# Usage ## Commands ```sh cd /path/to/project eukhe # start; run /login on first launch eukhe agents # search and open sessions eukhe attach # attach to a running session eukhe --resume [path|id] # resume a session eukhe chat view # print the memory view eukhe chat status [--json] # chat size, compactor progress eukhe chat browse # whole memory as one HTML page eukhe chat import sessions ... eukhe status # daemon state eukhe state [agent] # session state tree eukhe provider caps # provider capabilities; no network eukhe provider models

# models of provider

eukhe tiny-models # list or download local models eukhe config # show and set installed resources; no network eukhe package install # install an npm or git package; add it to settings eukhe package update [s] # update installed packages eukhe doctor [--fix] # check or repair the daemon eukhe shutdown [--force] # stop all agents and the daemon ``` Only `eukhe package install`/`update` download (npm, git). Sessions and `eukhe config` use installed packages only and warn per missing configured package with its `eukhe package install` command. ## Slash commands - `/` at prompt start: command menu, fuzzy on names and aliases. - Space after a command name, or Tab/Enter on a command row: `/ `, its subcommand menu opens at once. - Subcommand menu: the rows whose name starts with the typed word (any case). Tab/Enter on a row: ` `, the next level opens if any. Ghost text after the cursor: rest of the selected row, then its usage. - Enter on a fully typed path runs it: `/memory model`, `/workflows size large`. Bare command with subcommands: its default (`/memory` = `/memory view`, `/model` = the role hub). Bare path that needs a choice (`/dynamic-workflow --size`): its menu opens, nothing runs. - A path never runs its parent. Unknown subcommand: `Unknown: /memory x. Did you mean /memory view?`; nothing runs. - Paths: `/memory [view|model]`, `/model []`, `/workflows [size [small|medium|large]]`, `/dynamic-workflow [--size small|medium|large] `, `/tier []`, `/effort []` (the levels of the current model). ## Agents - Subagents: `rlm.spawn`, `rlm.spawn_batch` in the Python REPL. - Family = all agents under one root session. Message any member by name, id or `parent/child` path; `all` for everyone. - Subagent-to-subagent messages also show in the root chat as relay rows. - The roster note has each agent's latest report or progress. - `@name message`: message a family agent; reply shows as `name -> you`. - `/agents`: session agents with status, depth, model, work. - Agents view: Esc clears the search, then returns to the opening chat; Ctrl+D on empty search or double Ctrl+C exits. Settings: [settings.md](settings.md#agents-and-providers). ## Providers - `eukhe provider caps [provider]`: non-chat capabilities (count tokens, list models, embeddings, images, usage, web search, compaction); sends no request. - `eukhe provider models `: asks the provider: ID, name, context, max output. - `/model`: model role hub (chat, memory, auxiliary, image, subagent, gateway, titles, judge, translate, speech, dictation), each with model and source; `/model ` opens one role's models. Picks save globally; `chat` also switches the session. [settings.md](settings.md#model-roles). ## Session state State tree per session: conversations, tasks, jobs, documents. ```sh eukhe state [agent] # show the tree eukhe state --since # changes after a cursor eukhe state --follow # new changes until Ctrl+C eukhe state rewind-plan # what a rewind stops, starts, restores ```