everos-memory

{
  "name": "everos-memory",
  "version": "0.1.0",
  "description": "Automatic long-term memory for Claude Code, backed by a self-hosted EverOS server (recall on prompt submit, deferred capture on turn stop, batched extraction on flush).",
  "author": {
    "name": "EverMind AI (EverOS)"
  },
  "keywords": [
    "memory",
    "context",
    "recall",
    "everos"
  ],
  "userConfig": {
    "baseUrl": {
      "type": "string",
      "title": "EverOS base URL",
      "description": "EverOS server base URL",
      "default": "http://127.0.0.1:8124"
    },
    "apiVersion": {
      "type": "string",
      "title": "API version",
      "description": "EverOS API version to use: auto, v1, or v2",
      "default": "auto"
    },
    "appId": {
      "type": "string",
      "title": "App ID",
      "description": "EverOS application partition id",
      "default": "claude-code"
    },
    "projectId": {
      "type": "string",
      "title": "Project ID override",
      "description": "Fixed EverOS project partition id (defaults to a hash of the working directory if left empty)",
      "default": ""
    },
    "userId": {
      "type": "string",
      "title": "User ID override",
      "description": "Developer identity for the EverOS user track (defaults to the OS account name if left empty)",
      "default": ""
    },
    "agentId": {
      "type": "string",
      "title": "Agent ID",
      "description": "Agent identity for the EverOS agent track",
      "default": "claude-code"
    },
    "recallMethod": {
      "type": "string",
      "title": "Recall method",
      "description": "EverOS retrieval method: keyword, vector, hybrid, or agentic",
      "default": "hybrid"
    },
    "enableLlmRerank": {
      "type": "boolean",
      "title": "Enable LLM rerank",
      "description": "Route hybrid search through the LLM rerank lane instead of requiring a cross-encoder rerank provider",
      "default": true
    },
    "queryN": {
      "type": "number",
      "title": "Recall query history depth",
      "description": "Number of recent direct user messages blended into each recall query",
      "default": 3
    },
    "queryMaxChars": {
      "type": "number",
      "title": "Recall query character budget",
      "description": "Character budget for the recall query",
      "default": 2000
    },
    "recallTopK": {
      "type": "number",
      "title": "Recall top-K",
      "description": "Maximum results requested per EverOS owner track",
      "default": 2
    },
    "recallEpisodeTopK": {
      "type": "number",
      "title": "Episode recall top-K",
      "description": "Maximum main episode hits; each hit may load its immediate previous and next episode",
      "default": 2
    },
    "recallEpisodeMinScore": {
      "type": "number",
      "title": "Episode recall minimum score",
      "description": "Minimum calibrated score for injected episode recall results; surviving results are capped by Recall top-K",
      "default": 0.85
    },
    "recallCaseMinScore": {
      "type": "number",
      "title": "Case recall minimum score",
      "description": "Minimum calibrated score for injected agent case recall results; surviving results are capped by Recall top-K",
      "default": 0.6
    },
    "recallMaxChars": {
      "type": "number",
      "title": "Recall injection byte budget",
      "description": "Maximum UTF-8 bytes injected by recalled memory; complete items that do not fit are skipped",
      "default": 9500
    },
    "recallTimeoutMs": {
      "type": "number",
      "title": "Recall timeout (ms)",
      "description": "Per-attempt timeout for the primary recall method (hybrid+rerank makes a real LLM call; budget accordingly)",
      "default": 20000
    },
    "recallRetries": {
      "type": "number",
      "title": "Recall retry attempts",
      "description": "How many times to try the primary recall method before falling back",
      "default": 2
    },
    "recallFallbackMethod": {
      "type": "string",
      "title": "Recall fallback method",
      "description": "Method used if the primary recall method fails/times out on every retry (keyword never depends on rerank)",
      "default": "keyword"
    },
    "recallFallbackTimeoutMs": {
      "type": "number",
      "title": "Recall fallback timeout (ms)",
      "description": "Timeout for the fallback recall call",
      "default": 5000
    },
    "episodeSynthesisEnabled": {
      "type": "boolean",
      "title": "Enable episode/case synthesis",
      "description": "Evaluate recalled episode chains and compact recalled agent cases before injection",
      "default": true
    },
    "episodeSynthesisBaseUrl": {
      "type": "string",
      "title": "Memory synthesis API base URL",
      "description": "OpenAI-compatible API base URL; falls back to EVEROS_LLM__BASE_URL",
      "default": ""
    },
    "episodeSynthesisApiKey": {
      "type": "string",
      "title": "Memory synthesis API key",
      "description": "OpenAI-compatible API key; falls back to EVEROS_LLM__API_KEY",
      "default": ""
    },
    "episodeSynthesisModel": {
      "type": "string",
      "title": "Memory synthesis model",
      "description": "Model used for episode evaluation, same-task comparison, and case compression; falls back to EVEROS_LLM__MODEL",
      "default": ""
    },
    "episodeSynthesisTimeoutMs": {
      "type": "number",
      "title": "Episode synthesis timeout (ms)",
      "description": "Per-call timeout; the three calls share wall-clock time because they run concurrently",
      "default": 30000
    },
    "episodeSynthesisMaxTokens": {
      "type": "number",
      "title": "Episode synthesis output token limit",
      "description": "Maximum output tokens for each episode synthesis call",
      "default": 2000
    },
    "captureTimeoutMs": {
      "type": "number",
      "title": "Capture/flush timeout (ms)",
      "description": "Per-add/flush timeout in milliseconds",
      "default": 15000
    },
    "captureMaxChars": {
      "type": "number",
      "title": "Capture per-message character budget",
      "description": "Per-message text budget before capture",
      "default": 50000
    },
    "flushIdleMs": {
      "type": "number",
      "title": "Idle flush delay (ms)",
      "description": "Flush after this much inactivity following a captured turn",
      "default": 30000
    },
    "flushTokenThreshold": {
      "type": "number",
      "title": "Flush token threshold",
      "description": "Flush when the approximate buffered token count reaches this threshold",
      "default": 12000
    },
    "flushMessageThreshold": {
      "type": "number",
      "title": "Flush message threshold",
      "description": "Flush when the buffered message count reaches this threshold",
      "default": 50
    },
    "flushMaxDelayMs": {
      "type": "number",
      "title": "Max flush delay (ms)",
      "description": "Maximum time a non-empty buffer may remain unflushed",
      "default": 300000
    },
    "flushOnSessionSwitch": {
      "type": "boolean",
      "title": "Flush on session switch",
      "description": "Flush other pending sessions in the same project before a new session recalls",
      "default": true
    }
  }
}