@gmickel/gno 1.29.6 → 1.30.1

package/README.md

@@ -20,13 +20,23 @@ gno mcp install --target cursor    # or claude-code, claude-desktop, zed, ...
 
 ## What you get
 
-BM25 and vector retrieval fused and reranked over Markdown, code, PDFs, Office files, and exported mail, calendar and transcripts. A web workspace with cross-collection browse, a knowledge graph, a markdown editor and provenance-carrying capture. A fast CLI, a TypeScript SDK, a REST API, a headless daemon, and one-command install into ten agent clients. Optional hosted publishing when a slice needs a URL.
+**One local index across everything you have.** Markdown, PDFs, Office documents, plain text, source code, and portable mail, calendar and transcript exports. Point it at a folder that mixes all of them and it handles the mix.
 
-That is the table stakes. Here is the part that is harder to find elsewhere.
+**Three ways to search it.** Keyword (BM25), semantic (vector), and hybrid — fused, reranked, and explainable, with structured intent controls and metadata filters. Ask for "how we handle retries" and find the paragraph about exponential backoff that never uses the word.
+
+**A workspace, not a search box.** Cross-collection folder tree, per-tab browse context, a markdown editor, provenance-carrying quick capture, and a navigable knowledge graph.
+
+**Answers with citations.** Ask a question in natural language and get an answer built from your own documents, with citations that resolve to the source passage.
+
+**Six interfaces on one index.** A fast CLI, a web UI, a REST API, a TypeScript SDK, an MCP server, and a headless daemon — plus one-command install into ten agent clients. Nothing drifts between them.
+
+**Optional hosted publishing** at [gno.sh](https://gno.sh) when a slice needs a URL.
+
+No GPU required, no account required, no telemetry. Free and MIT licensed.
 
 ## Why not just another local RAG tool
 
-Most retrieval tools return a ranked list and leave the rest to optimism. Four things here are different, and each one is measurable rather than adjectival:
+The above is most of what people use day to day. Beyond it, four things here are different, and each one is measurable rather than adjectival:
 
 |                                                              | What it does                                                                                                                                                                   | Why it matters                                                                                                                                                                           |
 | :----------------------------------------------------------- | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
@@ -37,7 +47,7 @@ Most retrieval tools return a ranked list and leave the rest to optimism. Four t
 
 Everything runs on your machine. Zero telemetry. The three network boundaries are explicit: downloading a model, configuring an HTTP inference endpoint, and uploading an artifact you exported for publishing.
 
-**And when it fails, that ships too.** The CJK lexical benchmark missed its own promotion gates, so the analyzer was not shipped and [the failing numbers were published](#general-multilingual-embedding-benchmark) instead.
+**And the receipts ship too.** Every benchmark behind these numbers is committed to the repository against a pinned corpus, with its limitations stated alongside the result, so you can replay it rather than take it on trust.
 
 ## Use it when
 
@@ -107,7 +117,7 @@ gno daemon --detach  # headless indexing + resident MCP gateway
 
 <!-- public-truth:current-version -->
 
-> Current release: **v1.29.6** — see [CHANGELOG.md](./CHANGELOG.md)
+> Current release: **v1.30.1** — see [CHANGELOG.md](./CHANGELOG.md)
 
 <!-- /public-truth -->
 

package/browser-extension/artifacts/gno-browser-clipper-v1.30.1.zip.sha256

@@ -0,0 +1 @@
+7f28a8daa50eec43dcd95d5af91002a5e25cd87682ff7c99905af8c9f3457b24  gno-browser-clipper-v1.30.1.zip

package/browser-extension/dist/manifest.json

@@ -21,5 +21,5 @@
   "content_security_policy": {
     "extension_pages": "script-src 'self'; object-src 'none'; connect-src http://127.0.0.1:*"
   },
-  "version": "1.29.6"
+  "version": "1.30.1"
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@gmickel/gno",
-  "version": "1.29.6",
+  "version": "1.30.1",
   "description": "Local semantic search for your documents. Index Markdown, PDF, and Office files with hybrid BM25 + vector search.",
   "keywords": [
     "embeddings",
@@ -66,6 +66,8 @@
     "test:web": "bun test test/serve/public --timeout 30000",
     "test:e2e": "bun scripts/web-ui-smoke.ts",
     "test:e2e:clipper": "bun test test/clipper/e2e.test.ts --timeout 240000",
+    "test:e2e:pdf": "bun scripts/pdf-viewer-smoke.ts",
+    "smoke:pdf-viewer": "bun scripts/pdf-viewer-smoke.ts",
     "test:e2e:install": "bunx playwright install chromium",
     "test:package": "bun scripts/package-smoke.ts",
     "test:package:clipper": "bun scripts/package-smoke-clipper-cli.ts",
@@ -180,6 +182,7 @@
     "nanoid": "5.1.6",
     "node-llama-cpp": "3.18.1",
     "officeparser": "6.0.4",
+    "pdfjs-dist": "5.7.284",
     "picocolors": "1.1.1",
     "react": "19.2.4",
     "react-dom": "19.2.4",

package/src/core/network-boundary-inventory.ts

@@ -218,6 +218,57 @@ export const NETWORK_BOUNDARY_INVENTORY = [
     action: "serve",
     enforcement: "loopback_only",
   },
+  {
+    // Loopback self-request that warms the SPA shell cache from the server's
+    // own bound port (127.0.0.1). Carries no collection data and never leaves
+    // the host.
+    id: "serve-spa-shell",
+    key: "src/serve/server.ts::fetch#1",
+    path: "src/serve/server.ts",
+    primitive: "fetch",
+    action: "serve",
+    enforcement: "loopback_only",
+  },
+  {
+    // Windows-only private SPA bundle listener. Bun does not support Unix
+    // sockets on Windows, so this binds an ephemeral loopback port instead.
+    id: "serve-spa-bundle-windows-listener",
+    key: "src/serve/spa-bundle-source.ts::bun_serve#1",
+    path: "src/serve/spa-bundle-source.ts",
+    primitive: "bun_serve",
+    action: "serve",
+    enforcement: "loopback_only",
+  },
+  {
+    // Unix-only private SPA bundle listener. The generated assets are exposed
+    // through a process-owned Unix-domain socket, never a TCP interface.
+    id: "serve-spa-bundle-unix-listener",
+    key: "src/serve/spa-bundle-source.ts::bun_serve#2",
+    path: "src/serve/spa-bundle-source.ts",
+    primitive: "bun_serve",
+    action: "serve",
+    enforcement: "local_process_only",
+  },
+  {
+    // Windows-only request from the public server to its private, ephemeral
+    // loopback SPA bundle listener.
+    id: "serve-spa-bundle-windows-fetch",
+    key: "src/serve/spa-bundle-source.ts::fetch#1",
+    path: "src/serve/spa-bundle-source.ts",
+    primitive: "fetch",
+    action: "serve",
+    enforcement: "loopback_only",
+  },
+  {
+    // Unix-only request over the process-owned SPA bundle socket. The HTTP URL
+    // is a Bun fetch API requirement; the request is routed through `unix`.
+    id: "serve-spa-bundle-unix-fetch",
+    key: "src/serve/spa-bundle-source.ts::fetch#2",
+    path: "src/serve/spa-bundle-source.ts",
+    primitive: "fetch",
+    action: "serve",
+    enforcement: "local_process_only",
+  },
   {
     id: "http-mcp-tools",
     key: "logical::http-mcp-tools",

package/src/serve/AGENTS.md

@@ -74,6 +74,8 @@ Answer generation uses shared module to stay in sync with CLI:
 | `/api/publish/export` | POST   | Export gno.sh publish artifact JSON        |
 | `/api/docs`           | GET    | List documents                             |
 | `/api/doc`            | GET    | Get document content                       |
+| `/api/doc-asset`      | GET    | Original source file bytes (Range, HEAD)   |
+| `/vendor/pdfjs/*`     | GET    | Same-origin pdfjs worker/cmaps/fonts       |
 | `/api/search`         | POST   | BM25 search                                |
 | `/api/query`          | POST   | Hybrid search                              |
 | `/api/ask`            | POST   | AI answer with citations                   |
@@ -106,7 +108,9 @@ gno serve --port 3000
 - Binds to `127.0.0.1` only (no LAN exposure)
 - CSP headers on all responses
 - CORS protection on POST endpoints
-- No external font/script loading
+- No external font/script loading — the PDF.js worker, cMaps, and standard fonts
+  are served same-origin from the installed `pdfjs-dist` package via
+  `/vendor/pdfjs/*`, so the CSP keeps `worker-src 'self'` and `font-src 'self'`
 
 ## Bun.serve() Patterns
 

package/src/serve/CLAUDE.md

@@ -76,6 +76,8 @@ Answer generation uses shared module to stay in sync with CLI:
 | `/api/publish/export` | POST   | Export gno.sh publish artifact JSON        |
 | `/api/docs`           | GET    | List documents                             |
 | `/api/doc`            | GET    | Get document content                       |
+| `/api/doc-asset`      | GET    | Original source file bytes (Range, HEAD)   |
+| `/vendor/pdfjs/*`     | GET    | Same-origin pdfjs worker/cmaps/fonts       |
 | `/api/search`         | POST   | BM25 search                                |
 | `/api/query`          | POST   | Hybrid search                              |
 | `/api/ask`            | POST   | AI answer with citations                   |
@@ -108,7 +110,9 @@ gno serve --port 3000
 - Binds to `127.0.0.1` only (no LAN exposure)
 - CSP headers on all responses
 - CORS protection on POST endpoints
-- No external font/script loading
+- No external font/script loading — the PDF.js worker, cMaps, and standard fonts
+  are served same-origin from the installed `pdfjs-dist` package via
+  `/vendor/pdfjs/*`, so the CSP keeps `worker-src 'self'` and `font-src 'self'`
 
 ## Bun.serve() Patterns
 

package/src/serve/fn112-routes.ts

@@ -0,0 +1,232 @@
+/**
+ * Production route factories for fn-112 doc-asset + pdfjs vendor surfaces.
+ * server.ts and tests MUST consume these — no duplicated route maps.
+ *
+ * Vendor: ONE production dispatcher (`handlePdfjsVendorRequest`) is the sole
+ * path for every /vendor/pdfjs/* request (valid or malformed, any method).
+ * server.ts routes that prefix through fetch; tests call the same function.
+ */
+
+import type { Config } from "../config/types";
+import type { SqliteAdapter } from "../store/sqlite/adapter";
+import type { ResidentRuntime } from "./resident-runtime";
+
+import { handlePdfjsAsset, PDFJS_ASSET_CACHE_CONTROL } from "./pdfjs-assets";
+import { handleResidentRead } from "./resident-request";
+import { handleDocAsset } from "./routes/api";
+
+const PDFJS_WORKER_BOOTSTRAP = `
+if (typeof Math.sumPrecise !== "function") {
+  Object.defineProperty(Math, "sumPrecise", {
+    configurable: true,
+    writable: true,
+    value(values) {
+      let sum = 0;
+      let compensation = 0;
+      for (const value of values) {
+        const next = sum + value;
+        compensation += Math.abs(sum) >= Math.abs(value)
+          ? sum - next + value
+          : value - next + sum;
+        sum = next;
+      }
+      return sum + compensation;
+    },
+  });
+}
+await import("/vendor/pdfjs/pdf.worker.raw.min.mjs");
+`.trimStart();
+
+export type SecurityHeaderWrap = (
+  response: Response,
+  isDev: boolean
+) => Response;
+
+export type DocAssetRouteContext = {
+  store: SqliteAdapter;
+  /** Live config getter (ctxHolder.config may change). */
+  getConfig: () => Config;
+  runtime: ResidentRuntime;
+  isDev: boolean;
+  withSecurityHeaders: SecurityHeaderWrap;
+};
+
+export type MethodHandlers = {
+  GET: (req: Request) => Promise<Response> | Response;
+  HEAD: (req: Request) => Promise<Response> | Response;
+};
+
+/** Exact production error envelopes for vendor routes (I1-04 round 3). */
+export const PDFJS_VENDOR_ERRORS = {
+  NOT_FOUND: {
+    code: "NOT_FOUND",
+    message: "Asset not found",
+  },
+  METHOD_NOT_ALLOWED: {
+    code: "METHOD_NOT_ALLOWED",
+    message: "Only GET and HEAD are supported",
+  },
+} as const;
+
+/**
+ * Production `/api/doc-asset` GET+HEAD handlers.
+ * Both methods traverse handleResidentRead (admission) then withSecurityHeaders.
+ */
+export function createDocAssetRouteHandlers(
+  ctx: DocAssetRouteContext
+): MethodHandlers {
+  const dispatch = async (req: Request): Promise<Response> => {
+    const url = new URL(req.url);
+    return ctx.withSecurityHeaders(
+      await handleResidentRead(ctx.runtime, req, () =>
+        handleDocAsset(ctx.store, ctx.getConfig(), url, req)
+      ),
+      ctx.isDev
+    );
+  };
+  return {
+    GET: dispatch,
+    HEAD: dispatch,
+  };
+}
+
+export type PdfjsVendorDispatchOptions = {
+  isDev: boolean;
+  withSecurityHeaders: SecurityHeaderWrap;
+};
+
+/**
+ * Production vendor dispatcher — the ONLY production path for `/vendor/pdfjs/*`.
+ *
+ * Handles every method and every pathname under the prefix (worker, cmaps,
+ * standard_fonts, multi-segment, empty, encoded traversal, invalid encoding).
+ * ALL responses — success, 404, 405 — go through withSecurityHeaders exactly once.
+ *
+ * server.ts must call this for the prefix (via fetch); tests must call this
+ * identical function (no separate test-only fallback).
+ */
+export async function handlePdfjsVendorRequest(
+  request: Request,
+  options: PdfjsVendorDispatchOptions
+): Promise<Response> {
+  const wrap = options.withSecurityHeaders;
+  const isDev = options.isDev;
+  const method = request.method.toUpperCase();
+  const pathname = new URL(request.url).pathname;
+
+  // Never double-wrap: build the inner response, then wrap once.
+  const inner = await dispatchPdfjsVendorInner(method, pathname);
+  return wrap(inner, isDev);
+}
+
+/**
+ * True when this request should be handled by the vendor dispatcher.
+ * Used by server fetch to claim the whole prefix.
+ */
+export function isPdfjsVendorPath(pathname: string): boolean {
+  return pathname === "/vendor/pdfjs" || pathname.startsWith("/vendor/pdfjs/");
+}
+
+async function dispatchPdfjsVendorInner(
+  method: string,
+  pathname: string
+): Promise<Response> {
+  // Only GET/HEAD are ever valid on any vendor path
+  if (method !== "GET" && method !== "HEAD") {
+    return methodNotAllowedResponse();
+  }
+
+  // Public worker bootstrap: install the ES2026 Math.sumPrecise compatibility
+  // required by pdfjs-dist before evaluating its worker module.
+  if (pathname === "/vendor/pdfjs/pdf.worker.min.mjs") {
+    const headers = {
+      "Cache-Control": PDFJS_ASSET_CACHE_CONTROL,
+      "Content-Type": "text/javascript",
+      "Content-Length": String(
+        new TextEncoder().encode(PDFJS_WORKER_BOOTSTRAP).byteLength
+      ),
+    };
+    return new Response(method === "HEAD" ? null : PDFJS_WORKER_BOOTSTRAP, {
+      headers,
+    });
+  }
+
+  // Package worker stays behind a fixed same-origin route; never referenced
+  // directly by application code, only by the bootstrap above.
+  if (pathname === "/vendor/pdfjs/pdf.worker.raw.min.mjs") {
+    return handlePdfjsAsset({ kind: "worker", method });
+  }
+
+  // cmaps/:file — single segment only
+  const cmapMatch = pathname.match(/^\/vendor\/pdfjs\/cmaps\/([^/]*)$/u);
+  if (cmapMatch) {
+    const raw = cmapMatch[1] ?? "";
+    if (raw === "") {
+      return notFoundResponse();
+    }
+    const file = decodeRouteSegment(raw);
+    if (file === null) {
+      return notFoundResponse();
+    }
+    return handlePdfjsAsset({ kind: "cmaps", file, method });
+  }
+
+  // standard_fonts/:file — single segment only
+  const fontMatch = pathname.match(
+    /^\/vendor\/pdfjs\/standard_fonts\/([^/]*)$/u
+  );
+  if (fontMatch) {
+    const raw = fontMatch[1] ?? "";
+    if (raw === "") {
+      return notFoundResponse();
+    }
+    const file = decodeRouteSegment(raw);
+    if (file === null) {
+      return notFoundResponse();
+    }
+    return handlePdfjsAsset({ kind: "standard_fonts", file, method });
+  }
+
+  // Multi-segment, empty prefix, unknown subpaths, traversal that doesn't
+  // collapse to a single-segment match → 404
+  return notFoundResponse();
+}
+
+function decodeRouteSegment(segment: string): string | null {
+  try {
+    return decodeURIComponent(segment);
+  } catch {
+    return null;
+  }
+}
+
+function notFoundResponse(): Response {
+  return Response.json(
+    { error: { ...PDFJS_VENDOR_ERRORS.NOT_FOUND } },
+    { status: 404 }
+  );
+}
+
+function methodNotAllowedResponse(): Response {
+  return Response.json(
+    { error: { ...PDFJS_VENDOR_ERRORS.METHOD_NOT_ALLOWED } },
+    { status: 405 }
+  );
+}
+
+/**
+ * @deprecated Use handlePdfjsVendorRequest — kept only if any import remains.
+ * Prefer the single production dispatcher.
+ */
+export async function dispatchPdfjsVendorRoute(
+  _routeMap: unknown,
+  request: Request,
+  options?: PdfjsVendorDispatchOptions
+): Promise<Response> {
+  if (!options) {
+    throw new Error(
+      "dispatchPdfjsVendorRoute requires production options; use handlePdfjsVendorRequest"
+    );
+  }
+  return handlePdfjsVendorRequest(request, options);
+}