{
  "name": "FluctlightDB",
  "description": "FluctlightDB is an embedded memory database for AI agents. experience() writes, activate() recalls from a cue, checkpoint() keeps it. Open source, beta.",
  "status": "beta",
  "license": "MIT or Apache-2.0",
  "install": "pip install \"fluctlightdb[native]\"",
  "verbs": [
    {
      "name": "experience()",
      "title": "Write what happened",
      "body": "A tool result, a user fact, an observation. It carries context, salience and provenance, not just text. Nothing is extracted from a transcript later; the agent records the moment it happens.",
      "source": "README.md:47-52"
    },
    {
      "name": "activate(cue)",
      "title": "Recall from a cue",
      "body": "Full-text search, vector neighbours and graph spreading activation are fused inside one call. The question is not \"what is nearest\" but \"what should this cue bring back\".",
      "source": "README.md:113"
    },
    {
      "name": "checkpoint()",
      "title": "Survive the restart",
      "body": "A write-ahead log and atomic checkpoints put the brain directory on disk. Crash recovery is exercised in CI with SIGKILL mid-write and torn WAL tails.",
      "source": "crates/fluctlightdb/src/wal.rs:1"
    }
  ],
  "integrations": [
    {
      "name": "LangChain",
      "detail": "FluctlightMemory, chat history"
    },
    {
      "name": "LlamaIndex",
      "detail": "FluctlightLlamaMemory"
    },
    {
      "name": "OpenAI Agents SDK",
      "detail": "remember and search handlers"
    },
    {
      "name": "MCP",
      "detail": "Cursor, Claude and Codex"
    },
    {
      "name": "Docker",
      "detail": "ghcr.io/voxmastery/fluctlightdb"
    },
    {
      "name": "Node",
      "detail": "over HTTP from fluctlight-serve"
    }
  ],
  "stable": [
    "connect_embedded(), connect_agent() and the other connect_* entry points",
    "experience(), activate(), checkpoint(), recall()",
    "The v4 brain directory on disk",
    "POST /api/v1/experience and /api/v1/activate"
  ],
  "experimental": [
    "The Recall Fabric behind FLUCTLIGHT_FABRIC=1, off by default",
    "Every neuroscience-named internal module",
    "Governance, snapshots and retention policy",
    "Auth and multi-tenancy — adversarially tested in CI, never third-party audited"
  ],
  "limits": [
    {
      "body": "One active maintainer. No corporate backing, no foundation governance, no co-maintainer yet.",
      "source": "MAINTAINER.md:9",
      "href": "https://github.com/voxmastery/FluctlightDB/blob/main/MAINTAINER.md"
    },
    {
      "body": "No third-party security audit and no penetration test to date. No published CVEs either.",
      "source": "SECURITY.md:7,25",
      "href": "https://github.com/voxmastery/FluctlightDB/blob/main/SECURITY.md"
    },
    {
      "body": "Production readiness defaults to false. Rate limits are best-effort per tenant; there is no WAF and no mTLS.",
      "source": "docs/PRODUCTION.md:5,114",
      "href": "https://github.com/voxmastery/FluctlightDB/blob/main/docs/REPRODUCIBILITY.md"
    },
    {
      "body": "Approximate nearest-neighbour search beyond 100,000 memories has not been evaluated.",
      "source": "papers/arxiv-v1/main.tex:703",
      "href": "https://arxiv.org/abs/2608.12365"
    },
    {
      "body": "When many agents share one brain, provenance separation reaches only 18% top-1. Per-case isolation reaches 100%, which is a ceiling rather than multi-tenant readiness. Unaddressed in code.",
      "source": "papers/arxiv-v1/main.tex:668-675",
      "href": "https://github.com/voxmastery/FluctlightDB/blob/main/docs/BENCHMARKS.md"
    },
    {
      "body": "Offline, recall is lexical: a cue has to share tokens with what was stored. Paraphrase recall needs a semantic vector from an embedder you bring.",
      "source": "docs/EMBEDDINGS.md:9,24",
      "href": "https://github.com/voxmastery/FluctlightDB/blob/main/docs/GETTING_STARTED.md"
    }
  ],
  "benchmarks": {
    "conditions": [
      "Maintainer-reported",
      "Frozen July 2026",
      "Harnesses open",
      "No independent reproduction yet"
    ],
    "figures": [
      {
        "value": "96.8%",
        "label": "LoCoMo evidence recall",
        "detail": "at k=150 · 10 conversations, 1,982 questions · MiniLM-384",
        "source": "benchmarks/results/locomo-invented-stack-engine-2026-07-13.json",
        "href": "https://github.com/voxmastery/FluctlightDB/blob/main/docs/BENCHMARKS.md"
      },
      {
        "value": "72.6%",
        "label": "LoCoMo evidence recall",
        "detail": "at k=5 — the operational number",
        "source": "benchmarks/results/locomo-invented-stack-engine-2026-07-13.json",
        "href": "https://github.com/voxmastery/FluctlightDB/blob/main/docs/BENCHMARKS.md"
      },
      {
        "value": "97.6%",
        "label": "LongMemEval-S session recall",
        "detail": "at k=8 · 488 of 500",
        "source": "benchmarks/results/longmemeval-colab-v2-full-2026-07-04.json",
        "href": "https://github.com/voxmastery/FluctlightDB/blob/main/docs/BENCHMARKS.md"
      },
      {
        "value": "97.4%",
        "label": "LongMemEval end-to-end QA",
        "detail": "487 of 500 · locked run, gpt-4o reader and judge",
        "source": "benchmarks/results/e2e-cert-paper-v2-2026-07-07.json",
        "href": "https://github.com/voxmastery/FluctlightDB/blob/main/docs/BENCHMARKS.md"
      },
      {
        "value": "0.646",
        "label": "BEIR SciFact nDCG@10",
        "detail": "against Chroma 0.645 in a shared harness",
        "source": "benchmarks/results/paper-2026-07-09.json",
        "href": "https://github.com/voxmastery/FluctlightDB/blob/main/docs/BENCHMARKS.md"
      }
    ],
    "retraction": {
      "status": "retracted",
      "struck": "99.0%",
      "label": "LoCoMo evidence recall at k=150",
      "body": "The old figure expanded every retrieved turn by three neighbours on each side, then counted neighbours the engine never retrieved. A plain BM25 baseline also reaches about 99% under that protocol, so it distinguished nothing. It is not the headline any more, and it is not a number we will defend.",
      "source": "README.md:70 · CHANGELOG.md:49",
      "href": "https://github.com/voxmastery/FluctlightDB/blob/main/docs/BENCHMARKS.md",
      "note": "Retracted. Not a current claim."
    }
  },
  "links": {
    "github": "https://github.com/voxmastery/FluctlightDB",
    "pypi": "https://pypi.org/project/fluctlightdb/",
    "paper": "https://arxiv.org/abs/2608.12365",
    "doi": "https://doi.org/10.5281/zenodo.20949890"
  },
  "citation": "Ganesh S, “FluctlightDB: A Memory Model of Data for AI Agents”, 2026."
}