Familiar – Local AI Workspace with Chat, Notes, Wiki, and Automations

Familiar is a new open-source AI workspace that integrates chat, notes, wikis, and automations, designed to run locally with support for multiple AI models. The platform emphasizes security with passkey authentication and allows users to create data-driven agents called shards for task automation. Familiar is built with Go and offers deployment via Docker and systemd.

Familiar is an AI-enabled workspace where you can take Notes, collaborate in wikis, chat with models, and create smart automations. In arcane tradition, a familiar is a magical entity bound to a practitioner — it scouts, communicates, and serves as an extension of the caster's will. Familiar has five major concepts - Chat, Notes, Wikis, scheduled actions, and shards. A chat interface for your local AI model that is tied into all of your data - your notes, wikis, and automations. Supports imported skills and system skills like research mode for doing deep dives and searches. Your personal notes, all indexed for easy access by familiar. Edit in rich text or directly in markdown. Share a note publicly if you want to, flattened into a read only page for external users. Collaborate with other users just like you would expect in a simple mark-down enabled wiki. No bloat, just a simple way to share data with a team. Supports mermaid diagrams and images inline. Schedule familiar to execute a task and deliver the output to you via Chat, Notes, Mobile Notifications, Slack or more. From cleaning up your notes every day and providing you a summary to scheduling searches for your current interests, or whatever else you can imagine. Shards are basically data-driven agents. You define the scope of data an agent has access to, it's skills, and how you can access it. You can use an API to access data in your notes, talk to a shard on slack that can only read your recipes, or define a scope for a task runner and assign it via a scheduled action. Familiar is built with security taken seriously. You can login however you want, as long as it is with a passkey. Full instructions — prerequisites, Postgres, model backends, config, first-user passkey registration, systemd, and troubleshooting — are in DEPLOYMENT.md . The short version: 1. Postgres + pgvector docker compose up -d 2. build both binaries cd familiar-gateway && go build -o familiar-gateway ./cmd/gateway/ && cd .. cd familiar-workspace && make build && cd .. 3. write ~/.familiar/gateway.toml and ~/.familiar/workspace.toml see DEPLOYMENT.md 4. run note --http: it mounts the API + admin console ./familiar-gateway/familiar-gateway --http --config ~/.familiar/gateway.toml & ./familiar-workspace/familiar-workspace --config ~/.familiar/workspace.toml & 5. open the workspace URL and register the first passkey gateway unit tests needs a throwaway Postgres for the DB-backed ones cd familiar-gateway FAMILIAR TEST DSN="postgresql://familiar test:familiar test@localhost:5432/familiar test?sslmode=disable" \ go test ./... end-to-end Playwright — builds both binaries into a temp dir and drives a browser cd tests/e2e npm install FAMILIAR TEST DSN="postgresql://familiar test:familiar test@localhost:5432/familiar test?sslmode=disable" \ npx playwright test See tests/e2e/MAKE TESTS.md for the E2E harness details. config.example.toml documents every gateway block. System prompts live in prompts/ as a tiered set base.md , tier .md , tool policy.md ; point system prompt .dir at it. The single config knobs you must set for a usable instance: adapter.http .listen addr , memory .local dsn , at least one models chat model with "tools" in its capabilities , system prompt , and the admin WebAuthn relying-party rp id / rp origins . — new-instance setup and updates. DEPLOYMENT.md /sixvolts/familiar/blob/main/DEPLOYMENT.md — authoritative config reference. config.example.toml MIT