feat(agentswarm): optional c0mpute.com inference backend

createC0mputeModel() turns a connected c0mpute.com account (apiKey + OpenAI-
compatible base URL) into a LangChain chat model, usable as the model for
createDeepAgentRunner or the chatModel for the judge/router — so agent inference
can run on community-shared GPUs. Opt-in; default stays the hosted provider.
agentswarm performs no OAuth: the host's c0mpute connector supplies credentials
(resolveC0mputeConnector / c0mputeConnectorFromEnv normalize them). DeepAgent
runner now accepts a model instance, not just a provider string.

7 new tests (defaults, env mapping, not-connected + missing-peer errors); 34/34 pass.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

packages/agentswarm/src/compute.ts

@@ -0,0 +1,78 @@
+/**
+ * Optional c0mpute.com backend — run agent/judge/router inference on
+ * community-shared GPUs instead of (or alongside) a hosted provider.
+ *
+ * The host app connects a user's c0mpute.com account via the c0mpute OAuth
+ * connector and obtains an API key + inference base URL; this module turns that
+ * into a LangChain chat model (c0mpute exposes an OpenAI-compatible endpoint).
+ * agentswarm itself performs no OAuth — it only consumes the resolved credentials.
+ */
+
+export const C0MPUTE_DEFAULT_BASE_URL = "https://api.c0mpute.com/v1";
+export const C0MPUTE_DEFAULT_MODEL = "default";
+
+/** Result of a user connecting their c0mpute.com account. */
+export interface C0mputeConnector {
+  /** API key issued after the user OAuths/connects their c0mpute.com account. */
+  apiKey: string;
+  /** OpenAI-compatible inference base URL. Defaults to the c0mpute public endpoint. */
+  baseUrl?: string;
+  /** Model id to request from the worker pool. */
+  model?: string;
+}
+
+export interface ResolvedC0mputeConnector {
+  apiKey: string;
+  baseUrl: string;
+  model: string;
+}
+
+/** Apply defaults and validate a connector; throws if it has not been connected. */
+export function resolveC0mputeConnector(connector: C0mputeConnector): ResolvedC0mputeConnector {
+  if (!connector.apiKey) {
+    throw new Error("c0mpute connector requires an apiKey — connect a c0mpute.com account first");
+  }
+  return {
+    apiKey: connector.apiKey,
+    baseUrl: connector.baseUrl ?? C0MPUTE_DEFAULT_BASE_URL,
+    model: connector.model ?? C0MPUTE_DEFAULT_MODEL
+  };
+}
+
+/** Build a connector from env vars (shares names with the c0mpute plugin). */
+export function c0mputeConnectorFromEnv(
+  env: Record<string, string | undefined> = process.env
+): C0mputeConnector {
+  return {
+    apiKey: env.C0MPUTE_API_KEY ?? "",
+    baseUrl: env.C0MPUTE_API_URL,
+    model: env.C0MPUTE_MODEL
+  };
+}
+
+/** Indirection so TS does not statically resolve the optional `@langchain/openai` peer. */
+async function loadOptionalModule(specifier: string): Promise<any> {
+  return import(specifier);
+}
+
+/**
+ * Build a LangChain chat model pointed at c0mpute.com. Pass the result as the
+ * `model` of {@link createDeepAgentRunner}, or as the `chatModel` of
+ * `createLLMJudge` / `createLLMRouter`, to run that inference on c0mpute GPUs.
+ *
+ * Optional peer: requires `@langchain/openai` in the host app.
+ */
+export async function createC0mputeModel(connector: C0mputeConnector): Promise<any> {
+  const resolved = resolveC0mputeConnector(connector);
+  let openai: any;
+  try {
+    openai = await loadOptionalModule("@langchain/openai");
+  } catch {
+    throw new Error("createC0mputeModel requires '@langchain/openai'. Install it: npm i @langchain/openai");
+  }
+  return new openai.ChatOpenAI({
+    apiKey: resolved.apiKey,
+    model: resolved.model,
+    configuration: { baseURL: resolved.baseUrl }
+  });
+}