SnapMind AI CLI

Production-grade, persona-based AI workflows — directly in your terminal.

SnapMind AI is a local-first Command Line Interface that provides specialized Intelligence Architectures (personas) for researchers, developers, data analysts, and writers. It combines persistent vector storage (LanceDB), streaming AI responses, session memory, and a complexity-based model router into a single npm package.

Highlights

| Feature | Description | |---|---| | 4 Personas | Scholar, Coder, Analyst, Writer — each with tailored tools and system prompts | | Hybrid RAG | Vector similarity + Full-Text Search (FTS) via LanceDB for precision retrieval | | Local-First | 100% offline with Ollama; seamless cloud fallback to OpenAI, Mistral, Anthropic, Gemini | | Secure Vault | API keys stored in your OS-native keychain (via keytar) | | Session Memory | Persistent sessions with named snapshots for branching research paths | | Multi-Agent Handoff | Switch between personas mid-session while preserving full conversation context | | Auto-Routing | Semantic intent detection to automatically select the best persona for your query | | Re-ranking | 2-stage retrieval with LLM-based relevance scoring for high-precision context | | Global Search | Query across all indexed datasets from any persona | | Theme Engine | Cyberpunk, Matrix, and Default terminal color schemes |

Installation

npm install -g snapmind-ai

Prerequisites:

• Node.js >= 18.x
• (Optional) Ollama installed and running for local-first mode
• (Optional) API keys for cloud providers (OpenAI, Mistral, Anthropic, Gemini)

Quick Start

# Launch interactive mode
snapmind-ai

# Direct persona launch
snapmind-ai --persona scholar --mount ./research-papers

# Clone and analyze a GitHub repo
snapmind-ai --persona coder --repo https://github.com/user/repo.git

# Auto-route: let the AI pick the best persona
snapmind-ai "Summarize the key findings from my PDFs"

Personas

Scholar

Deep research, semantic PDF indexing, and academic citation.

| Command | Description | |---|---| | /cite | Generate inline citations from indexed documents | | /bibtex | Export a full BibTeX bibliography from your session | | /global <query> | Search across ALL indexed datasets (Relational RAG) | | /snapshot <name> | Save a named checkpoint of your session state | | /handoff | Transfer context to another persona | | /export | Save session history to a Markdown file | | /stats | View token usage and system health | | exit | Return to main menu |

snapmind-ai --persona scholar --mount <dir>     # Mount a folder of PDFs
snapmind-ai --persona scholar --pages <range>   # Index specific pages (e.g., 1-10)

Coder

Elite software architecture companion with AST-aware indexing.

| Command | Description | |---|---| | /diagram | Generate a Mermaid.js architecture diagram of the codebase | | /skill <script> | Execute an npm script (e.g., /skill test, /skill lint) | | /global <query> | Cross-repo Relational RAG search | | /snapshot <name> | Save named session checkpoint | | /handoff | Transfer context to another persona | | /export | Save session to Markdown | | /stats | View token usage | | exit | Return to main menu |

snapmind-ai --persona coder --repo <url>        # Clone and index a GitHub repo
snapmind-ai --persona coder --mount <dir>       # Mount a local project
snapmind-ai --persona coder --watch             # Live-sync index on file changes

Tech-Stack Intelligence: Automatically detects package.json, requirements.txt, go.mod, and Cargo.toml to inject framework-specific system prompts.

Analyst

High-fidelity data querying, trend analysis, and visualization.

| Command | Description | |---|---| | /chart | Render an ASCII line chart from numerical trends | | /table | Format data snippets into a clean Markdown table | | /global <query> | Cross-dataset Relational RAG search | | /snapshot <name> | Save named session checkpoint | | /handoff | Transfer context to another persona | | /stats | View token usage | | exit | Return to main menu |

snapmind-ai --persona analyst --mount <dir>     # Mount a folder of CSV files

Writer

Web-to-draft synthesis with tone control and collaborative memory.

| Command | Description | |---|---| | /import | Import context from another persona's indexed data | | /global <query> | Cross-dataset Relational RAG search | | /snapshot <name> | Save named session checkpoint | | /handoff | Transfer context to another persona | | /export | Save synthesized drafts to Markdown | | /stats | View token usage | | exit | Return to main menu |

Tone Options: Professional · Creative · Technical · Academic · Concise

CLI Reference

Commands

| Command | Description | |---|---| | snapmind-ai | Launch interactive persona selector | | snapmind-ai config | Open the configuration wizard | | snapmind-ai config set <key> <value> | Set a config value directly | | snapmind-ai search <query> | Global search across all indexed datasets | | snapmind-ai maintenance | Clean up orphan sessions and stale caches | | snapmind-ai vault set <provider> | Securely store an API key in the OS Keychain | | snapmind-ai vault delete <provider> | Remove an API key from the Keychain |

Flags

| Flag | Description | |---|---| | --airgap | Force 100% offline mode (requires Ollama) | | --persona <name> | Launch directly into a persona (scholar, coder, analyst, writer) | | --mount <path> | Mount a local directory for indexing | | --repo <url> | Clone and index a GitHub repository | | --watch | Auto-sync the vector index on file changes | | --pages <range> | Index specific PDF page ranges (Scholar only) |

Configuration

snapmind-ai config

Interactive wizard to configure:

• Default AI Provider — ollama, openai, mistral, anthropic, gemini
• API Keys — Stored securely in your OS Keychain
• Temperature — Control response creativity (0.0 – 1.0)

Supported Providers

| Provider | Models | Local | |---|---|---| | Ollama | Llama 3, Mistral, etc. | Yes | | OpenAI | GPT-4o, GPT-4o-mini (auto-routed) | No | | Mistral | Mistral Large | No | | Anthropic | Claude 3.5 Sonnet | No | | Gemini | Gemini 1.5 Pro | No |

Project Structure

snapmind-ai/
├── src/
│   ├── index.js              # CLI entry point (Commander.js)
│   ├── cli/
│   │   └── menu.js           # Interactive persona selector & handoff loop
│   ├── personas/
│   │   ├── scholar.js        # PDF indexing, /cite, /bibtex
│   │   ├── coder.js          # AST parsing, /diagram, /skill
│   │   ├── analyst.js        # CSV querying, /chart, /table
│   │   └── writer.js         # Web scraping, tone control, /import
│   └── utils/
│       ├── llm.js            # Unified LLM + Embeddings factory
│       ├── lance_store.js    # LanceDB wrapper (Hybrid Search, Re-ranking)
│       ├── vector_storage.js # Namespace manager, Global Search
│       ├── credentials.js    # OS Keychain integration (keytar)
│       ├── session.js        # Persistent session & snapshot manager
│       ├── streamer.js       # Real-time token streaming to terminal
│       ├── router.js         # Complexity-based model routing
│       ├── monitor.js        # Token usage tracking & health dashboard
│       ├── ast_parser.js     # JS/TS AST code block extractor
│       ├── charts.js         # ASCII chart rendering (asciichart)
│       ├── themes.js         # CLI theme engine
│       ├── watcher.js        # File system change watcher (chokidar)
│       ├── exporter.js       # Session-to-Markdown export
│       ├── errors.js         # Centralized error handling
│       ├── config.js         # Persistent config store (conf)
│       └── constants.js      # NLP tuning parameters per persona
├── package.json
└── README.md

Data Storage

All persistent data is stored locally in .snapmind_cache/:

.snapmind_cache/
├── lancedb/           # Vector tables (one per namespace)
└── sessions/          # Conversation history & snapshots

Development

# Clone the repository
git clone https://github.com/roshankumar0036singh/SnapMind.git
cd snapmind-ai

# Install dependencies
npm install

# Link for local development
npm link

# Run locally
snapmind-ai

NLP Tuning

Each persona has configurable retrieval parameters in src/utils/constants.js:

| Persona | Chunk Size | Chunk Overlap | Similarity K | |---|---|---|---| | Scholar | 1000 | 200 | 5 | | Coder | 2000 | 200 | 4 | | Analyst | 1000 | 100 | 5 | | Writer | 1500 | 200 | 6 |

Security

• API Keys are never stored in plaintext. All credentials use the OS-native keychain via keytar.
• Airgap Mode (--airgap) ensures zero network traffic — all processing stays on your machine.
• No telemetry. SnapMind does not phone home or collect usage data.

License

ISC © 2026