Skip to content

quick-start.md

Latest commit

 

History

History
109 lines (70 loc) · 4.86 KB

quick-start.md

File metadata and controls

109 lines (70 loc) · 4.86 KB
Raw

Quick Start

Get HotM running in under 10 minutes. HotM is a React UI that talks to a Fortemi API. Pick the path that matches your setup.

Choose your install path

Path When to use Footprint What gets installed
Desktop bundle First time trying HotM, no existing LLM setup ~10 GB HotM + Fortemi sidecar + Postgres + Ollama + models
Docker UI + external Fortemi You already run Fortemi somewhere ~200 MB Just the UI container; you provide Fortemi
Bring Your Own LLM You already run llama.cpp / vLLM / an OpenAI-compatible endpoint HotM is agnostic UI + Fortemi configured to point at your endpoint (skip the bundled Ollama)

The UI is inference-agnostic — it only needs VITE_API_BASE_URL pointing at any Fortemi instance. The inference backend (Ollama, llama.cpp, vLLM, OpenAI-compatible) is configured entirely on the Fortemi side.

BYO-LLM users: the desktop installer installs Ollama by default. Run with --no-ollama to skip it. See Bring Your Own LLM below.

Desktop Installation

The desktop app bundles the Fortemi API sidecar — no separate server setup required.

Linux

One command installs everything (Postgres 18 via PGDG + pgvector + postgis + Ollama + HotM):

curl -fsSL https://raw.githubusercontent.com/Fortemi/HotM/main/scripts/install.sh | bash

The bootstrap downloads + checksum-verifies the latest .deb, runs apt install (which pulls Postgres 18, pgvector, and postgis from PGDG via Depends/Recommends), seeds the matric database via the postinst hook, then installs the Ollama daemon and pulls models in the background. Idempotent — re-running is safe.

Launch:

hotm

See desktop-linux.md for the AppImage path, manual install steps, troubleshooting, and the full list of installer flags.

macOS

1. Install prerequisites:

curl -fsSL https://raw.githubusercontent.com/Fortemi/HotM/main/scripts/setup-macos.sh | bash

2. Install HotM: Download the latest HotM_*_aarch64.dmg from the releases page, open it, and drag HotM to /Applications.

3. Launch: Open HotM from Applications. On first launch macOS may show a Gatekeeper warning — right-click and choose Open.

See desktop-macos.md for full details.

Docker (Web UI Only)

For server/headless deployments. Requires a separately running Fortemi instance.

docker run -d \
  --name hotm \
  -e VITE_API_BASE_URL=http://your-fortemi-host:3000 \
  -p 8080:80 \
  ghcr.io/fortemi/hotm-ui:latest

Open http://localhost:8080. See docker.md for compose setup.

Bring Your Own LLM

If you already run llama.cpp, vLLM, or any OpenAI-compatible endpoint, skip the bundled Ollama install — Fortemi is configurable for any inference backend. The Linux installer still starts Speaches/Whisper by default for audio/video transcription unless you pass --no-whisper.

Linux desktop bundle, no Ollama:

curl -fsSL https://raw.githubusercontent.com/Fortemi/HotM/main/scripts/install.sh | bash -s -- --no-ollama

This installs HotM + Fortemi + Postgres but skips the Ollama daemon and model pulls. Configure your inference endpoint in Fortemi (env vars or inference.toml):

  • MATRIC_INFERENCE_DEFAULT=llamacpp (or openai)
  • LLAMACPP_BASE_URL=http://localhost:8080/v1 (or your endpoint)
  • OPENAI_API_KEY=... (if using OpenAI or a compatible service)

See the Fortemi configuration docs for the full inference backend matrix.

Linux desktop bundle, no local transcription:

curl -fsSL https://raw.githubusercontent.com/Fortemi/HotM/main/scripts/install.sh | bash -s -- --no-whisper

This leaves media upload enabled but disables local transcription scheduling in the desktop sidecar profile.

HotM in Docker doesn't run inference at all — that's all on the Fortemi side. The Docker quickstart above already supports any Fortemi instance via VITE_API_BASE_URL.

Status Indicators

After launch, the sidebar shows the connection state:

Indicator Meaning
API Connected (green) Fully operational
Degraded (yellow) Fortemi reachable, inference unavailable (Ollama not running)
Offline Mode (red) Cannot reach Fortemi — check PostgreSQL and sidecar logs

To recover from Degraded: start Ollama (ollama serve or sudo systemctl start ollama), then close and reopen HotM.

Next Steps