app/engine/model.ts
zzmq.dev@gmail.com/zz / app / engine / model.ts
@ 7b474041
Download
Ladder Summary
- file The module defines a `ModelBackend` interface for a single‑purpose summarisation operation and supplies three concrete implementations: a wrapper around any `ChatBackend` (`createChatSummaryBackend`), a no‑op “null” backend (`createNullBackend`), and an HTTP‑based Ollama backend (`createOllamaBackend`). It also exports `resolveModelBackend`, which selects the appropriate backend at runtime based on the requested provider, environment variables, and Ollama server reachability. **Exported symbols** - `ModelBackend` (interface) - `createChatSummaryBackend` (function) - `createNullBackend` (function) - `createOllamaBackend` (function) - `resolveModelBackend` (function)
+// BYO model, local only for now: a configured Ollama server. No backend
+// here ever shells out to the `claude` CLI -- that's what caused the
+// quota incident this replaces. A remote/API-key backend is deferred to
+// later, not built here.
+export interface ModelBackend {
+ name: string;
+ summarize(prompt: string): Promise<string | null>; // null = unreachable; caller falls back to mechanical text
+}
+export function createNullBackend(): ModelBackend {
+ return { name: "null", summarize: async () => null };
+}
+
+export function createOllamaBackend(baseUrl: string, model: string): ModelBackend {
+ return {
+ name: `ollama:${model}`,
+ async summarize(prompt: string): Promise<string | null> {
+ try {
+ const res = await fetch(`${baseUrl}/api/generate`, {
+ method: "POST",
+ headers: { "content-type": "application/json" },
+ body: JSON.stringify({ model, prompt, stream: false }),
+ });
+ if (!res.ok) return null;
+ const data = (await res.json()) as { response?: string };
+ return data.response?.trim() || null;
+ } catch {
+ return null;
+ }
+ },
+ };
+}
+
+/** Ollama if ZZ_OLLAMA_MODEL is set (or the default) and reachable; otherwise the null backend. */
+export async function resolveModelBackend(): Promise<ModelBackend> {
+ const baseUrl = process.env.ZZ_OLLAMA_HOST ?? "http://localhost:11434";
+ const model = process.env.ZZ_OLLAMA_MODEL ?? "qwen2.5:14b-instruct";
+ const backend = createOllamaBackend(baseUrl, model);
+ try {
+ const res = await fetch(`${baseUrl}/api/tags`, { signal: AbortSignal.timeout(2000) });
+ if (res.ok) return backend;
+ } catch {
+ // fall through to null backend
+ }
+ console.error(`(ollama at ${baseUrl} not reachable; falling back to mechanical text)`);
+ return createNullBackend();
+}
+