fix: improve first-run indexing diagnostics

src/config/constants.ts

@@ -102,6 +102,13 @@ export const EMBEDDING_MODELS = {
 export const DEFAULT_PROVIDER_MODELS = {
   "github-copilot": "text-embedding-3-small",
   "openai": "text-embedding-3-small",
-  "google": "text-embedding-005",
+  "google": "gemini-embedding-001",
   "ollama": "nomic-embed-text",
 } as const
+
+export const AUTO_DETECT_PROVIDER_ORDER = [
+  "github-copilot",
+  "openai",
+  "google",
+  "ollama",
+] as const;

src/config/schema.ts

@@ -1,6 +1,6 @@
 // Config schema without zod dependency to avoid version conflicts with OpenCode SDK
 
-import { DEFAULT_INCLUDE, DEFAULT_EXCLUDE, EMBEDDING_MODELS, DEFAULT_PROVIDER_MODELS } from "./constants.js";
+import { AUTO_DETECT_PROVIDER_ORDER, DEFAULT_INCLUDE, DEFAULT_EXCLUDE, EMBEDDING_MODELS, DEFAULT_PROVIDER_MODELS } from "./constants.js";
 import { substituteEnvString } from "./env-substitution.js";
 
 export type IndexScope = "project" | "global";
@@ -426,6 +426,10 @@ export type EmbeddingProvider = keyof typeof EMBEDDING_MODELS;
 
 export const availableProviders: EmbeddingProvider[] = Object.keys(EMBEDDING_MODELS) as EmbeddingProvider[]
 
+export const autoDetectProviders: EmbeddingProvider[] = AUTO_DETECT_PROVIDER_ORDER.filter(
+  (provider): provider is EmbeddingProvider => provider in EMBEDDING_MODELS,
+);
+
 export type ProviderModels = {
   [P in keyof typeof EMBEDDING_MODELS]: keyof (typeof EMBEDDING_MODELS)[P]
 }

src/embeddings/detector.ts

@@ -1,4 +1,4 @@
-import { type EmbeddingProvider, type CustomProviderConfig, type BaseModelInfo, getDefaultModelForProvider, isValidModel, availableProviders, EmbeddingModelName, EMBEDDING_MODELS } from "../config";
+import { type EmbeddingProvider, type CustomProviderConfig, type BaseModelInfo, getDefaultModelForProvider, isValidModel, autoDetectProviders, EmbeddingModelName, EMBEDDING_MODELS } from "../config";
 import { existsSync, readFileSync } from "fs";
 import * as path from "path";
 import * as os from "os";
@@ -91,7 +91,7 @@ export async function detectEmbeddingProvider<P extends EmbeddingProvider>(
 }
 
 export async function tryDetectProvider(): Promise<ConfiguredProviderInfo> {
-  for (const provider of availableProviders) {
+  for (const provider of autoDetectProviders) {
     const credentials = await getProviderCredentials(provider);
     if (credentials) {
       return {
@@ -103,7 +103,7 @@ export async function tryDetectProvider(): Promise<ConfiguredProviderInfo> {
   }
 
   throw new Error(
-    `No embedding-capable provider found. Please authenticate with OpenCode using one of: ${availableProviders.join(", ")}.`
+    `No embedding-capable provider found. Please authenticate with OpenCode using one of: ${autoDetectProviders.join(", ")}.`
   );
 }
 

src/indexer/index.ts

@@ -128,6 +128,8 @@ export interface StatusResult {
   currentBranch: string;
   baseBranch: string;
   compatibility: IndexCompatibility | null;
+  failedBatchesCount: number;
+  failedBatchesPath?: string;
 }
 
 export interface IndexProgress {
@@ -1527,22 +1529,37 @@ export class Indexer {
   private saveFailedBatches(batches: FailedBatch[]): void {
     if (batches.length === 0) {
       if (existsSync(this.failedBatchesPath)) {
-        fsPromises.unlink(this.failedBatchesPath).catch(() => { });
+        try {
+          unlinkSync(this.failedBatchesPath);
+        } catch {
+          // Ignore cleanup failures; stale diagnostics are best-effort only.
+        }
       }
       return;
     }
     writeFileSync(this.failedBatchesPath, JSON.stringify(batches, null, 2));
   }
 
-  private addFailedBatch(batch: PendingChunk[], error: string): void {
-    const existing = this.loadFailedBatches();
-    existing.push({
-      chunks: batch,
-      error,
-      attemptCount: 1,
-      lastAttempt: new Date().toISOString(),
-    });
-    this.saveFailedBatches(existing);
+  private collectRetryableFailedChunks(
+    currentFileHashes: Map<string, string>,
+    unchangedFilePaths: Set<string>
+  ): PendingChunk[] {
+    const retryableById = new Map<string, PendingChunk>();
+
+    for (const batch of this.loadFailedBatches()) {
+      for (const chunk of batch.chunks) {
+        const filePath = chunk.metadata.filePath;
+        if (!currentFileHashes.has(filePath)) {
+          continue;
+        }
+        if (!unchangedFilePaths.has(filePath)) {
+          continue;
+        }
+        retryableById.set(chunk.id, chunk);
+      }
+    }
+
+    return Array.from(retryableById.values());
   }
 
   private getProviderRateLimits(provider: string): {
@@ -2087,6 +2104,7 @@ export class Indexer {
       skippedFiles: [],
       parseFailures: [],
     };
+    const failedBatchesForCurrentRun: FailedBatch[] = [];
 
     onProgress?.({
       phase: "scanning",
@@ -2245,6 +2263,18 @@ export class Indexer {
       }
     }
 
+    const retryableFailedChunks = this.collectRetryableFailedChunks(currentFileHashes, unchangedFilePaths);
+    if (retryableFailedChunks.length > 0) {
+      const pendingChunkIds = new Set(pendingChunks.map((chunk) => chunk.id));
+      for (const chunk of retryableFailedChunks) {
+        if (!pendingChunkIds.has(chunk.id)) {
+          pendingChunks.push(chunk);
+          pendingChunkIds.add(chunk.id);
+          currentChunkIds.add(chunk.id);
+        }
+      }
+    }
+
     if (chunkDataBatch.length > 0) {
       database.upsertChunksBatch(chunkDataBatch);
     }
@@ -2376,6 +2406,7 @@ export class Indexer {
       database.addSymbolsToBranchBatch(this.currentBranch, Array.from(allSymbolIds));
       this.fileHashCache = currentFileHashes;
       this.saveFileHashCache();
+      this.saveFailedBatches([]);
       stats.durationMs = Date.now() - startTime;
       onProgress?.({
         phase: "complete",
@@ -2397,6 +2428,7 @@ export class Indexer {
       invertedIndex.save();
       this.fileHashCache = currentFileHashes;
       this.saveFileHashCache();
+      this.saveFailedBatches([]);
       stats.durationMs = Date.now() - startTime;
       onProgress?.({
         phase: "complete",
@@ -2530,7 +2562,12 @@ export class Indexer {
           });
         } catch (error) {
           stats.failedChunks += batch.length;
-          this.addFailedBatch(batch, getErrorMessage(error));
+          failedBatchesForCurrentRun.push({
+            chunks: batch,
+            error: getErrorMessage(error),
+            attemptCount: 1,
+            lastAttempt: new Date().toISOString(),
+          });
           this.logger.recordEmbeddingError();
           this.logger.embedding("error", `Failed to embed batch after retries`, {
             batchSize: batch.length,
@@ -2541,6 +2578,7 @@ export class Indexer {
     }
 
     await queue.onIdle();
+    this.saveFailedBatches(failedBatchesForCurrentRun);
 
     onProgress?.({
       phase: "storing",
@@ -2962,6 +3000,7 @@ export class Indexer {
 
   async getStatus(): Promise<StatusResult> {
     const { store, configuredProviderInfo } = await this.ensureInitialized();
+    const failedBatchesCount = this.getFailedBatchesCount();
 
     return {
       indexed: store.count() > 0,
@@ -2972,6 +3011,8 @@ export class Indexer {
       currentBranch: this.currentBranch,
       baseBranch: this.baseBranch,
       compatibility: this.indexCompatibility,
+      failedBatchesCount,
+      failedBatchesPath: failedBatchesCount > 0 ? this.failedBatchesPath : undefined,
     };
   }
 

src/mcp-server.ts

@@ -1,9 +1,9 @@
 import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
 import { z } from "zod";
 
-import { Indexer, type IndexStats } from "./indexer/index.js";
+import { Indexer } from "./indexer/index.js";
 import type { ParsedCodebaseIndexConfig, LogLevel } from "./config/schema.js";
-import { formatDefinitionLookup } from "./tools/utils.js";
+import { formatDefinitionLookup, formatIndexStats, formatStatus } from "./tools/utils.js";
 import { formatCostEstimate } from "./utils/cost.js";
 import type { LogEntry } from "./utils/logger.js";
 
@@ -18,88 +18,6 @@ function truncateContent(content: string): string {
   );
 }
 
-function formatIndexStats(stats: IndexStats, verbose: boolean = false): string {
-  const lines: string[] = [];
-
-  if (stats.indexedChunks === 0 && stats.removedChunks === 0) {
-    lines.push(`Indexed. ${stats.totalFiles} files processed, ${stats.existingChunks} code chunks already up to date.`);
-  } else if (stats.indexedChunks === 0) {
-    lines.push(`Indexed. ${stats.totalFiles} files, removed ${stats.removedChunks} stale chunks, ${stats.existingChunks} chunks remain.`);
-  } else {
-    let main = `Indexed. ${stats.totalFiles} files processed, ${stats.indexedChunks} new chunks embedded.`;
-    if (stats.existingChunks > 0) {
-      main += ` ${stats.existingChunks} unchanged chunks skipped.`;
-    }
-    lines.push(main);
-
-    if (stats.removedChunks > 0) {
-      lines.push(`Removed ${stats.removedChunks} stale chunks.`);
-    }
-
-    if (stats.failedChunks > 0) {
-      lines.push(`Failed: ${stats.failedChunks} chunks.`);
-    }
-
-    lines.push(`Tokens: ${stats.tokensUsed.toLocaleString()}, Duration: ${(stats.durationMs / 1000).toFixed(1)}s`);
-  }
-
-  if (verbose) {
-    if (stats.skippedFiles.length > 0) {
-      const tooLarge = stats.skippedFiles.filter(f => f.reason === "too_large");
-      const excluded = stats.skippedFiles.filter(f => f.reason === "excluded");
-      const gitignored = stats.skippedFiles.filter(f => f.reason === "gitignore");
-
-      lines.push("");
-      lines.push(`Skipped files: ${stats.skippedFiles.length}`);
-      if (tooLarge.length > 0) {
-        lines.push(`  Too large (${tooLarge.length}): ${tooLarge.slice(0, 5).map(f => f.path).join(", ")}${tooLarge.length > 5 ? "..." : ""}`);
-      }
-      if (excluded.length > 0) {
-        lines.push(`  Excluded (${excluded.length}): ${excluded.slice(0, 5).map(f => f.path).join(", ")}${excluded.length > 5 ? "..." : ""}`);
-      }
-      if (gitignored.length > 0) {
-        lines.push(`  Gitignored (${gitignored.length}): ${gitignored.slice(0, 5).map(f => f.path).join(", ")}${gitignored.length > 5 ? "..." : ""}`);
-      }
-    }
-
-    if (stats.parseFailures.length > 0) {
-      lines.push("");
-      lines.push(`Files with no extractable chunks (${stats.parseFailures.length}): ${stats.parseFailures.slice(0, 10).join(", ")}${stats.parseFailures.length > 10 ? "..." : ""}`);
-    }
-  }
-
-  return lines.join("\n");
-}
-
-function formatStatus(status: {
-  indexed: boolean;
-  vectorCount: number;
-  provider: string;
-  model: string;
-  indexPath: string;
-  currentBranch: string;
-  baseBranch: string;
-}): string {
-  if (!status.indexed) {
-    return "Codebase is not indexed. Run index_codebase to create an index.";
-  }
-
-  const lines = [
-    `Index status:`,
-    `  Indexed chunks: ${status.vectorCount.toLocaleString()}`,
-    `  Provider: ${status.provider}`,
-    `  Model: ${status.model}`,
-    `  Location: ${status.indexPath}`,
-  ];
-
-  if (status.currentBranch !== "default") {
-    lines.push(`  Current branch: ${status.currentBranch}`);
-    lines.push(`  Base branch: ${status.baseBranch}`);
-  }
-
-  return lines.join("\n");
-}
-
 const CHUNK_TYPE_ENUM = [
   "function", "class", "method", "interface", "type",
   "enum", "struct", "impl", "trait", "module", "other",

src/tools/utils.ts

@@ -14,6 +14,14 @@ function truncateContent(content: string): string {
 
 export function formatIndexStats(stats: IndexStats, verbose: boolean = false): string {
   const lines: string[] = [];
+
+  if (stats.failedChunks > 0) {
+    lines.push(`INDEXING WARNING: ${stats.failedChunks} chunks failed to embed.`);
+    if (stats.failedBatchesPath) {
+      lines.push(`Inspect failed batches at: ${stats.failedBatchesPath}`);
+    }
+    lines.push("");
+  }
 
   if (stats.indexedChunks === 0 && stats.removedChunks === 0) {
     lines.push(`${stats.totalFiles} files processed, ${stats.existingChunks} code chunks already up to date.`);
@@ -67,6 +75,19 @@ export function formatIndexStats(stats: IndexStats, verbose: boolean = false): s
 
 export function formatStatus(status: StatusResult): string {
   if (!status.indexed) {
+    if (status.failedBatchesCount > 0) {
+      const lines = [
+        "Codebase is not indexed. The last indexing run left failed embedding batches.",
+        "Fix the provider/model configuration, then rerun index_codebase normally to retry the saved failed batches. Use force=true only for a full rebuild or compatibility reset.",
+      ];
+
+      if (status.failedBatchesPath) {
+        lines.push(`Failed batches: ${status.failedBatchesPath}`);
+      }
+
+      return lines.join("\n");
+    }
+
     return "Codebase is not indexed. Run index_codebase to create an index.";
   }
 
@@ -82,6 +103,14 @@ export function formatStatus(status: StatusResult): string {
     lines.push(`Base branch: ${status.baseBranch}`);
   }
 
+  if (status.failedBatchesCount > 0) {
+    lines.push("");
+    lines.push(`INDEXING WARNING: ${status.failedBatchesCount} failed embedding batch${status.failedBatchesCount === 1 ? " remains" : "es remain"}.`);
+    if (status.failedBatchesPath) {
+      lines.push(`Failed batches: ${status.failedBatchesPath}`);
+    }
+  }
+
   if (status.compatibility && !status.compatibility.compatible) {
     lines.push("");
     lines.push(`COMPATIBILITY WARNING: ${status.compatibility.reason}`);

tests/config.test.ts

@@ -7,6 +7,7 @@ import {
 } from "../src/config/schema.js";
 import {
   EMBEDDING_MODELS,
+  AUTO_DETECT_PROVIDER_ORDER,
   DEFAULT_PROVIDER_MODELS,
 } from "../src/config/constants.js";
 
@@ -836,8 +837,8 @@ describe("config schema", () => {
     it("should return correct model for google", () => {
       const model = getDefaultModelForProvider("google");
       expect(model.provider).toBe("google");
-      expect(model.model).toBe("text-embedding-005");
-      expect(model.dimensions).toBe(768);
+      expect(model.model).toBe("gemini-embedding-001");
+      expect(model.dimensions).toBe(1536);
     });
 
     it("should return correct model for ollama", () => {
@@ -937,5 +938,12 @@ describe("config schema", () => {
       const defaultProviders = Object.keys(DEFAULT_PROVIDER_MODELS);
       expect(defaultProviders.sort()).toEqual(providers.sort());
     });
+
+    it("should prefer GitHub Copilot before Google for auto-detection", () => {
+      expect(AUTO_DETECT_PROVIDER_ORDER[0]).toBe("github-copilot");
+      expect(AUTO_DETECT_PROVIDER_ORDER.indexOf("google")).toBeGreaterThan(
+        AUTO_DETECT_PROVIDER_ORDER.indexOf("github-copilot"),
+      );
+    });
   });
 });