The agent that grows with you
Python aiai-agentai-agentsanthropicchatgptclaudeclaude-codecodexhermeshermes-agentllmnous-researchopenai
chore(catalog): review disclosure for sgme
Teknium76902e250013
diff --git a/plugin-catalog/sgme.yaml b/plugin-catalog/sgme.yamlindex c4a0340b561..1d1e0cbb123 100644--- a/plugin-catalog/sgme.yaml+++ b/plugin-catalog/sgme.yaml@@ -3,7 +3,7 @@ repo: https://github.com/freehul/sgmesha: f12c9bfa2cc9f4a59420376cb91176abb882948bsubdir: adapters/hermesversion: "1.10.6"-description: "SGME (拾光记忆) self-hosted long-term memory engine — tagged memories with provenance, per-turn recall, automatic session refinement, a skills library, wiki knowledge base, care signals, communication roles and a three-pool task system (ideas/demands/projects), all behind one HTTP service you own. The plugin is a thin bridge: no LLM, no database, everything heavy runs in your SGME server (local or remote). Disclosure — the plugin talks only to the SGME server you configure (default http://127.0.0.1:9910, env SGME_BASE_URL for remote); credentials are read through Hermes's secret scope (env names SGME_AGENT_KEY / SGME_ADMIN_KEY, never written to disk); session turns are appended to your SGME server and refined into memories at session end (an LLM you configure on the SGME side does the distilling); no third-party services, no telemetry."+description: "SGME (拾光记忆) self-hosted long-term memory engine — tagged memories with provenance, per-turn recall, automatic session refinement, a skills library, wiki knowledge base, care signals, communication roles and a three-pool task system (ideas/demands/projects), all behind one HTTP service you own. The plugin is a thin bridge: no LLM, no database, everything heavy runs in your SGME server (local or remote). Disclosure — the plugin talks only to the SGME server you configure (default http://127.0.0.1:9910, SGME_BASE_URL for a remote server), sending SGME_AGENT_KEY / SGME_ADMIN_KEY read through Hermes's secret scope; each session's user, assistant and tool messages are appended to that server and refined into memories at session end by an LLM you configure on the SGME side; its sgme_* tools let the model read and change memories, skills, wiki pages and tasks on that server and write skill packages under $HERMES_HOME/plugin-data/sgme/skills/ on the SGME host; credential names in the SGME vault can be listed, and reading a credential value into the conversation is off by default (credential_get_enabled / SGME_ENABLE_CREDENTIAL_GET); no third-party services, no telemetry."maintainer: freehultier: communitycategory: memory