@plurnk/plurnk-service 0.46.0 → 0.50.0

package/docs/log.md

@@ -0,0 +1,25 @@
+# `log://` — your run's event history, and how to manage it
+
+Every operation you emit is recorded as a row at `log:///<loop>/<turn>/<seq>`. The log is your working memory's ledger — what you've done and what came back. READ a row to recover an earlier op's actual result body (chain a matcher to extract a value). The engine writes it; you read and CURATE it.
+
+## The two budget surfaces
+
+You manage working-memory tokens across two surfaces:
+
+- **The log** — every operation, in order. The curatable surface.
+- **`plurnk:///manifest.json`** — the directory of every entry available to READ. Discovery, not history.
+
+## Curation — OPEN, FOLD, KILL
+
+Your context carries a `tokensFree` budget; keeping it lean keeps your reasoning sharp, because irrelevant tokens degrade signal. Three moves, all on the log:
+
+- **FOLD** a row → contracts it to its one-line summary, reclaiming the body's tokens. Non-destructive: a folded row stays listed and re-OPENable. Your primary hygiene tool — fold a row the moment its body has served its purpose.
+- **OPEN** a folded row → restores its full body, spending from `tokensFree`. Use it to revisit a result you folded too early.
+- **KILL** a row → erases it. Sparingly, for rows whose very existence (not just body) is noise.
+
+## Patterns
+
+- **Fold-as-you-go.** Once you've extracted what you need from a large READ or EXEC result, FOLD it. Don't carry raw bodies you've already digested.
+- **Bulk fold.** `FOLD(log:///**/get)<101,200>` collapses a range of matching rows in one move — clear a burst of exploratory reads once they've paid out.
+- **Recover on demand.** Folded ≠ gone. If a later turn needs a folded result, OPEN it (or READ the row directly with a matcher).
+- **Don't wait for the grinder.** When `tokensFree` runs low the engine's budget grinder folds for you to fit the turn — but it's a mechanical last resort. Curating deliberately keeps the rows YOU value and avoids the overflow in the first place.

package/docs/run.md

@@ -0,0 +1,3 @@
+# `run://` — sibling agent runs
+
+Sister agent runs in this session. `run:///.` is you; `run:///<name>` is a sibling. EDIT(run:///name):prompt SPAWNS a new sibling seeded with that prompt; SEND(run:///name):msg messages a sibling, waking it if idle; COPY(run:///.):prompt FORKS — branches a run with your log so far, then continues it with the prompt; KILL(run:///name) ENDS a sibling. Siblings share this session's files and entries; only the conversation log is private to each.

package/docs/unknown.md

@@ -0,0 +1,3 @@
+# `unknown://` — the catch-all scheme
+
+The catch-all an address falls into when its scheme isn't one you recognize. It behaves like `known:///` — the unrecognized URI still becomes an addressable entry you can EDIT, READ, FIND, and SEND — so a typo or an unmodeled scheme degrades to a usable entry rather than an error.

package/migrations/0000-00-00.01_schema.sql

@@ -49,6 +49,12 @@ CREATE TABLE IF NOT EXISTS loops (
     status   INTEGER NOT NULL DEFAULT 102 CHECK (status IN (100, 102, 200, 202, 413, 429, 499, 500, 508)),
     prompt   TEXT    NOT NULL,
     flags    TEXT    NOT NULL DEFAULT '{}' CHECK (json_valid(flags)),
+    -- #249 — attribution tags of the loop's active plugins (string[] JSON); the activity tagged
+    -- with what its plugins offer. Same set the engine rides on each turn's generate() wire.
+    attributions TEXT NOT NULL DEFAULT '[]' CHECK (json_valid(attributions)),
+    -- #260 — client-passed @file paths foisted as turn-0 READs (string[] JSON). The daemon owns the
+    -- workspace, so it READs them in instead of the client inlining bytes (co-location law).
+    open_paths TEXT NOT NULL DEFAULT '[]' CHECK (json_valid(open_paths)),
     -- §run-scheme loop-termination delta: terminated_at is stamped by the trigger
     -- below when status crosses into terminal (every death-path, uniformly);
     -- terminal_message is the deliverable — the SEND[200] body or the abandonment

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@plurnk/plurnk-service",
-    "version": "0.46.0",
+    "version": "0.50.0",
     "description": "Plurnk agent runtime — the server / engine / core. Implements the @plurnk/plurnk-grammar contract.",
     "type": "module",
     "license": "MIT",
@@ -20,6 +20,7 @@
     "files": [
         ".env.example",
         "dist/**/*",
+        "docs/**/*",
         "migrations/**/*.sql",
         "requirements.md",
         "README.md",
@@ -48,18 +49,20 @@
         "build:local:install": "node scripts/install-sandbox.mjs install",
         "build:local:uninstall": "node scripts/install-sandbox.mjs uninstall",
         "test:installation": "node scripts/test-installation.mjs",
-        "prepare": "npm run build"
+        "deps:audit": "node scripts/deps-audit.ts",
+        "prepare": "npm run build",
+        "prepublishOnly": "npm run deps:audit"
     },
     "dependencies": {
-        "@plurnk/plurnk-execs": "0.4.13",
-        "@plurnk/plurnk-execs-all": "0.1.5",
-        "@plurnk/plurnk-grammar": "0.73.3",
-        "@plurnk/plurnk-mimetypes": "0.15.21",
-        "@plurnk/plurnk-mimetypes-all": "0.5.1",
+        "@plurnk/plurnk-execs": "0.4.15",
+        "@plurnk/plurnk-execs-all": "0.1.6",
+        "@plurnk/plurnk-grammar": "0.73.8",
+        "@plurnk/plurnk-mimetypes": "0.15.23",
+        "@plurnk/plurnk-mimetypes-all": "0.5.2",
         "@plurnk/plurnk-providers": "0.13.0",
         "@plurnk/plurnk-providers-all": "0.8.0",
-        "@plurnk/plurnk-schemes": "0.29.2",
-        "@plurnk/plurnk-schemes-all": "0.1.14",
+        "@plurnk/plurnk-schemes": "0.29.5",
+        "@plurnk/plurnk-schemes-all": "0.1.19",
         "@possumtech/sqlrite": "5.2.0",
         "diff": "^9.0.0",
         "ws": "^8.20.1"