Config Schema

Complete configuration reference for .logician.json.

The current flat runtime keys include:

{
  "baseUrl": "http://127.0.0.1:8080",
  "model": "model-id",
  "memory": false,
  "memoryEmbeddings": false,
  "memoryEmbeddingModel": "Xenova/all-MiniLM-L6-v2",
  "memoryExtractor": {
    "baseUrl": "http://127.0.0.1:8081",
    "model": "small-model-id"
  },
  "memoryViewer": true,
  "memoryViewerPort": 3200,
  "reasoner": "none",
  "reasonerConfig": {}
}

reasoner defaults to none. A selected reasoner runs before the ordinary tool-capable agent loop and its reasonerConfig values override that reasoner's registry defaults.

memoryEmbeddings defaults to false. When enabled, Logician lazily loads the configured local embedding model and fuses its semantic results with SQLite FTS memory retrieval. Lexical retrieval remains available while the model warms.

Full schema

{
  "llm": {
    "url": "string",           // Required: LLM API endpoint
    "model": "string",         // Required: Model name
    "apiKey": "string",        // Optional: API key
    "maxTokens": "number",     // Optional: Default 8192
    "temperature": "number",   // Optional: Default 0.7
    "topP": "number",          // Optional: Default 1.0
    "stream": "boolean"        // Optional: Default true
  },
  "permissions": "plan|ask|acceptEdits|acceptAll",
  "thinkingLevel": "low|medium|high|full",
  "reasoning": {
    "mode": "string",          // Reasoner: cot, reflexion, tot, etc.
    "maxIterations": "number"  // Default 10
  },
  "compaction": {
    "enabled": "boolean",      // Default true
    "triggerFraction": "number", // Default 0.75
    "strategy": "summarize"    // Summarization strategy
  },
  "mcp": {
    "servers": {
      "name": {
        "command": "string",
        "args": "string[]",
        "env": "Record<string, string>",
        "url": "string",       // For HTTP transport
        "transport": "stdio|http"
      }
    }
  },
  "plugins": ["string"],       // Plugin file paths
  "tools": {
    "allowed": ["string"],     // Tool name allowlist (empty = all)
    "denied": ["string"]      // Tool name denylist
  },
  "access": {
    "allowedPaths": ["string"], // Directories the agent can access
    "deniedPaths": ["string"]   // Directories the agent cannot access
  },
  "headless": {
    "timeout": "number",       // Default 300
    "maxIterations": "number", // Default 20
    "streamOutput": "boolean"  // Default true
  }
}

Environment variable overrides

All config options can be overridden via environment variables:

Config path	Environment variable
llm.url	LOGICIAN_LLM_URL
llm.model	LOGICIAN_LLM_MODEL
llm.apiKey	LOGICIAN_LLM_API_KEY
llm.maxTokens	LOGICIAN_MAX_TOKENS
llm.temperature	LOGICIAN_TEMPERATURE
permissions	LOGICIAN_PERMISSIONS
thinkingLevel	LOGICIAN_THINKING