Features

Features

Feature inventory grounded in the current desktop app (≈ v0.4.0) and supporting packages.

Projects & chat

• Projects with instructions, folders, threads, optional model lock, default tools toggle
• Reference docs — ingest text-like files; chunk + embed; RAG at chat time; summary fallback
• Standalone chats and recents outside a project
• Streaming chat against local Ollama models
• Reasoning / thinking stream with a global Working strip
• Descriptive AI activity trail (tool steps, agent steps)
• Chat attachments UX (design/spec under docs/superpowers/)
• Customize personal context injected into prompts
• Conversation compact helpers

Tools & skills

• Builtin tools: time, FS suite (sandboxed to workspace), web search, HTTP fetch, shell, open app/URL, ask_user, generate_document, …
• Custom tools (UI): shell or HTTP with {param} templates
• Skills: Web research, Google Calendar (code present; connector limited), Outlook
• Custom skills (UI): instructions + selected tools
• Risky tool confirm dialog (renderer must approve)
• URL chip to enable web research for a turn / conversation overrides

Agents & load

• Multi-agent orchestration — classify → plan → parallel workers → synthesize
• Load protection — soft-stop when RAM crosses configured threshold
• Agent settings panel (enable, limits, metering)

Coding & documents

• Project-first coding — prefers real project folder scaffold; disables multi-agent for that turn
• Workspace sandbox for file tools (workspaceGet / Pick / Clear)
• Document generation — pdf / docx / pptx / md via generate_document
• Artifacts explorer — browse/open/reveal/delete generated files
• Standalone generated docs under Documents/AnyLM/generated (platform-dependent)

Knowledge

• Chroma-backed project RAG, memory, general/org vector stores
• Knowledge graph extraction alongside memory
• Soft-degrade when Chroma/embeddings unavailable

Auth, orgs, governance

• Email/password + Google + GitHub sign-in (Firebase)
• Orgs, members, invites, teams
• Policies (model allowlists, content/PII rules — cooperative limits)
• Usage recording, API keys (anylm_…), audit/compliance logs
• Connectors collection for OAuth skill providers

Local proxy

• OpenAI-compatible /v1/models and /v1/chat/completions on loopback (default 3227)
• Authenticated with org/user API keys — not session cookies
• Does not execute AnyLM tools; governance/routing only

Platform UX

• Boot splash + Ollama setup gate / first-run wizard
• Models browser (list/grid, RAM cues)
• Settings hub (left-nav sections, Agents, tabbed project settings)
• Themes / site-theme glass chrome
• Auto-updates via electron-updater when feeds are published
• Marketing site with skills/tools catalog and download picker

Explicit non-features (today)

Item	Status
MCP client/server	Not implemented
Cloud LLM backends (Claude/OpenAI API)	Roadmap (mentioned in app README)
Cloud Functions / Nest server	Retired; in-process API instead
Adversarial usage enforcement	Cooperative only on Spark plan
Google Calendar connector in production	Disabled (token endpoint needs client secret)

Release trail

See docs/releases/ for tagged notes. Website download page reads GitHub Release assets.