Lite and Full modes

\n\nEverything runs on your machine. Choose a tier during `neurostack init`:\n\n- **Lite** (~130 MB) — keyword search, link-based connections between notes, stale detection, MCP server. No GPU or Ollama required.\n- **Full** (~560 MB) — adds semantic search (finds notes by meaning, not just keywords), AI-generated summaries, connections between notes, and topic clustering via local [Ollama](https://ollama.ai). GPU or 6+ core CPU recommended.\n\nNon-interactive setup:\n\n```bash\nneurostack init --mode lite ~/my-notes # lite mode\nneurostack init --mode full ~/my-notes # full mode\n```\n\n

\n\n

Alternative install methods (PyPI, pip, curl)

\n\n```bash\n# PyPI\npipx install neurostack\npip install neurostack # inside a venv\nuv tool install neurostack\n\n# One-line script\ncurl -fsSL https://raw.githubusercontent.com/raphasouthall/neurostack/main/install.sh | bash\n\n# Lite mode (no ML deps)\ncurl -fsSL https://raw.githubusercontent.com/raphasouthall/neurostack/main/install.sh | NEUROSTACK_MODE=lite bash\n```\n\nOn Ubuntu 23.04+, Debian 12+, and Fedora 38+, bare `pip install` outside a virtual environment is blocked by the operating system. Use `npm`, `pipx`, or `uv tool install` instead.\n\nTo uninstall: `neurostack uninstall`\n\n

\n\n---\n\n## What it actually feels like\n\n**The researcher.** You ask Claude to help write the methodology section. Instead of starting from scratch, it already knows you've read 50 papers on complementary learning systems, that you settled on a particular framing in January, and that the meta-analysis you were relying on has been flagged as stale — it keeps appearing in searches where it no longer fits. You check it, update the note, and the AI's next answer reflects where your thinking actually is.\n\n**The writer.** You ask Cursor to help with chapter eleven. It knows Elena's backstory from chapter two, the decision you made in your world-building notes to keep magic systems implicit, and that you changed her last name in a revision three weeks ago. No contradictions.\n\n**The DevOps engineer.** You ask about the deployment runbook for the auth service. NeuroStack surfaces it — but also flags it as stale. You check it. The endpoint was renamed six weeks ago. You fix the note. The next time anyone asks, they get the right answer.\n\n**The student.** You're revising three weeks before exams. You ask your AI what's on the syllabus for Module 4. It searches your notes — and stale detection tells you two of the topics were in last year's module structure, which you replaced when the course was restructured. You know what to revise. You don't waste time on dropped content.\n\n**The data scientist.** You ask about the hyperparameters from your best experiment. NeuroStack returns the results from the rerun, not the original — because you updated that note, and the update is reflected in the index.\n\n---\n\n## What makes it different\n\nNeuroStack is not a replacement for Obsidian, Notion, or any note-taking app. It sits on top of what you already use and adds what they don't have.\n\n| Capability | Note apps | Basic RAG | NeuroStack |\n|-----------|-----------|-----------|------------|\n| Stores your notes | Yes | No | No (read-only by default; opt-in git-backed write tools) |\n| AI can search your notes | Some | Yes | Yes |\n| Detects stale/outdated notes | No | No | Yes |\n| AI memories persist across sessions | No | No | Yes |\n| Works with any MCP-compatible AI | No | Varies | Yes |\n| Tiered retrieval (saves 80-95% tokens) | No | No | Yes |\n| Profession-specific workflows | No | No | Yes |\n| Open source, self-hostable | Varies | Varies | Yes (Apache 2.0) |\n\nStale detection is the feature no other tool offers. When a note keeps appearing in contexts where it no longer fits — a deprecated API, a reversed decision, a superseded paper — NeuroStack flags it and demotes it in future results. Without this, your AI confidently cites information that is no longer true.\n\n---\n\n## Profession packs\n\nWhen you run `neurostack init`, you choose a profession pack. Each one configures NeuroStack with templates, folder structures, and AI guidance suited to how your profession actually uses notes.\n\n| Pack | Built for |\n|------|-----------|\n| `researcher` | Literature review, citation tracking, evolving arguments, stale paper detection |\n| `writer` | Character sheets, world-building, chapter outlines, continuity tracking |\n| `student` | Course notes, spaced repetition, exam prep, syllabus change detection |\n| `developer` | Code decisions, architecture notes, runbooks, deprecated API detection |\n| `devops` | Infrastructure runbooks, incident notes, change logs |\n| `data-scientist` | Experiment tracking, model notes, dataset documentation |\n\nApply a pack to an existing vault without losing any notes:\n\n```bash\nneurostack scaffold researcher ~/my-notes # or: writer, student, developer, devops, data-scientist\n```\n\nYou can also import an existing Markdown directory:\n\n```bash\nneurostack onboard ~/my-notes\n```\n\n---\n\n## How retrieval works\n\nMost memory tools give your AI a wall of text and let it figure out what's relevant. NeuroStack is tiered. It starts with the cheapest retrieval that answers the question and escalates only when it needs to.\n\n| Level | Tokens | What your AI gets |\n|-------|--------|-------------------|\n| Quick facts | ~15 | Structured facts extracted from your notes: `experiment-3 used learning-rate 0.001` |\n| Summaries | ~75 | AI-generated overview of a note |\n| Full content | ~300 | Actual Markdown content |\n| Auto (default) | Varies | Starts at quick facts, escalates only if the answer isn't there |\n\nSimple factual questions resolve at ~15 tokens. Deep dives get full context. Your AI spends its attention budget where it matters.\n\n---\n\n## Your AI remembers decisions\n\nAcross sessions, your AI can save and retrieve typed memories: observations, decisions, conventions, learnings, bugs. When you start a new session, those memories are surfaced automatically.\n\n> \"We decided to keep authentication stateless.\"\n> \"The thesis framing shifted from consolidation to complementary learning systems.\"\n> \"Elena's surname changed from Vasquez to Reyes in the chapter 7 revision.\"\n\nThese aren't just notes. They're things your AI remembers you decided together. They survive `/clear`. They survive closing the terminal. They survive switching machines.\n\n```bash\nneurostack memories add \"revised thesis framing to CLS, not just consolidation\" --type decision --tags \"thesis,neuroscience\"\nneurostack memories search \"thesis direction\"\n```\n\n---\n\n## Learns from your AI sessions\n\nNeuroStack can scan your past AI conversations, extract the key decisions, observations, and learnings, and save them as memories — automatically. No manual work.\n\n```bash\nneurostack harvest --sessions 5 # extract insights from last 5 sessions\nneurostack hooks install # set up hourly auto-harvest\n```\n\nSupports Claude Code, VS Code, Codex CLI, Aider, and Gemini CLI session formats.\n\n---\n\n## Keeps itself current\n\nYour vault changes. NeuroStack watches it.\n\n```bash\nneurostack watch # auto-index on vault changes\n```\n\nThe index updates as you write. Stale detection runs continuously. You don't maintain it — it maintains itself.\n\n---\n\n## What changes day-to-day\n\n| Without NeuroStack | With NeuroStack |\n|-------------------|-----------------|\n| AI answers from training data | AI answers from your actual notes |\n| Cites the runbook you deprecated | Flags it as stale, demotes it automatically |\n| No memory of yesterday's session | `session_brief` reconstructs working context |\n| Reading 10 notes to find one fact | Tiered retrieval: ~15 tokens for a structured fact |\n| Decisions lost after `/clear` | Typed memories persist indefinitely |\n\n---\n\n## How your vault is stored\n\n```\n~/your-vault/ # your Markdown files (not modified by indexing; AI clients can edit via opt-in MCP write tools)\n~/.config/neurostack/config.toml # configuration\n~/.local/share/neurostack/\n neurostack.db # SQLite + FTS5 knowledge graph\n sessions.db # session transcript index\n```\n\nNeuroStack reads your vault. By default, it writes nothing back — all index data lives in its own SQLite databases. The opt-in MCP write tools (`vault_write_file` / `vault_delete_file`) are the one exception: they create or edit `.md` files in the vault and commit + push the change to your git remote on the spot.\n\n### Memory write-back (opt-in)\n\nMemories live in SQLite by default, so they're invisible in Obsidian and vanish if the database is lost. Turn on write-back to persist qualifying memories as markdown files you own:\n\n```toml\n[writeback]\nenabled = true # opt-in; default false\npath = \".neurostack\" # quarantine dir, relative to vault_root\ninclude_observations = false # also write the noisier observation/context types\n```\n\n- Files land under `{vault_root}/.neurostack/memories///.md`. NeuroStack only ever writes inside that one directory — your notes are never touched.\n- Only persistent (no-TTL) `decision` / `convention` / `learning` / `bug` memories are written; ephemeral (TTL) memories never are.\n- The database stays the source of truth; files are readable exports. `vault_remember` / `vault_update_memory` / `vault_forget` / `vault_merge` keep the files in step automatically.\n- The directory self-ignores via its own `.gitignore` so memories stay out of git until you opt in (delete that file to version them). NeuroStack never commits on your behalf.\n- `neurostack migrate write-back [--dry-run]` exports existing memories; `neurostack sync` reconciles files against the DB (the DB wins on conflict).\n\n---\n\n

All 24 MCP tools

\n\n**Search & retrieval**\n\n| Tool | Description |\n|------|-------------|\n| `vault_search` | Hybrid search with tiered depth (`triples`, `summaries`, `full`, `auto`) |\n| `vault_ask` | RAG Q&A with inline citations |\n| `vault_summary` | Pre-computed note summary |\n| `vault_graph` | Wiki-link neighborhood with PageRank scores |\n| `vault_related` | Semantically similar notes by embedding distance |\n| `vault_triples` | Knowledge graph facts (subject-predicate-object) |\n| `vault_communities` | GraphRAG queries across topic clusters |\n| `vault_context` | Task-scoped context assembly within token budget |\n\n**Context & insights**\n\n| Tool | Description |\n|------|-------------|\n| `session_brief` | Compact session briefing |\n| `vault_stats` | Index health, excitability breakdown, memory stats |\n| `vault_record_usage` | Track note hotness |\n| `vault_prediction_errors` | Surface stale notes |\n\n**Memories**\n\n| Tool | Description |\n|------|-------------|\n| `vault_remember` | Store a memory (returns duplicate warnings + tag suggestions) |\n| `vault_update_memory` | Update a memory in place |\n| `vault_merge` | Merge two memories (unions tags, audit trail) |\n| `vault_forget` | Delete a memory |\n| `vault_memories` | List or search memories |\n| `vault_harvest` | Extract insights from session transcripts |\n\n**Sessions**\n\n| Tool | Description |\n|------|-------------|\n| `vault_session_start` | Begin a memory session |\n| `vault_session_end` | End session with optional summary and auto-harvest |\n\n**Vault files** (opt-in write surface — git-backed)\n\n| Tool | Description |\n|------|-------------|\n| `vault_read_file` | Read a `.md` file under your vault root |\n| `vault_list_files` | List `.md` files; hidden segments (`.git`, `.obsidian`, …) always excluded |\n| `vault_write_file` | Create or overwrite a `.md` file; commits + pushes `origin/main`. Hard-rejects writes without required frontmatter (`date`, `tags`, `type`). On push conflict: `git pull --rebase --autostash` + retry once, then rollback. |\n| `vault_delete_file` | Delete a `.md` file; commits + pushes `origin/main` |\n\n

\n\n

Full CLI reference

\n\n```\n# Setup\nneurostack init # one-command setup: deps, vault, index\nneurostack init --mode full ~/brain # non-interactive full mode\nneurostack onboard ~/my-notes # import existing Markdown notes\nneurostack scaffold researcher # apply a profession pack\nneurostack scaffold --list # see all packs\nneurostack update # pull latest source + re-sync deps\nneurostack uninstall # complete removal\n\n# Search & retrieval\nneurostack search \"query\" # hybrid search\nneurostack ask \"question\" # RAG Q&A with citations\nneurostack tiered \"query\" # tiered: triples -> summaries -> full\nneurostack triples \"query\" # knowledge graph triples\nneurostack summary \"note.md\" # AI-generated note summary\nneurostack related \"note.md\" # semantically similar notes\nneurostack graph \"note.md\" # wiki-link neighborhood\nneurostack communities query \"topic\" # GraphRAG across topic clusters\nneurostack context \"task\" --budget 2000 # task-scoped context recovery\nneurostack brief # session briefing\n\n# Maintenance\nneurostack index # build/rebuild knowledge graph\nneurostack watch # auto-index on vault changes\nneurostack decay # excitability report\nneurostack prediction-errors # stale note detection\nneurostack backfill [summaries|triples|all]\nneurostack communities build # rebuild topic clusters\nneurostack reembed-chunks # re-embed all chunks\n\n# Memories\nneurostack memories add \"text\" --type observation\nneurostack memories search \"query\"\nneurostack memories list\nneurostack memories update --content \"revised\"\nneurostack memories merge \nneurostack memories forget \nneurostack memories prune --expired\n\n# Sessions\nneurostack harvest --sessions 5 # extract session insights\nneurostack sessions search \"query\" # search transcripts\nneurostack hooks install # hourly harvest timer\n\n# Client setup\nneurostack setup-client cursor # or: windsurf, gemini, vscode, claude-code\nneurostack setup-client --list\nneurostack setup-desktop # Claude Desktop\n\n# Diagnostics\nneurostack stats # index health\nneurostack doctor # validate all subsystems\nneurostack demo # interactive demo with sample vault\n```\n\n

\n\n

Neuroscience basis

\n\nEach feature models a specific mechanism from memory neuroscience:\n\n| Feature | Mechanism | Citation |\n|---------|-----------|----------|\n| Stale detection + demotion | Prediction error signals trigger reconsolidation | Sinclair & Bhatt 2022 |\n| Excitability decay | CREB-elevated neurons preferentially join new memories | Han et al. 2007 |\n| Co-occurrence learning | Hebbian \"fire together, wire together\" plasticity | Hebb 1949 |\n| Topic clusters | Hopfield attractor basin dynamics, inverse temperature | Ramsauer et al. 2020 |\n| Convergence confidence | Energy landscape retrieval, basin width = robustness | Krotov & Hopfield 2016 |\n| Lateral inhibition | PV+/SOM+ interneuron winner-take-all competition | Rashid et al. 2016 |\n| Tiered retrieval | Complementary learning systems | McClelland et al. 1995 |\n\nFull citations: [docs/neuroscience-appendix.md](https://github.com/raphasouthall/neurostack/blob/main/docs/neuroscience-appendix.md)\n\n

\n\n---\n\n## FAQ\n\n**Does it modify my vault files?** Not by default. Indexing, search, summaries, and every read tool leave your files untouched — all index data lives in NeuroStack's own SQLite databases. Four opt-in MCP write tools (`vault_write_file`, `vault_delete_file`, plus `vault_read_file` / `vault_list_files`) let an AI client author and edit notes; every write commits and pushes to your git remote, so changes are tracked and revertable. If your vault is not a git repo, the file is still written to disk but the commit step is skipped. Separately, opt-in [memory write-back](#memory-write-back-opt-in) persists memories as markdown, but only ever inside the quarantined `.neurostack/` directory — never alongside your own notes.\n\n**Do I need a GPU?** No. Lite mode has zero ML dependencies. Full mode runs on CPU but summarization is slow without a GPU.\n\n**Do I need to know Python?** No. The npm package handles everything. You never touch a virtualenv.\n\n**How large a vault can it handle?** Tested with ~5,000 notes. FTS5 search stays fast at any size.\n\n**Can I use it without an AI client?** Yes. The CLI works standalone and pipes into any LLM.\n\n**Is my vault private?** Yes. Nothing leaves your machine, unless you point Full mode at a third-party LLM provider instead of local Ollama. In that case the text you index goes to that provider under its own policy.\n\n**What AI clients does it work with?** Claude Code, Claude Desktop, Cursor, Windsurf, Gemini CLI, VS Code, and Codex — anything that supports MCP.\n\n---\n\n## Requirements\n\n- Linux or macOS\n- **Lite mode:** Node.js + Python 3.11+. No GPU or Ollama required.\n- **Full mode:** [Ollama](https://ollama.ai) with `nomic-embed-text` and a summary model. GPU or 6+ core CPU recommended.\n\n---\n\n## Get started\n\n```bash\nnpm install -g neurostack\nneurostack init\n```\n\nTwo minutes. One wizard. Your AI stops forgetting.\n\n- **Contributing:** [CONTRIBUTING.md](https://github.com/raphasouthall/neurostack/blob/main/CONTRIBUTING.md)\n- **GitHub:** [github.com/raphasouthall/neurostack](https://github.com/raphasouthall/neurostack)\n- **Sponsor:** [GitHub Sponsors](https://github.com/sponsors/raphasouthall) | [Buy me a coffee](https://buymeacoffee.com/raphasouthall)\n\n---\n\nApache-2.0 — see [LICENSE](https://github.com/raphasouthall/neurostack/blob/main/LICENSE). No GPL dependencies.\n","changelogMarkdown","videoKey","requiredSecretsJson","publishedAt","priceCents","trialDays",0,"categoryDisplayName","Developer Tools","upstream",{"_95":96,"_97":98,"_99":100,"_101":102,"_103":104},"requiredSecrets",[],"tags",[],"assets",[],"stats",{"_93":80,"_94":80},"ratingAvg","reviewCount","topics",[105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121],"license","Apache-2.0","language","Python","registryName","io.github.raphasouthall/neurostack","homepage","https://neurostack.sh","ai-memory","knowledge-graph","local-ai","local-first","markdown","mcp","mcp-server","neuroscience","note-taking","obsidian","open-source","pkm","python","rag","second-brain","semantic-search","stale-detection","ENV",{"_124":125,"_126":127,"_128":129},"ENVIRONMENT","production","TURNSTILE_SITE_KEY","0x4AAAAAAB8-rEI2gNDf_QhU","SENTRY_DSN","https://45fd1229f78089d5c3ee78be128030e1@o4510583261757440.ingest.de.sentry.io/4510584253448272"]