@comfanion/usethis_search 3.0.0-dev.15 → 3.0.0-dev.17

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@comfanion/usethis_search",
-  "version": "3.0.0-dev.15",
+  "version": "3.0.0-dev.17",
   "description": "OpenCode plugin: semantic search with graph-based context (v3: graph relations, 1-hop context, LSP + regex analyzers)",
   "type": "module",
   "main": "./index.ts",

package/vectorizer/chunkers/code-chunker.ts

@@ -23,6 +23,8 @@ export interface CodeChunk {
   content: string
   function_name?: string
   class_name?: string
+  start_line?: number
+  end_line?: number
 }
 
 // ── Block detection ─────────────────────────────────────────────────────────
@@ -172,31 +174,74 @@ function findPythonBlockEnd(lines: string[], startLine: number): number {
   return lines.length - 1
 }
 
-// ── Fallback: line-based splitting ──────────────────────────────────────────
+  // ── Fallback: line-based splitting ──────────────────────────────────────────
+
+  function splitByLines(lines: string[], maxChars: number): CodeChunk[] {
+    const chunks: CodeChunk[] = []
+    let current: string[] = []
+    let currentLen = 0
+    let startLine = 0
+
+    for (let i = 0; i < lines.length; i++) {
+      const line = lines[i]
+      if (currentLen + line.length + 1 > maxChars && current.length > 0) {
+        chunks.push({ content: current.join("\n"), start_line: startLine, end_line: i - 1 })
+        current = []
+        currentLen = 0
+        startLine = i
+      }
+      current.push(line)
+      currentLen += line.length + 1
+    }
 
-function splitByLines(lines: string[], maxChars: number): CodeChunk[] {
-  const chunks: CodeChunk[] = []
-  let current: string[] = []
-  let currentLen = 0
-
-  for (const line of lines) {
-    if (currentLen + line.length + 1 > maxChars && current.length > 0) {
-      chunks.push({ content: current.join("\n") })
-      current = []
-      currentLen = 0
+    if (current.length > 0) {
+      chunks.push({ content: current.join("\n"), start_line: startLine, end_line: lines.length - 1 })
     }
-    current.push(line)
-    currentLen += line.length + 1
-  }
 
-  if (current.length > 0) {
-    chunks.push({ content: current.join("\n") })
+    return chunks
   }
 
-  return chunks
-}
+  // ── Split large chunks preserving line numbers ────────────────────────────
+
+  function splitChunkByLines(chunk: CodeChunk, maxChars: number): CodeChunk[] {
+    const lines = chunk.content.split("\n")
+    const baseLine = chunk.start_line || 0
+
+    const parts: CodeChunk[] = []
+    let current: string[] = []
+    let currentLen = 0
+    let startLine = baseLine
+
+    for (let i = 0; i < lines.length; i++) {
+      const line = lines[i]
+      if (currentLen + line.length + 1 > maxChars && current.length > 0) {
+        parts.push({
+          ...chunk,
+          content: current.join("\n"),
+          start_line: startLine,
+          end_line: baseLine + i - 1,
+        })
+        current = []
+        currentLen = 0
+        startLine = baseLine + i
+      }
+      current.push(line)
+      currentLen += line.length + 1
+    }
+
+    if (current.length > 0) {
+      parts.push({
+        ...chunk,
+        content: current.join("\n"),
+        start_line: startLine,
+        end_line: baseLine + lines.length - 1,
+      })
+    }
+
+    return parts
+  }
 
-// ── Public API ──────────────────────────────────────────────────────────────
+  // ── Public API ──────────────────────────────────────────────────────────────
 
 /**
  * Chunk source code by functions/classes.
@@ -231,12 +276,13 @@ export function chunkCode(
     if (block.startLine > lastEnd + 1) {
       const gapContent = lines.slice(lastEnd + 1, block.startLine).join("\n").trim()
       if (gapContent.length >= config.min_chunk_size) {
-        chunks.push({ content: gapContent })
+        chunks.push({ content: gapContent, start_line: lastEnd + 1, end_line: block.startLine - 1 })
       } else if (gapContent.length > 0 && chunks.length > 0) {
         // Merge small gap with previous chunk
         chunks[chunks.length - 1].content += "\n\n" + gapContent
+        chunks[chunks.length - 1].end_line = block.startLine - 1
       } else if (gapContent.length > 0) {
-        chunks.push({ content: gapContent })
+        chunks.push({ content: gapContent, start_line: lastEnd + 1, end_line: block.startLine - 1 })
       }
     }
 
@@ -259,6 +305,8 @@ export function chunkCode(
               chunks.push({
                 content: gap,
                 class_name: block.name,
+                start_line: classLastEnd + 1,
+                end_line: method.startLine - 1,
               })
             }
           }
@@ -267,6 +315,8 @@ export function chunkCode(
             content: lines.slice(method.startLine, method.endLine + 1).join("\n"),
             function_name: method.name,
             class_name: block.name,
+            start_line: method.startLine,
+            end_line: method.endLine,
           })
           classLastEnd = method.endLine
         }
@@ -275,7 +325,7 @@ export function chunkCode(
         if (classLastEnd < block.endLine) {
           const tail = lines.slice(classLastEnd + 1, block.endLine + 1).join("\n").trim()
           if (tail) {
-            chunks.push({ content: tail, class_name: block.name })
+            chunks.push({ content: tail, class_name: block.name, start_line: classLastEnd + 1, end_line: block.endLine })
           }
         }
       } else {
@@ -312,9 +362,9 @@ export function chunkCode(
   const result: CodeChunk[] = []
   for (const chunk of chunks) {
     if (chunk.content.length > config.max_chunk_size) {
-      const parts = splitByLines(chunk.content.split("\n"), config.max_chunk_size)
+      const parts = splitChunkByLines(chunk, config.max_chunk_size)
       for (const p of parts) {
-        result.push({ ...chunk, content: p.content })
+        result.push(p)
       }
     } else {
       result.push(chunk)

package/vectorizer/chunkers/markdown-chunker.ts

@@ -22,14 +22,18 @@ export const DEFAULT_MD_CONFIG: MarkdownChunkConfig = {
 export interface MarkdownChunk {
   content: string
   heading_context: string   // "H1 > H2 > H3"
+  start_line?: number
+  end_line?: number
 }
 
 // ── Internal types ──────────────────────────────────────────────────────────
 
 interface Section {
-  level: number       // 1-6 for headings, 0 for preamble
+  level: number       //1-6 for headings, 0 for preamble
   heading: string
   body: string
+  start_line: number
+  end_line: number
 }
 
 // ── Parsing ─────────────────────────────────────────────────────────────────
@@ -38,19 +42,23 @@ interface Section {
 function parseSections(content: string): Section[] {
   const lines = content.split("\n")
   const sections: Section[] = []
-  let currentSection: Section = { level: 0, heading: "", body: "" }
+  let currentSection: Section = { level: 0, heading: "", body: "", start_line: 0, end_line: 0 }
 
-  for (const line of lines) {
+  for (let i = 0; i < lines.length; i++) {
+    const line = lines[i]
     const headingMatch = line.match(/^(#{1,6})\s+(.+)$/)
     if (headingMatch) {
       // Push previous section
       if (currentSection.body.trim() || currentSection.heading) {
+        currentSection.end_line = i - 1
         sections.push(currentSection)
       }
       currentSection = {
         level: headingMatch[1].length,
         heading: headingMatch[2].trim(),
         body: "",
+        start_line: i,
+        end_line: 0,
       }
     } else {
       currentSection.body += line + "\n"
@@ -59,6 +67,7 @@ function parseSections(content: string): Section[] {
 
   // Push last section
   if (currentSection.body.trim() || currentSection.heading) {
+    currentSection.end_line = lines.length - 1
     sections.push(currentSection)
   }
 
@@ -97,6 +106,45 @@ function splitLargeText(text: string, maxSize: number): string[] {
   return chunks
 }
 
+function splitLargeTextWithLines(text: string, maxSize: number, startLine: number): Array<{ content: string; start_line: number; end_line: number }> {
+  if (text.length <= maxSize) {
+    const lines = text.split("\n")
+    return [{ content: text, start_line: startLine, end_line: startLine + lines.length - 1 }]
+  }
+
+  const chunks: Array<{ content: string; start_line: number; end_line: number }> = []
+  const lines = text.split("\n")
+  let current: string[] = []
+  let currentLen = 0
+  let chunkStartLine = startLine
+
+  for (let i = 0; i < lines.length; i++) {
+    const line = lines[i]
+    if (currentLen + line.length + 1 > maxSize && current.length > 0) {
+      chunks.push({
+        content: current.join("\n"),
+        start_line: chunkStartLine,
+        end_line: startLine + i - 1,
+      })
+      current = []
+      currentLen = 0
+      chunkStartLine = startLine + i
+    }
+    current.push(line)
+    currentLen += line.length + 1
+  }
+
+  if (current.length > 0) {
+    chunks.push({
+      content: current.join("\n"),
+      start_line: chunkStartLine,
+      end_line: startLine + lines.length - 1,
+    })
+  }
+
+  return chunks
+}
+
 // ── Public API ──────────────────────────────────────────────────────────────
 
 /**
@@ -138,7 +186,12 @@ export function chunkMarkdown(
       ? `${"#".repeat(section.level)} ${section.heading}\n${section.body}`
       : section.body
 
-    rawChunks.push({ content: sectionText.trim(), heading_context: headingContext })
+    rawChunks.push({
+      content: sectionText.trim(),
+      heading_context: headingContext,
+      start_line: section.start_line,
+      end_line: section.end_line,
+    })
   }
 
   // Merge small sections with previous
@@ -150,7 +203,11 @@ export function chunkMarkdown(
     ) {
       const prev = merged[merged.length - 1]
       prev.content += "\n\n" + chunk.content
-      // Keep the deepest heading context
+      // Merge end_line
+      if (chunk.end_line !== undefined) {
+        prev.end_line = chunk.end_line
+      }
+      // Keep deepest heading context
       if (chunk.heading_context) {
         prev.heading_context = chunk.heading_context
       }
@@ -163,9 +220,14 @@ export function chunkMarkdown(
   const result: MarkdownChunk[] = []
   for (const chunk of merged) {
     if (chunk.content.length > config.max_chunk_size) {
-      const parts = splitLargeText(chunk.content, config.max_chunk_size)
+      const parts = splitLargeTextWithLines(chunk.content, config.max_chunk_size, chunk.start_line || 0)
       for (const part of parts) {
-        result.push({ content: part, heading_context: chunk.heading_context })
+        result.push({
+          content: part.content,
+          heading_context: chunk.heading_context,
+          start_line: part.start_line,
+          end_line: part.end_line,
+        })
       }
     } else {
       result.push(chunk)

package/vectorizer/graph-builder.ts

@@ -41,7 +41,7 @@ export class GraphBuilder {
     content: string,
     chunks: ChunkWithId[],
     fileType: "code" | "docs"
-  ): Promise<void> {
+  ): Promise<number> {
     let relations: Array<RegexRelation | LSPRelation> = []
 
     if (fileType === "docs") {
@@ -74,6 +74,7 @@ export class GraphBuilder {
     }))
 
     await this.graphDB.putEdges(triples)
+    return triples.length
   }
 
   resolveChunkId(filePath: string, line: number): string | null {

package/vectorizer/index.ts

@@ -562,7 +562,23 @@ class CodebaseIndexer {
 
     // v3: Delete old edges for this file and build new ones
     await this.graphDB.deleteByFile(relPath);
-    await this.graphBuilder.buildEdges(relPath, content, chunksWithIds, fileMeta.file_type);
+    const graphEdgesBuilt = await this.graphBuilder.buildEdges(relPath, content, chunksWithIds, fileMeta.file_type);
+
+    // Log graph creation to indexer.log
+    if (graphEdgesBuilt > 0 || DEBUG) {
+      const timestamp = new Date().toISOString().slice(11, 19);
+      const logMsg = `${timestamp} Graph built: ${relPath} (${chunksWithIds.length} chunks)`;
+      if (DEBUG) console.log(`[vectorizer] ${logMsg}`);
+      
+      // Write to indexer.log in .opencode directory
+      try {
+        const logPath = path.join(this.root, ".opencode", "indexer.log");
+        const fsSync = await import("fs");
+        fsSync.appendFileSync(logPath, `${logMsg}\n`);
+      } catch {
+        // non-fatal — logging is advisory
+      }
+    }
 
     // FR-054: Store graph build timestamp + file hash as metadata triple
     try {
@@ -590,6 +606,9 @@ class CodebaseIndexer {
         function_name: chunksWithIds[i].function_name || "",
         class_name: chunksWithIds[i].class_name || "",
         tags: (fileMeta.tags || []).join(","),
+        // Line numbers for "from-to" extraction
+        start_line: chunksWithIds[i].start_line,
+        end_line: chunksWithIds[i].end_line,
       });
     }