@chatpanel/events 0.70.0 → 0.71.0

package/index.js

@@ -201,7 +201,8 @@ export {
 } from './memory.js';
 export { SOURCE_TRUST, SkillSourceError, defineSkillSource, createSkillSourceRegistry } from './skill-sources.js';
 export { SKILL_MANIFEST_VERSION, SKILL_CONTEXTS, SKILL_HISTORY_SCOPES, SKILL_MCP_MODES, SKILL_TRUST, SKILL_FILE_KINDS, SKILL_UPCASTERS, SkillManifestError, isSafeSkillPath, originOf, trustOf, skillFiles, needsBridge, declaredAccess, originLabel, sameSkillOrigin, skillIsStale, validateSkill, upcastSkill, upcastSkills, normalizeSkill } from './skill-manifest.js';
-export { SKILL_VARS, SKILL_VAR_NAMES, skillVar, skillVarPattern, parseSkillVars, lintSkillPrompt, suggestSkillVar, substituteSkillVars, skillVarGuidance, SkillVarError } from './skill-vars.js';
+export { SKILL_VARS, SKILL_VAR_NAMES, skillVar, skillVarPattern, parseSkillVars, lintSkillPrompt, suggestSkillVar, substituteSkillVars, expandSkillPrompt, skillVarGuidance, SkillVarError } from './skill-vars.js';
+export { SLASH_TYPING_RE, enabledSkills, slashCommandItems, matchSlashSkill, matchSlashRecipe, recipeInvocationText, slashCommandInsert, skillInvocationOf, skillInvocationLabel } from './slash-commands.js';
 export { outlineOf, parseListItem, continueList, indentSelection, toggleWrap, toggleLinePrefix, toggleTask, toggleLink, docStats, selectionStats } from './markdown-authoring.js';
 export {
   MAX_TAG_LENGTH, MAX_TAGS, normalizeTag, normalizeTags, hasTag, addTag, removeTag, toggleTag,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@chatpanel/events",
-  "version": "0.70.0",
+  "version": "0.71.0",
   "description": "The canonical ChatPanel event-log and capability contracts \u2014 typed durable facts, clock-free deterministic linearization, schema upcasting, and the invariants the replay harness asserts. Pure, dependency-free ESM shared by the ChatPanel extension, gateway and bridge.",
   "type": "module",
   "main": "index.js",
@@ -67,6 +67,7 @@
     "./skill-scan.js": "./skill-scan.js",
     "./skill-sources.js": "./skill-sources.js",
     "./skill-vars.js": "./skill-vars.js",
+    "./slash-commands.js": "./slash-commands.js",
     "./sources-retrieval.js": "./sources-retrieval.js",
     "./sources.js": "./sources.js",
     "./store.js": "./store.js",
@@ -179,6 +180,7 @@
     "skill-scan.js",
     "skill-sources.js",
     "skill-vars.js",
+    "slash-commands.js",
     "sources-retrieval.js",
     "sources.js",
     "store.js",

package/skill-vars.js

@@ -224,6 +224,22 @@ export async function substituteSkillVars(text, { args = '', resolvers = {} } =
   return { text: out, filled, empty, unknown: lint.unknown };
 }
 
+/**
+ * The prompt a `/command args` sends.
+ *
+ * The typed args fill `{{input}}` when the author left a slot for them, and are appended
+ * after the prompt when not — never both, which is how "/fix this sentence" once landed
+ * in the prompt twice. Everything else `substituteSkillVars` reports (an empty selection,
+ * an invented placeholder) comes back untouched, so the caller can say so out loud.
+ */
+export async function expandSkillPrompt(prompt, { args = '', resolvers = {} } = {}) {
+  const src = String(prompt || '');
+  const a = String(args || '').trim();
+  const inline = lintSkillPrompt(src).hasInput;
+  const body = src + (!inline && a ? `\n\n${a}` : '');
+  return substituteSkillVars(body, { args: a, resolvers });
+}
+
 /**
  * The sentence prompt-assist needs so the model stops inventing placeholders.
  * Generated from SKILL_VARS so a variable added here reaches the assist prompt

package/slash-commands.js

@@ -0,0 +1,133 @@
+// slash-commands.js — the `/command` grammar of a chat composer, declared ONCE.
+//
+// A skill, a saved recipe and a client's built-in command all answer to a slash. Which
+// ones exist, how a half-typed prefix is matched, what a chosen item inserts, and how a
+// sent run is LABELLED all lived in the extension's side panel, and the desktop had to copy
+// them the day its composer wanted `/summarize` too. This module is that copy, made the
+// original: a third client (mobile, a channel bot) inherits the grammar instead of
+// re-deriving it.
+//
+// What is NOT here is deliberate:
+//   • the BUILT-IN list — `/search`, `/history`, `/monitor` — is what a client can DO,
+//     and a desktop with no live meeting has no `/tldr`. Each client passes its own.
+//   • the prompt expansion — `expandSkillPrompt` in skill-vars.js, so a composer that only
+//     needs the menu never pays for the variable layer.
+//   • the entitlement — `skillsAllowed` is decided by the caller against its licence.
+
+function normalizePrefix(prefix) {
+  return String(prefix || '')
+    .toLowerCase()
+    .replace(/^\//, '')
+    .replace(/\s+/g, ' ');
+}
+
+function skillItem(skill) {
+  return {
+    type: 'skill',
+    command: skill.command || '',
+    icon: skill.icon || '🎓',
+    description: skill.description || skill.name || '',
+    skill,
+  };
+}
+
+// A saved recipe is a /command too. It is not a skill: nothing is expanded into a prompt.
+// The line becomes a plain request to run it, and the `recipe` tool does the rest.
+function recipeItem(recipe) {
+  return { type: 'recipe', command: recipe.name || '', icon: '🧩', description: recipe.description || 'Saved recipe', recipe };
+}
+
+/** Skills that are switched on. Absence of the flag means enabled (older records have none). */
+export function enabledSkills(skills) {
+  return (Array.isArray(skills) ? skills : []).filter((s) => !!s && s.enabled !== false);
+}
+
+/**
+ * The items a composer offers for what has been typed so far.
+ *
+ * @param builtins      the client's own commands — `{ command, icon, description, feature? }`.
+ *                      One with a `feature` is shown LOCKED when `features[feature]` is
+ *                      false: a Free user should still discover what Pro unlocks.
+ * @param skills        the user's skills; only enabled ones with a command are offered
+ * @param recipes       saved recipes; enabled ones by name
+ * @param prefix        what follows the slash, possibly with a subcommand ("history m")
+ * @param skillsAllowed whether skills are offered at all (Pro on the extension)
+ * @param features      `{ liveMeetings: true }` — what the licence unlocks
+ */
+export function slashCommandItems({
+  builtins = [],
+  skills = [],
+  recipes = [],
+  prefix = '',
+  skillsAllowed = false,
+  features = {},
+} = {}) {
+  const normalized = normalizePrefix(prefix);
+  const own = (Array.isArray(builtins) ? builtins : []).map((item) => ({
+    type: 'builtin',
+    ...item,
+    locked: !!item.feature && !features[item.feature],
+  }));
+  const skillItems = skillsAllowed ? enabledSkills(skills).map(skillItem) : [];
+  const recipeItems = (recipes || []).filter((r) => r && r.enabled !== false && r.name).map(recipeItem);
+  return [...own, ...skillItems, ...recipeItems]
+    .filter((item) => item.command && item.command.toLowerCase().startsWith(normalized))
+    .slice(0, 12);
+}
+
+/** Is the composer's text a slash command still being typed — the moment to show the menu? */
+export const SLASH_TYPING_RE = /^\/([a-z0-9_-]*(?:\s+[a-z0-9_-]*)?)$/i;
+
+/** "/summarize the thread" → the enabled skill whose command is `summarize`, and the rest. */
+export function matchSlashSkill(text, skills = []) {
+  const m = /^\/([a-z0-9_-]+)\s*([\s\S]*)$/i.exec(String(text || ''));
+  if (!m) return null;
+  const skill = enabledSkills(skills).find((s) => String(s.command || '').toLowerCase() === m[1].toLowerCase());
+  return skill ? { skill, args: m[2].trim() } : null;
+}
+
+/** "/open_bug Crash on start" → the recipe, and the rest of the line as its input. */
+export function matchSlashRecipe(text, recipes = []) {
+  const m = /^\/([a-z0-9_-]+)\s*([\s\S]*)$/i.exec(String(text || ''));
+  if (!m) return null;
+  const recipe = (recipes || []).find((r) => r && r.enabled !== false && String(r.name || '').toLowerCase() === m[1].toLowerCase());
+  return recipe ? { recipe, args: m[2].trim() } : null;
+}
+
+/** What the model receives for a recipe command: a request, not a prompt expansion. */
+export function recipeInvocationText(recipe, args = '') {
+  const a = String(args || '').trim();
+  return `Run the saved recipe "${recipe.name}"${a ? ` with this input: ${a}` : ''}. Use the recipe tool; if a parameter is missing, ask for it.`;
+}
+
+export function slashCommandInsert(item) {
+  return item?.command ? `/${item.command} ` : '/';
+}
+
+/**
+ * How a skill invocation should READ once it has been sent.
+ *
+ * The model must receive the whole prompt — that is what a skill IS. The user's own bubble
+ * echoing it back is a separate question with a different answer: a screen of instructions
+ * they wrote themselves buries the thread the skill was asked to read. So the send keeps
+ * the expansion as the message CONTENT (model, exports and memory capture see exactly what
+ * they always saw) and carries this beside it, for display only.
+ *
+ * Null for anything that is not a command, so callers attach it unconditionally and older
+ * messages — which have none — keep rendering as they did.
+ */
+export function skillInvocationOf(skill, args = '') {
+  if (!skill?.command) return null;
+  return {
+    command: skill.command,
+    args: String(args || '').trim(),
+    name: skill.name || '',
+    icon: skill.icon || '🎓',
+  };
+}
+
+/** The one line an invocation is worth: `/command args`, exactly as typed. */
+export function skillInvocationLabel(inv) {
+  if (!inv?.command) return '';
+  return `/${inv.command}${inv.args ? ` ${inv.args}` : ''}`;
+}