opencodekit 0.13.1 → 0.14.0

package/dist/template/.opencode/tool/memory-embed.ts

@@ -0,0 +1,183 @@
+import { tool } from "@opencode-ai/plugin";
+
+// Configuration - Qwen3-Embedding-0.6B
+// Better for code, multilingual (100+ langs), instruction-aware
+// See: https://ollama.com/library/qwen3-embedding
+const OLLAMA_MODEL = "qwen3-embedding:0.6b";
+const OLLAMA_DIMENSIONS = 1024;
+const OLLAMA_BASE_URL = process.env.OLLAMA_HOST || "http://127.0.0.1:11434";
+
+interface EmbeddingResult {
+	text: string;
+	embedding: number[];
+	model: string;
+	dimensions: number;
+}
+
+interface BatchEmbeddingResult {
+	results: EmbeddingResult[];
+	failed: { text: string; error: string }[];
+	model: string;
+}
+
+async function checkOllamaRunning(): Promise<boolean> {
+	try {
+		const response = await fetch(`${OLLAMA_BASE_URL}/api/version`, {
+			method: "GET",
+			signal: AbortSignal.timeout(3000),
+		});
+		return response.ok;
+	} catch {
+		return false;
+	}
+}
+
+async function ensureModelAvailable(): Promise<{
+	ok: boolean;
+	error?: string;
+}> {
+	try {
+		const listResponse = await fetch(`${OLLAMA_BASE_URL}/api/tags`, {
+			method: "GET",
+			signal: AbortSignal.timeout(5000),
+		});
+
+		if (!listResponse.ok) {
+			return { ok: false, error: "Failed to list models" };
+		}
+
+		const data = (await listResponse.json()) as {
+			models?: { name: string }[];
+		};
+		const models = data.models || [];
+		const modelExists = models.some(
+			(m) => m.name === OLLAMA_MODEL || m.name === `${OLLAMA_MODEL}:latest`,
+		);
+
+		if (modelExists) {
+			return { ok: true };
+		}
+
+		return {
+			ok: false,
+			error: `Model '${OLLAMA_MODEL}' not found. Run: ollama pull ${OLLAMA_MODEL}`,
+		};
+	} catch (err) {
+		const message = err instanceof Error ? err.message : String(err);
+		return { ok: false, error: `Failed to check models: ${message}` };
+	}
+}
+
+async function embedWithOllama(
+	texts: string[],
+): Promise<{ embeddings: number[][]; error?: string }> {
+	try {
+		const embeddings: number[][] = [];
+
+		for (const text of texts) {
+			const response = await fetch(`${OLLAMA_BASE_URL}/api/embeddings`, {
+				method: "POST",
+				headers: { "Content-Type": "application/json" },
+				body: JSON.stringify({
+					model: OLLAMA_MODEL,
+					prompt: text,
+				}),
+				signal: AbortSignal.timeout(30000),
+			});
+
+			if (!response.ok) {
+				const errorText = await response.text();
+				return {
+					embeddings: [],
+					error: `Ollama API error (${response.status}): ${errorText}`,
+				};
+			}
+
+			const data = (await response.json()) as { embedding?: number[] };
+			if (!data.embedding) {
+				return { embeddings: [], error: "No embedding in response" };
+			}
+
+			embeddings.push(data.embedding);
+		}
+
+		return { embeddings };
+	} catch (err) {
+		const message = err instanceof Error ? err.message : String(err);
+
+		if (message.includes("ECONNREFUSED") || message.includes("fetch failed")) {
+			return {
+				embeddings: [],
+				error: "Cannot connect to Ollama. See .opencode/README.md for setup.",
+			};
+		}
+
+		return { embeddings: [], error: `Ollama error: ${message}` };
+	}
+}
+
+export default tool({
+	description: `Generate embeddings using Ollama (${OLLAMA_MODEL}). Requires Ollama running locally.`,
+	args: {
+		text: tool.schema
+			.union([tool.schema.string(), tool.schema.array(tool.schema.string())])
+			.describe("Text or array of texts to embed"),
+	},
+	execute: async (args: { text: string | string[] }) => {
+		const texts = Array.isArray(args.text) ? args.text : [args.text];
+
+		if (texts.length === 0) {
+			return "Error: No text provided";
+		}
+
+		const maxChars = 32000;
+		const longTexts = texts.filter((t) => t.length > maxChars);
+		if (longTexts.length > 0) {
+			return `Error: ${longTexts.length} text(s) exceed ${maxChars} char limit`;
+		}
+
+		const ollamaRunning = await checkOllamaRunning();
+		if (!ollamaRunning) {
+			return "Error: Ollama not running. See .opencode/README.md for setup.";
+		}
+
+		const modelCheck = await ensureModelAvailable();
+		if (!modelCheck.ok) {
+			return `Error: ${modelCheck.error}. See .opencode/README.md for setup.`;
+		}
+
+		const result = await embedWithOllama(texts);
+
+		if (result.error) {
+			return `Error: ${result.error}`;
+		}
+
+		const response: BatchEmbeddingResult = {
+			results: texts.map((text, i) => ({
+				text: text.substring(0, 100) + (text.length > 100 ? "..." : ""),
+				embedding: result.embeddings[i],
+				model: OLLAMA_MODEL,
+				dimensions: OLLAMA_DIMENSIONS,
+			})),
+			failed: [],
+			model: OLLAMA_MODEL,
+		};
+
+		return JSON.stringify(response, null, 2);
+	},
+});
+
+// Export for other tools (memory-index, observation)
+export async function generateEmbedding(
+	text: string,
+): Promise<{ embedding: number[]; model: string } | null> {
+	const ollamaRunning = await checkOllamaRunning();
+	if (!ollamaRunning) return null;
+
+	const result = await embedWithOllama([text]);
+	if (result.error || result.embeddings.length === 0) return null;
+
+	return { embedding: result.embeddings[0], model: OLLAMA_MODEL };
+}
+
+export const EMBEDDING_DIMENSIONS = OLLAMA_DIMENSIONS;