feat: add Voyage AI embedding provider — closes #1

src/cli/init.ts

@@ -92,7 +92,7 @@ async function main(): Promise<void> {
 
   if (profile === "full") {
     console.log("\nFull profile uses embeddings for vector search.");
-    const provider = await ask("Embedding provider [local/openai] (default: local): ");
+    const provider = await ask("Embedding provider [local/openai/voyage] (default: local): ");
     if (provider.trim().toLowerCase() === "openai") {
       const key = await ask("OpenAI API key (or set OPENAI_API_KEY later): ");
       if (key.trim()) {

src/lib/embeddings.ts

@@ -26,9 +26,9 @@
 
 // --- Local Provider (Xenova/transformers) ---
 
 let extractor: any = null;
 
 async function getExtractor(): Promise<any> {
   if (!extractor) {
     extractor = await pipeline("feature-extraction", "Xenova/all-MiniLM-L6-v2");
   }
@@ -92,7 +92,7 @@
 
       const data = await resp.json();
       // Sort by index to preserve order
       const sorted = data.data.sort((a: any, b: any) => a.index - b.index);
       for (const item of sorted) {
         results.push(item.embedding);
       }
@@ -102,17 +102,77 @@
   }
 }
 
+// --- Voyage AI Provider ---
+
+class VoyageEmbeddingProvider implements EmbeddingProvider {
+  dimensions = 1024;
+  private apiKey: string;
+  private model: string;
+
+  constructor(apiKey: string, model = "voyage-3") {
+    this.apiKey = apiKey;
+    this.model = model;
+    // voyage-3 outputs 1024 dims, voyage-3-lite outputs 512
+    if (model === "voyage-3-lite") this.dimensions = 512;
+  }
+
+  async embed(text: string): Promise<number[]> {
+    const [result] = await this.embedBatch([text]);
+    return result;
+  }
+
+  async embedBatch(texts: string[]): Promise<number[][]> {
+    const results: number[][] = [];
+    const processed = texts.map(preprocessText);
+
+    // Voyage supports up to 128 texts per batch
+    for (let i = 0; i < processed.length; i += 128) {
+      const batch = processed.slice(i, i + 128);
+      const resp = await fetch("https://api.voyageai.com/v1/embeddings", {
+        method: "POST",
+        headers: {
+          Authorization: `Bearer ${this.apiKey}`,
+          "Content-Type": "application/json",
+        },
+        body: JSON.stringify({
+          model: this.model,
+          input: batch,
+          input_type: "document",
+        }),
+      });
+
+      if (!resp.ok) {
+        const err = await resp.text();
+        throw new Error(`Voyage AI embeddings API error ${resp.status}: ${err}`);
+      }
+
+      const data = await resp.json();
+      const sorted = data.data.sort((a: any, b: any) => a.index - b.index);
+      for (const item of sorted) {
+        results.push(item.embedding);
+      }
+    }
+
+    return results;
+  }
+}
+
 // --- Factory ---
 
 export interface EmbeddingConfig {
-  provider: "local" | "openai";
+  provider: "local" | "openai" | "voyage";
   apiKey?: string;
+  model?: string;
 }
 
 export function createEmbeddingProvider(config: EmbeddingConfig): EmbeddingProvider {
   if (config.provider === "openai") {
     if (!config.apiKey) throw new Error("OpenAI API key required for openai embedding provider");
     return new OpenAIEmbeddingProvider(config.apiKey);
   }
+  if (config.provider === "voyage") {
+    if (!config.apiKey) throw new Error("Voyage AI API key required for voyage embedding provider");
+    return new VoyageEmbeddingProvider(config.apiKey, config.model);
+  }
   return new LocalEmbeddingProvider();
 }

src/lib/timeline-db.ts

@@ -55,7 +55,7 @@ export interface ProjectInfo {
 }
 
 export interface TimelineConfig {
-  embedding_provider: "local" | "openai";
+  embedding_provider: "local" | "openai" | "voyage";
   embedding_model: string;
   openai_api_key?: string;
   indexed_projects: Record<string, {

src/tools/onboard-project.ts

@@ -30,7 +30,7 @@ export function registerOnboardProject(server: McpServer) {
     "Index a project's Claude Code sessions and git history into the timeline database for semantic search and chronological viewing.",
     {
       project_dir: z.string().describe("Absolute path to the project directory"),
-      embedding_provider: z.enum(["local", "openai"]).default("local"),
+      embedding_provider: z.enum(["local", "openai", "voyage"]).default("local"),
       openai_api_key: z.string().optional(),
       git_depth: z.enum(["all", "6months", "1year", "3months"]).default("all"),
       git_since: z.string().optional().describe("Override git_depth with exact start date (ISO: '2025-08-01')"),

tests/lib/embeddings.test.ts

@@ -63,4 +63,27 @@ describe("createEmbeddingProvider", () => {
     });
     expect(provider.dimensions).toBe(1536);
   });
+
+  it("throws when voyage provider has no API key", () => {
+    expect(() =>
+      createEmbeddingProvider({ provider: "voyage" }),
+    ).toThrow("API key required");
+  });
+
+  it("returns voyage provider with 1024 dimensions (voyage-3)", () => {
+    const provider = createEmbeddingProvider({
+      provider: "voyage",
+      apiKey: "pa-test-key",
+    });
+    expect(provider.dimensions).toBe(1024);
+  });
+
+  it("returns voyage provider with 512 dimensions for voyage-3-lite", () => {
+    const provider = createEmbeddingProvider({
+      provider: "voyage",
+      apiKey: "pa-test-key",
+      model: "voyage-3-lite",
+    });
+    expect(provider.dimensions).toBe(512);
+  });
 });