Skip to content

Guides

One task per page, in the order the work usually happens. Each guide assumes a running install (start here if you don't have one) and links back to the concept it applies.

00Manage slotsCreate, load, swap, restart, rename, edit, and delete inference slots with the hal0 CLI or the dashboard's inline pencil editor. Each slot is a podman container behind a systemd unit and the local API.guides/manage-slots.mdx01Pull and register modelsPull models from Hugging Face into hal0's registry, register files already on disk, check for weight updates, scan for new files, and change the model store with the hal0 model CLI.guides/pull-and-register-models.mdx02Choose modelshal0's curated catalogue of recommended chat, embed, rerank, and image models, how launch flags now live on the model instead of the slot, and how to edit a model from the dashboard's row menu.guides/choose-models.mdx03Connect external providersWire up OpenAI, Anthropic, OpenRouter, Ollama, or any OpenAI-compatible provider with the hal0 upstream CLI — credentials, testing, and reachability.guides/connect-external-providers.mdx04Enable memoryhal0's memory subsystem is toggled with hal0 memory enable/disable — Hindsight-backed recall, optional graph extraction, and per-bank management.guides/enable-memory.mdx05Run agentsInstall and manage the bundled Hermes agent with the hal0 agent CLI — provisioning, personas, gated-action approvals, and per-persona spending budgets.guides/run-agents.mdx06Generate imagesRun Stable-Diffusion-family image generation on hal0 with ComfyUI — enable the img capability, read engine status, switch the iGPU between inference and generation, and call /v1/images/generations.guides/generate-images.mdx07Speech to text and text to speechEnable hal0's voice capability with hal0 capabilities set: transcription via Moonshine (CPU) or whisper-class models on the XDNA NPU (FLM), synthesis via Kokoro (CPU) or Qwen3-TTS (GPU), then call /v1/audio/transcriptions and /v1/audio/speech.guides/voice-stt-tts.mdx08Connect MCP serversInstall MCP servers into hal0 with the hal0 mcp CLI, and connect an external MCP client to hal0's own bundled hal0-admin and hal0-memory servers.guides/connect-mcp.mdx09Edit configurationManage hal0's TOML config under /etc/hal0: hal0.toml, upstreams, providers, profiles, stacks, and per-slot files. Validate, reload, and migrate with the hal0 config CLI.guides/configure.mdx10Update and roll backCosign-verified, atomic self-update for hal0 — check channels, apply or pin a version, roll back to the previous tree. Read the profiles.toml reseed warning before your first v1.0 update.guides/update-and-rollback.mdx11Observe the systemTail slot and API logs with the hal0 CLI or the dashboard, and query the durable Activity/Audit log of config changes and slot transitions.guides/logs-and-activity.mdx