pi-ocr 1.0.0 → 1.0.1

package/README.md

@@ -7,7 +7,7 @@
 
 Multi-backend OCR for [Pi Coding Agent](https://pi.dev) — extract text, LaTeX math formulas, and tables from images and PDFs. Choose the backend that fits your needs: free cloud API, local GPU, or pure Python.
 
-> Bridges the multimodal gap for non-vision LLMs like **DeepSeek**. When your model can't see images, `minimodel_ocr` acts as its eyes.
+> Bridges the multimodal gap for non-vision LLMs like **DeepSeek**. When your model can't see images, `pi_ocr` acts as its eyes.
 
 ## Three Backends — One Tool
 
@@ -158,6 +158,9 @@ Free tier limits:
 - ≤ 20 pages per request
 - IP-based rate limiting
 
+> 💡 PDFs >20 pages: auto-splitting needs `python3` + `pypdfium2` (`pip install pypdfium2`).
+> Most PDFs are under 20 pages — you'll likely never need this.
+
 For files >10MB, compress first at [ilovepdf.com/compress_pdf](https://ilovepdf.com/compress_pdf).
 
 ---
@@ -210,7 +213,7 @@ Opens an interactive `SettingsList` with keyboard navigation:
 
 ### LLM-invoked (automatic)
 
-The extension registers a `minimodel_ocr` tool. The agent invokes it automatically:
+The extension registers a `pi_ocr` tool. The agent invokes it automatically:
 
 ```
 > What formula is written in this screenshot?
@@ -371,7 +374,7 @@ sudo pacman -S poppler
 
 ```
 ┌──────────────────┐     ┌──────────────────┐     ┌──────────────────────┐
-│  pi (DeepSeek)   │────▶│  minimodel_ocr   │────▶│  Ollama / MinerU    │
+│  pi (DeepSeek)   │────▶│  pi_ocr   │────▶│  Ollama / MinerU    │
 │  (no vision)     │     │  pi extension    │     │  / Pix2Text        │
 └──────────────────┘     └──────────────────┘     └──────────────────────┘
         │                         │                           │

package/extensions/index.ts

@@ -1,7 +1,7 @@
 /**
- * pi-minimodel-ocr — Multi-backend OCR for Pi Coding Agent
+ * pi-ocr — Multi-backend OCR for Pi Coding Agent
  *
- * Registers a `minimodel_ocr` tool that the LLM can call to read images and PDFs
+ * Registers a `pi_ocr` tool that the LLM can call to read images and PDFs
  * using one of three backends:
  *   - Ollama (local vision models like glm-ocr)
  *   - MinerU API (free Agent API, ≤10MB, ≤20 pages)
@@ -19,7 +19,7 @@
  *   Pix2Text:  pip install pix2text
  *   PDF tools:  brew install poppler (macOS multi-page PDF for Ollama)
  *
- * Install: pi install npm:pi-minimodel-ocr
+ * Install: pi install npm:pi-ocr
  */
 
 import { Type } from "@earendil-works/pi-ai";
@@ -111,7 +111,7 @@ const ocrSchema = Type.Object({
 });
 
 const ocrTool = defineTool({
-  name: "minimodel_ocr",
+  name: "pi_ocr",
   label: "Minimodel OCR",
   description:
     "Extract text, math formulas (LaTeX), and tables from images or PDFs using local Ollama vision models. " +
@@ -120,9 +120,9 @@ const ocrTool = defineTool({
   promptSnippet:
     "Extract text/formulas/tables from images and PDFs using local Ollama OCR",
   promptGuidelines: [
-    "When the user asks about the content of an image or PDF, use minimodel_ocr to extract the text first.",
-    "For mathematical documents, use minimodel_ocr with task='formula' or task='auto' to get LaTeX output.",
-    "Use minimodel_ocr with task='auto' for general document OCR to extract all text, formulas, tables, and figures.",
+    "When the user asks about the content of an image or PDF, use pi_ocr to extract the text first.",
+    "For mathematical documents, use pi_ocr with task='formula' or task='auto' to get LaTeX output.",
+    "Use pi_ocr with task='auto' for general document OCR to extract all text, formulas, tables, and figures.",
   ],
   parameters: ocrSchema,
   async execute(_toolCallId, params, signal, onUpdate, _ctx) {
@@ -408,7 +408,7 @@ export default function ocrExtension(pi: ExtensionAPI) {
     const text = config.backend === "ollama"
       ? `OCR: ollama ${config.model}`
       : `OCR: ${config.backend}`;
-    ctx.ui.setStatus("minimodel-ocr", text);
+    ctx.ui.setStatus("pi-ocr", text);
   }
 
   // ── Startup ────────────────────────────────────────────────────────────────
@@ -430,5 +430,5 @@ export default function ocrExtension(pi: ExtensionAPI) {
     }
   });
 
-  console.log("[pi-ocr] Loaded — /ocr (file or settings), tool: minimodel_ocr, default: mineru");
+  console.log("[pi-ocr] Loaded — /ocr (file or settings), tool: pi_ocr, default: mineru");
 }

package/extensions/mineru.ts

@@ -1,5 +1,5 @@
 /**
- * pi-minimodel-ocr — MinerU API backend
+ * pi-ocr — MinerU API backend
  *
  * Uses the free Agent Lightweight API (no token required):
  *   - File ≤10MB, ≤20 pages → one free request

package/extensions/ollama.ts

@@ -1,5 +1,5 @@
 /**
- * pi-minimodel-ocr — Ollama backend
+ * pi-ocr — Ollama backend
  *
  * Uses any locally-running Ollama vision model (default: glm-ocr) to OCR
  * images and PDFs. Converts PDF pages to PNG before sending to Ollama.

package/extensions/pix2text.ts

@@ -1,5 +1,5 @@
 /**
- * pi-minimodel-ocr — Pix2Text backend
+ * pi-ocr — Pix2Text backend
  *
  * Uses Pix2Text (https://github.com/breezedeus/Pix2Text) — an open-source
  * Python alternative to Mathpix. Recognizes layouts, text, math formulas (LaTeX),

package/extensions/types.ts

@@ -1,5 +1,5 @@
 /**
- * pi-minimodel-ocr — shared types for OCR backends
+ * pi-ocr — shared types for OCR backends
  */
 
 export const TASKS = ["text", "formula", "table", "figure", "auto"] as const;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-ocr",
-  "version": "1.0.0",
+  "version": "1.0.1",
   "description": "Pi extension: Zero-setup multi-backend OCR — MinerU (free cloud), Ollama (local GPU, LaTeX formulas), Pix2Text (local Python). Extract text, formulas, and tables from images and PDFs. Default: zero config, works out of the box.",
   "keywords": [
     "pi-package",