Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.
Python agent-contextagent-memoryagent-pluginsagentic-ragcontext-databasedsh-pluginknowledge-basememorymemory-managementself-evolvingvector-database
/**
* One deadline for every request a hook sends.
*
* A hook's requests run one after another — recall falls back through three
* endpoints, capture sends then commits — so per-request timeouts alone add up
* past the timeout the host gives the hook, and the host kills it mid-write.
* Each request still gets its own timeout, or `defaultTimeoutMs`, but never
* more than what is left of `budgetMs`.
*/
export function withRequestBudget(fetchJSON, budgetMs, defaultTimeoutMs = 0) {
if (!budgetMs) return fetchJSON;
const deadline = Date.now() + budgetMs;
return (path, init = {}, options = {}) => {
const remaining = deadline - Date.now();
// ov-http deliberately clamps individual requests to one second. Do not
// start one inside that final second or the host-level total can overrun.
if (remaining < 1000) {
return Promise.resolve({
ok: false,
status: 0,
result: null,
error: { name: "AbortError", aborted: true, message: "hook request budget exhausted" },
});
}
const requested = Number(options.timeoutMs) || Number(defaultTimeoutMs) || remaining;
return fetchJSON(path, init, { ...options, timeoutMs: Math.min(requested, remaining) });
};
}