-agent\n```\n\n**Windows (PowerShell)**\n\n```powershell\nirm https://omp.sh/install.ps1 | iex\n```\n\n**Pinned versions (mise)**\n\n```sh\nmise use -g github:can1357/oh-my-pi\n```\n\nmacOS · Linux · Windows · bun ≥ 1.3.14\n\n### Shell completions\n\n`omp` generates its own completion scripts for **bash**, **zsh**, and **fish** from the live command/flag metadata, so they never drift from the actual CLI. Subcommands, flags, and enum values complete statically; model names (`--model`, `--smol`, `--slow`, `--plan`) resolve against the bundled model catalog and `--resume` against your on-disk sessions.\n\n```sh\n# zsh — add to ~/.zshrc (or write the output into a file on your $fpath)\neval \"$(omp completions zsh)\"\n\n# bash — add to ~/.bashrc\neval \"$(omp completions bash)\"\n\n# fish\nomp completions fish > ~/.config/fish/completions/omp.fish\n```\n\n## Every tool, _benchmaxxed_.\n\nEdits that land on the first attempt. Reads that summarize files instead of dumping their content. Searches that return instantly. Pick any model — omp will get it right.\n\n| model | metric | what |\n| ---------------- | ------------ | --------------------------------------------------------------------- |\n| Grok Code Fast 1 | 6.7% → 68.3% | Tenfold lift the moment the edit format stops eating the model alive. |\n| Gemini 3 Flash | +5 pp | Over str_replace — beats Google's own best attempt at the format. |\n| Grok 4 Fast | −61% tokens | Output collapses once the retry loop on bad diffs disappears. |\n| MiniMax | 2.1× | Pass rate more than doubles. Same weights, same prompt. |\n\n- `read` : summarized snippets · ideal defaults · selector hit rate\n- `search` : fastest in the west\n- `lsp` : everything your IDE knows, the agent knows\n- `prompts` : adjusted relentlessly for each model\n\n[Read the full post ↗](https://blog.can.ac/2026/02/12/the-harness-problem/)\n\n## The Pi _you love_, with **batteries included**.\n\nOriginally built on [Mario Zechner](https://github.com/mariozechner)'s wonderful [Pi](https://github.com/badlogic/pi-mono), omp adds everything you're missing.\n\n### 01 · Code execution w/ tool-calling\n\nMost harnesses give the agent a Python sandbox and call it done. Ours runs persistent Python and a Bun worker, and either kernel can call back into the agent's own tools — read, search, task — over a loopback bridge. The agent loads a CSV with tool.read from inside Python, charts it from JavaScript, and never leaves the cell.\n\n![omp TUI: a single eval session with `[1/2] pandas describe` (Python) printing a real DataFrame.describe() table, followed by `[2/2] top scorer` (JavaScript) running a reduce. Footer: 'Both kernels ran in one session.'](https://omp.sh/captures/eval.webp)\n\n### 02 · LSP wired into every write\n\nAsk for a rename and you get a rename. The call goes through workspace/willRenameFiles, so re-exports, barrel files, and aliased imports update before the file moves. Everything your IDE knows, the agent knows.\n\n\n\n### 03 · Drives a real debugger\n\nA C binary segfaults: the agent attaches lldb, steps to the bad pointer, reads the frame. A Go service hangs: it attaches dlv and walks the goroutines. A Python process is wedged: debugpy, pause, inspect, evaluate. Most agents are still sprinkling print statements.\n\n\n\n_[Watch the capture ↗](https://omp.sh/clips/dap.mp4)_\n\n### 04 · Time-traveling stream rules\n\nYour rules sit dormant until the model goes off-script. A regex match aborts the stream mid-token, injects the rule as a system reminder, and retries from the same point. You get course-correction without paying context tax on every turn. Injections survive compaction, so the fix sticks.\n\n![omp TUI: agent reading src.rs and about to write Box::leak when the request aborts (red `Error: Request was aborted`), an amber `⚠ Injecting rule: box-leak` card injects the rule body `Don't reach for Box::leak in production code paths`, and the agent then course-corrects by proposing `Arc