Frequently Asked
Questions

SEEV is a local-first, private AI workspace for macOS. It can chat, generate code, summarize documents, search the web, and manage local memory while keeping core workspace data on your device. SEEV includes a bundled LFM 2.5 1.2B model for first-launch use, an optional LFM2-VL-3B reasoning download, and Smart Hybrid routing between them.

Download SEEV from this website, open the DMG file, and drag the application to your Applications folder. Launch it, and you are ready to begin — no account registration, no API keys, and no configuration wizard required. The bundled 1.2B model is included for first launch, and the optional 3B reasoning model can be installed later if you want it.

No. SEEV requires no account, no email address, no phone number, and no sign-in. The application is designed for complete anonymity. Download and begin using it immediately.

SEEV runs as a macOS app and is optimized for local inference. Because the bundled model cache footprint is about 1.2 GB, plan for enough free disk space for the app plus its local cache. Performance will vary by hardware, especially when using the deep reasoning model or voice input.

Yes. SEEV is entirely free to download and use. There are no subscriptions, no usage limits, no premium tiers, and no in-app purchases. The AI model runs on your own hardware, eliminating per-token costs.

Yes. SEEV has no backend servers and no telemetry systems. Your conversations, memories, and settings are stored locally on your device, and core AI flows run locally as well. Optional features such as web search or an external provider use the network only when you explicitly invoke them.

Core AI functionality — including chat, code generation, and summarization — operates entirely offline with zero network requests. The only feature that uses the internet is Web Search, which is explicitly opt-in. Even then, only the specific search query is transmitted; your conversation history is never exposed.

No. Your conversations never leave your device, making it impossible for them to be used for training purposes. The AI model bundled with SEEV is fixed and pre-trained — it does not learn or update from your usage. This is fundamentally different from cloud AI services that may use your inputs to improve their models.

All data resides in your device's local storage: IndexedDB for conversations and memories, localStorage for settings and preferences, and Cache Storage for model files. Uninstalling SEEV removes all associated data. You can also clear all data manually from within the Control Center.

SEEV ships with multiple local model paths rather than a single fixed model. The bundled default is LFM 2.5 1.2B Flash for speed, the optional local download is LFM2-VL-3B • Deep Reasoning for tougher prompts and attachments, and Smart Hybrid • Routed switches between them when appropriate.

SEEV uses a bundled local runtime inside the macOS app to keep model execution on-device rather than depending on a remote inference service. The exact performance still depends heavily on your Mac's hardware.

SEEV's 1.2B parameter model is smaller than GPT-4, which operates at a significantly larger scale. For standard tasks — code generation, summarization, Q&A, and structured output — the local model performs well. For complex multi-step reasoning or highly creative writing, larger cloud models may still hold an advantage. SEEV prioritizes privacy and offline capability over raw parameter count. For users requiring cloud-level performance, an optional external provider connection is available.

SEEV allows you to optionally connect your own API key for external AI providers such as OpenAI or Anthropic. This is entirely optional — the local model operates independently. When used, SEEV sends only the current prompt to the external provider; no conversation history is shared. This provides flexibility to access cloud capabilities while maintaining SEEV as your primary local workspace.

Live Knowledge is SEEV's Retrieval-Augmented Generation system. When enabled, SEEV references your stored memories and previous conversations to provide more contextual and personalized responses. The entire retrieval and injection process occurs locally — no data leaves your device.

Yes for core use. Chat, code generation, summarization, memory, and most attachment workflows can run offline with the bundled local model. Connectivity is only needed for optional features such as Web Search or an external provider connection.

SEEV supports a broad range of file types: images (PNG, JPG, GIF), text files (.txt, .md), code files (.py, .js, .html, .css, .swift, and others), documents (.pdf), and structured data files (.json, .csv, .tsv). Files can be attached via drag-and-drop or the attachment button in the composer.

Activate the microphone button in the composer bar to begin voice input. SEEV uses WhisperKit speech recognition to transcribe your speech in real time. All transcription processing occurs on your device — no audio is transmitted to any external service. A live waveform visualization is displayed during recording, and the transcribed text populates the composer automatically.

SEEV's composer provides four specialized modes: Code (optimized for generating structured code output), Summarize (condense lengthy texts), Explain (step-by-step topic breakdowns), and Web Search (search the internet and synthesize results). These modes can be activated via the mode selector buttons or using slash commands: /code, /summarize, /explain, /web.

SEEV keeps memory local and editable. Personal memory, project memory, and uploaded document text can be reviewed, updated, or cleared inside the app, so the saved context stays under your control instead of living in a remote black box.

Yes. The Control Center includes an AI Persona setting where you can define a custom system prompt. This shapes the AI's tone, focus, and communication style. For example, you might configure a more concise, technical, or conversational persona. The persona is applied to all conversations and stored locally.

On the initial launch, SEEV loads the AI model into memory, which typically requires 15 to 60 seconds depending on your hardware configuration. Subsequent launches are faster due to caching. If the model appears to stall during loading, restarting the application generally resolves the issue. Apple Silicon Macs with 16 GB or more of RAM will experience the fastest load times.

Navigate to Settings, scroll to the data management section, and select "Erase All Data and Chats." This action permanently deletes all conversations, memories, and settings from your local storage. Alternatively, uninstalling the application removes all associated data, as everything is stored within the application's local sandbox.

Several optimizations are available: (1) Use the Quick preset, which generates shorter, faster responses. (2) Reduce the Max Tokens setting in the Control Center. (3) Close other memory-intensive applications during use. (4) Enable Compact UI mode to reduce rendering overhead. On Intel-based Macs, generation will be slower compared to Apple Silicon — this performance difference is inherent to the hardware architecture for AI inference workloads.