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
import { createHash } from "node:crypto";
import { deriveHarnessSessionId } from "./shared/session-model.mjs";
// Keep identifiers valid as a single path component on Windows. The `dsh-`
// prefix also prevents reserved device names such as CON and NUL from matching
// the whole component.
const WINDOWS_UNSAFE = /[<>:"/\\|?*\u0000-\u001F]/g;
const WINDOWS_UNSAFE_TRAILING = /[. ]+$/g;
export function deriveDshSessionId(sessionId) {
const original = deriveHarnessSessionId("dsh-", String(sessionId));
const portable = original
.replace(WINDOWS_UNSAFE, "_")
.replace(WINDOWS_UNSAFE_TRAILING, match => "_".repeat(match.length));
if (portable === original) return original;
// Replacement alone is lossy (`im:a` and `im?a` both become `im_a`). Keep a
// readable stem and add a stable digest whenever normalization was needed.
const digest = createHash("sha256").update(original).digest("hex").slice(0, 12);
return `${portable}__${digest}`;
}