@gmickel/gno 1.36.0 → 1.36.1

package/browser-extension/artifacts/gno-browser-clipper-v1.36.1.zip.sha256

@@ -0,0 +1 @@
+8ece31111978af50978a899662a7bd14eddcc90f2c90f6b0f04b67b3b602d980  gno-browser-clipper-v1.36.1.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": "1.36.0"
+  "version": "1.36.1"
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@gmickel/gno",
-  "version": "1.36.0",
+  "version": "1.36.1",
   "description": "Local semantic search for your documents. Index Markdown, PDF, and Office files with hybrid BM25 + vector search.",
   "keywords": [
     "embeddings",

package/src/core/context-evidence.ts

@@ -413,13 +413,19 @@ export const materializeContextEvidenceCandidates = async (
     const content = contentByHash.get(document.mirrorHash);
     const chunk = getChunk(document.mirrorHash, metadata.seq);
     const chunkKey = `${document.mirrorHash}:${metadata.seq}`;
+    // Display snippets may omit leading YAML frontmatter, so startLine can sit
+    // after the stored chunk start. endLine must still identify the same chunk.
+    const displayRangeWithinChunk =
+      chunk !== undefined &&
+      snippetRange.endLine === chunk.endLine &&
+      snippetRange.startLine >= chunk.startLine &&
+      snippetRange.startLine <= chunk.endLine;
     if (
       !content ||
       metadata.mirrorHash !== document.mirrorHash ||
       !chunk ||
       chunk.mirrorHash !== document.mirrorHash ||
-      snippetRange.startLine !== chunk.startLine ||
-      snippetRange.endLine !== chunk.endLine
+      !displayRangeWithinChunk
     ) {
       throw new ContextEvidenceError(
         "chunk_coordinate_mismatch",

package/src/pipeline/hybrid.ts

@@ -58,6 +58,7 @@ import {
 } from "./query-modes";
 import { rerankCandidates } from "./rerank";
 import { attachSearchResultContexts } from "./result-context";
+import { cleanDisplaySnippet } from "./snippet";
 import {
   isWithinTemporalRange,
   resolveRecencyTimestamp,
@@ -1003,6 +1004,7 @@ export async function searchHybrid(
           ) ?? chunk);
 
     let snippet = snippetChunk.text;
+    let snippetStartLine = snippetChunk.startLine;
     let snippetRange: { startLine: number; endLine: number } | undefined = {
       startLine: snippetChunk.startLine,
       endLine: snippetChunk.endLine,
@@ -1021,6 +1023,18 @@ export async function searchHybrid(
         snippetRange = undefined; // Full content has no range
       }
       // Fallback to chunk text if content unavailable
+    } else {
+      const cleanedSnippet = cleanDisplaySnippet(
+        snippetChunk.text,
+        snippetChunk.text
+      );
+      snippet = cleanedSnippet.text;
+      snippetStartLine =
+        snippetChunk.startLine + cleanedSnippet.startLineOffset;
+      snippetRange = {
+        startLine: snippetStartLine,
+        endLine: snippetChunk.endLine,
+      };
     }
 
     for (const doc of candidateDocs) {
@@ -1050,7 +1064,7 @@ export async function searchHybrid(
         title: doc.title ?? undefined,
         contentType: doc.contentType ?? undefined,
         categories: doc.categories ?? undefined,
-        line: snippetChunk.startLine,
+        line: snippetStartLine,
         snippet,
         snippetLanguage: chunk.language ?? undefined,
         snippetRange,

package/src/pipeline/search.ts

@@ -30,6 +30,7 @@ import { selectBestChunkForSteering } from "./intent";
 import { hasProjectAffinity } from "./project-affinity";
 import { detectQueryLanguage } from "./query-language";
 import { attachSearchResultContexts } from "./result-context";
+import { cleanDisplaySnippet } from "./snippet";
 import {
   resolveRecencyTimestamp,
   resolveTemporalRange,
@@ -107,6 +108,7 @@ function buildSearchResult(ctx: BuildResultContext): SearchResult {
   // Determine snippet content and range
   let snippet: string;
   let snippetRange: { startLine: number; endLine: number } | undefined;
+  let line = chunk?.startLine;
 
   if (options?.full && fullContent) {
     // --full: use full content, no range (full doc)
@@ -117,11 +119,18 @@ function buildSearchResult(ctx: BuildResultContext): SearchResult {
     snippet = chunk.text;
     snippetRange = { startLine: chunk.startLine, endLine: chunk.endLine };
   } else {
-    // Default: use FTS snippet or chunk text
-    snippet = fts.snippet ?? chunk?.text ?? "";
-    snippetRange = chunk
-      ? { startLine: chunk.startLine, endLine: chunk.endLine }
-      : undefined;
+    // Default: FTS snippet or chunk text, with leading frontmatter stripped
+    const cleaned = cleanDisplaySnippet(
+      fts.snippet ?? chunk?.text ?? "",
+      chunk?.text
+    );
+    snippet = cleaned.text;
+    if (chunk) {
+      line = chunk.startLine + cleaned.startLineOffset;
+      snippetRange = { startLine: line, endLine: chunk.endLine };
+    } else {
+      snippetRange = undefined;
+    }
   }
 
   const result: SearchResult = {
@@ -131,7 +140,7 @@ function buildSearchResult(ctx: BuildResultContext): SearchResult {
     title: fts.title,
     contentType: fts.contentType,
     categories: fts.categories,
-    line: chunk?.startLine,
+    line,
     snippet,
     snippetLanguage: chunk?.language ?? undefined,
     snippetRange,

package/src/pipeline/snippet.ts

@@ -0,0 +1,203 @@
+/**
+ * Display-layer snippet cleaning for search/query results.
+ * Strips leading YAML frontmatter so snippets prefer prose. Does not change
+ * indexed text, `--full` content, or `--line-numbers` raw chunks.
+ *
+ * @module src/pipeline/snippet
+ */
+
+import { stripFrontmatter } from "../ingestion/frontmatter";
+
+/** FTS5 highlight markers from snippet(documents_fts, ..., '<mark>', '</mark>', '...', 32). */
+const MARK_TAG_REGEX = /<\/?mark>/g;
+
+/** Leading blank lines after a closed frontmatter fence. */
+const LEADING_BLANK_LINES_REGEX = /^(?:[ \t]*\r?\n)+/;
+
+/** YAML mapping line (`key: value` or `key:`). */
+const YAML_MAPPING_LINE_REGEX = /^[\w./-]+\s*:/;
+
+/** YAML sequence item. */
+const YAML_SEQUENCE_LINE_REGEX = /^- /;
+
+export interface DisplaySnippet {
+  text: string;
+  /**
+   * Lines to add to the chunk's startLine when the emitted text is derived
+   * from stripped chunk prose (not a kept FTS window).
+   */
+  startLineOffset: number;
+  /** True when a frontmatter-dominated FTS snippet was replaced by chunk prose. */
+  usedChunkFallback: boolean;
+}
+
+/**
+ * Clean a default-path snippet: strip a leading closed YAML fence, or replace
+ * an FTS window that is only frontmatter with stripped chunk prose.
+ * Never returns an empty string when the original text had content.
+ */
+export function cleanDisplaySnippet(
+  snippet: string,
+  chunkText?: string
+): DisplaySnippet {
+  const strippedSnippet = stripLeadingFrontmatterBlock(snippet);
+  if (strippedSnippet.didStrip) {
+    return {
+      text: strippedSnippet.text,
+      startLineOffset: strippedSnippet.lineCount,
+      usedChunkFallback: false,
+    };
+  }
+
+  const afterEmbeddedFence = proseAfterEmbeddedFrontmatterFence(snippet);
+  if (afterEmbeddedFence !== undefined) {
+    const cleanedChunk =
+      chunkText === undefined
+        ? undefined
+        : stripLeadingFrontmatterBlock(chunkText);
+    return {
+      text: afterEmbeddedFence,
+      startLineOffset: cleanedChunk?.didStrip ? cleanedChunk.lineCount : 0,
+      usedChunkFallback: false,
+    };
+  }
+
+  const canFallback =
+    chunkText !== undefined &&
+    chunkText !== snippet &&
+    isFrontmatterDominatedSnippet(snippet);
+  if (canFallback) {
+    const cleanedChunk = stripLeadingFrontmatterBlock(chunkText);
+    if (cleanedChunk.text.length > 0) {
+      return {
+        text: cleanedChunk.text,
+        startLineOffset: cleanedChunk.didStrip ? cleanedChunk.lineCount : 0,
+        usedChunkFallback: true,
+      };
+    }
+  }
+
+  return {
+    text: snippet,
+    startLineOffset: 0,
+    usedChunkFallback: false,
+  };
+}
+
+/** True for FTS-style snippets that are only (or start as) YAML frontmatter. */
+export function isFrontmatterDominatedSnippet(text: string): boolean {
+  const unmarked = text.replace(MARK_TAG_REGEX, "");
+  const trimmed = unmarked.trimStart();
+  const withoutLeadingEllipsis = trimmed.startsWith("...")
+    ? trimmed.slice(3).trimStart()
+    : trimmed;
+  if (withoutLeadingEllipsis.startsWith("---")) {
+    return true;
+  }
+
+  const contentLines = unmarked.split(/\r?\n/).filter((line) => {
+    const trimmedLine = line.trim();
+    return trimmedLine.length > 0 && trimmedLine !== "...";
+  });
+  if (contentLines.length === 0) {
+    return false;
+  }
+  if (contentLines.every(isYamlFrontmatterLine)) {
+    return true;
+  }
+  return proseAfterEmbeddedFrontmatterFence(text) !== undefined;
+}
+
+/**
+ * FTS windows often straddle the closing fence (`...yaml\n---\n# Heading`).
+ * Keep the prose after that fence when the prefix looks like YAML.
+ */
+function proseAfterEmbeddedFrontmatterFence(text: string): string | undefined {
+  const lines = text.split(/\r?\n/);
+  for (let i = 1; i < lines.length; i++) {
+    const line = lines[i];
+    if (line === undefined) {
+      continue;
+    }
+    if (line.replace(MARK_TAG_REGEX, "").trim() !== "---") {
+      continue;
+    }
+    const prefixLines = lines.slice(0, i);
+    if (!prefixLooksLikeFrontmatter(prefixLines)) {
+      continue;
+    }
+    const after = lines
+      .slice(i + 1)
+      .join("\n")
+      .replace(LEADING_BLANK_LINES_REGEX, "");
+    if (after.trim().length === 0) {
+      continue;
+    }
+    return after;
+  }
+  return undefined;
+}
+
+function prefixLooksLikeFrontmatter(lines: string[]): boolean {
+  const content = lines.filter((line) => {
+    const trimmed = line.replace(MARK_TAG_REGEX, "").trim();
+    return trimmed.length > 0 && trimmed !== "...";
+  });
+  if (content.length === 0) {
+    return true;
+  }
+  return content.every((line, index) => {
+    const trimmed = line.replace(MARK_TAG_REGEX, "").trim();
+    if (index === 0 && trimmed.startsWith("...")) {
+      return true;
+    }
+    return isYamlFrontmatterLine(trimmed);
+  });
+}
+
+function stripLeadingFrontmatterBlock(text: string): {
+  text: string;
+  didStrip: boolean;
+  lineCount: number;
+} {
+  const afterFence = stripFrontmatter(text);
+  if (afterFence === text) {
+    return { text, didStrip: false, lineCount: 0 };
+  }
+
+  const withoutBlanks = afterFence.replace(LEADING_BLANK_LINES_REGEX, "");
+  if (withoutBlanks.trim().length === 0) {
+    return { text, didStrip: false, lineCount: 0 };
+  }
+
+  const prefix = text.slice(0, text.length - withoutBlanks.length);
+  return {
+    text: withoutBlanks,
+    didStrip: true,
+    lineCount: countConsumedLines(prefix),
+  };
+}
+
+function isYamlFrontmatterLine(line: string): boolean {
+  const trimmed = line.trim();
+  if (trimmed === "---") {
+    return true;
+  }
+  if (YAML_SEQUENCE_LINE_REGEX.test(trimmed)) {
+    return true;
+  }
+  return YAML_MAPPING_LINE_REGEX.test(trimmed);
+}
+
+function countConsumedLines(prefix: string): number {
+  if (prefix.length === 0) {
+    return 0;
+  }
+  let newlineCount = 0;
+  for (const char of prefix) {
+    if (char === "\n") {
+      newlineCount += 1;
+    }
+  }
+  return prefix.endsWith("\n") ? newlineCount : newlineCount + 1;
+}

package/src/pipeline/vsearch.ts

@@ -28,6 +28,7 @@ import { selectBestChunkForSteering } from "./intent";
 import { hasProjectAffinity } from "./project-affinity";
 import { detectQueryLanguage } from "./query-language";
 import { attachSearchResultContexts } from "./result-context";
+import { cleanDisplaySnippet } from "./snippet";
 import {
   resolveRecencyTimestamp,
   isWithinTemporalRange,
@@ -236,6 +237,8 @@ export async function searchVectorWithEmbedding(
         continue;
       }
 
+      const cleanedSnippet = cleanDisplaySnippet(chunk.text, chunk.text);
+      const snippetStartLine = chunk.startLine + cleanedSnippet.startLineOffset;
       const scoredResult = applyContentTypeBoost(
         {
           docid: doc.docid,
@@ -244,11 +247,11 @@ export async function searchVectorWithEmbedding(
           title: doc.title ?? undefined,
           contentType: doc.contentType ?? undefined,
           categories: doc.categories ?? undefined,
-          line: chunk.startLine,
-          snippet: chunk.text,
+          line: snippetStartLine,
+          snippet: cleanedSnippet.text,
           snippetLanguage: chunk.language ?? undefined,
           snippetRange: {
-            startLine: chunk.startLine,
+            startLine: snippetStartLine,
             endLine: chunk.endLine,
           },
           source: {
@@ -338,6 +341,12 @@ export async function searchVectorWithEmbedding(
 
       const collectionPath = collectionPaths.get(doc.collection);
       const sourceRelPath = doc.recordSourcePath ?? doc.relPath;
+      const cleanedChunk = fullContent
+        ? undefined
+        : cleanDisplaySnippet(chunk.text, chunk.text);
+      const snippetStartLine = cleanedChunk
+        ? chunk.startLine + cleanedChunk.startLineOffset
+        : chunk.startLine;
 
       const result = applyContentTypeBoost(
         {
@@ -347,13 +356,13 @@ export async function searchVectorWithEmbedding(
           title: doc.title ?? undefined,
           contentType: doc.contentType ?? undefined,
           categories: doc.categories ?? undefined,
-          line: chunk.startLine,
-          snippet: fullContent ?? chunk.text,
+          line: snippetStartLine,
+          snippet: fullContent ?? cleanedChunk?.text ?? chunk.text,
           snippetLanguage: chunk.language ?? undefined,
           // --full: no snippetRange (full doc content)
           snippetRange: fullContent
             ? undefined
-            : { startLine: chunk.startLine, endLine: chunk.endLine },
+            : { startLine: snippetStartLine, endLine: chunk.endLine },
           source: {
             relPath: sourceRelPath,
             absPath: collectionPath

package/browser-extension/artifacts/gno-browser-clipper-v1.36.0.zip.sha256

@@ -1 +0,0 @@
-8d8b73a55081409e3dbba4e6927d864839329cbcdb98a8f78274c29688edeb9e  gno-browser-clipper-v1.36.0.zip