@samuelfaj/distill 1.4.6 → 1.4.8

package/README.md

@@ -23,6 +23,8 @@ D inspect remotes + MR meta
 R merge/update may block w/o token
 ```
 
+Inline variables use dynamic `<term>=#<letter><digit>` assignments chosen by the model from repeated terms. They stay thread-local unless `distill dsl learn-thread --stdin` sees the explicit variable more than 5 times; learned entries are removed when absent from the next learned thread.
+
 `/distill` also has DSL memory:
 
 ```bash

package/bin/distill.js

@@ -41,23 +41,23 @@ function resolveBinaryPath() {
     process.exit(1);
   }
 
+  const workspaceBinaryPath = path.resolve(
+    __dirname,
+    "..",
+    "..",
+    `distill-${target}`,
+    "bin",
+    targetSpec.binaryName
+  );
+
+  if (fs.existsSync(workspaceBinaryPath)) {
+    return workspaceBinaryPath;
+  }
+
   try {
     const packageJsonPath = requireFromHere.resolve(`${targetSpec.packageName}/package.json`);
     return path.join(path.dirname(packageJsonPath), "bin", targetSpec.binaryName);
   } catch (error) {
-    const workspaceBinaryPath = path.resolve(
-      __dirname,
-      "..",
-      "..",
-      `distill-${target}`,
-      "bin",
-      targetSpec.binaryName
-    );
-
-    if (fs.existsSync(workspaceBinaryPath)) {
-      return workspaceBinaryPath;
-    }
-
     console.error(
       `[distill] Missing platform package ${targetSpec.packageName}. Reinstall @samuelfaj/distill for this platform.`
     );

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@samuelfaj/distill",
-  "version": "1.4.6",
+  "version": "1.4.8",
   "description": "Compress command output for downstream LLMs.",
   "license": "MIT",
   "bin": {
@@ -16,11 +16,11 @@
     "node": ">=18"
   },
   "optionalDependencies": {
-    "@samuelfaj/distill-darwin-arm64": "1.4.6",
-    "@samuelfaj/distill-darwin-x64": "1.4.6",
-    "@samuelfaj/distill-linux-arm64": "1.4.6",
-    "@samuelfaj/distill-linux-x64": "1.4.6",
-    "@samuelfaj/distill-win32-x64": "1.4.6"
+    "@samuelfaj/distill-darwin-arm64": "1.4.8",
+    "@samuelfaj/distill-darwin-x64": "1.4.8",
+    "@samuelfaj/distill-linux-arm64": "1.4.8",
+    "@samuelfaj/distill-linux-x64": "1.4.8",
+    "@samuelfaj/distill-win32-x64": "1.4.8"
   },
   "publishConfig": {
     "access": "public"

package/skills/distill/SKILL.md

@@ -165,6 +165,37 @@ C PATH pkg bin => repo patch ignored
 R missing token => merge blocked
 ```
 
+## Variable Dict
+
+Every thread must use DSL/Dict when it helps compression.
+Start with `Dict:` when meanings are not already shared.
+Define short thread variables inline when a noun repeats or is likely to repeat.
+The model chooses the variables dynamically from the current task; there is no fixed variable list.
+
+```text
+S cache=#c1 warmed model=#m1
+D inspect #c1 hit rate
+D compare #m1 latency
+N no extra vars for one-off nouns
+```
+
+After definition, use the variable:
+
+```text
+D tune #c1 ttl
+D benchmark #m1 output
+```
+
+Rules:
+
+- variable key format: `#` + letter + digit
+- one stable meaning per variable inside the thread
+- do not define variables for secrets, people, IDs, paths, URLs, or one-off terms
+- do not redefine an active variable; add a new key if meaning changed
+- inline `#` variables are thread-local immediately after explicit assignment
+- persist only variables used more than 5 times in a `distill dsl learn-thread --stdin` transcript
+- remove a learned variable when it is absent from the next learned thread
+
 ## Good Response Forms
 
 Tiny status:
@@ -206,7 +237,12 @@ R not committed
 
 ## Glossary And Memory
 
-Keep an internal alias dict per conversation. Do not create files.
+Use two memory layers:
+
+- thread dict: internal aliases for the current conversation
+- persisted dict: entries saved by `distill dsl learn`, `distill dsl learn-thread --stdin`, or explicit user action
+
+Do not manually create memory files. Let the `distill` CLI own JSON memory writes.
 
 Use aliases only when they stay obvious:
 
@@ -245,7 +281,15 @@ Dict: S=state C=context D=action R=risk O=outcome N=no-go P=proof 1=failing-test
 Dict+: A1=authentication bug fix
 ```
 
-Expire learned terms mentally if they stop appearing. A term should not become part of thread DSL unless it appears at least twice in a short window or the user explicitly approves it.
+Persisted learned terms start as candidates and promote only through lifecycle rules. A term should not become part of active DSL unless it appears at least twice in a short window or the user explicitly approves it.
+
+At every thread end or handoff, when transcript export is available, run:
+
+```bash
+distill dsl learn-thread --stdin < transcript.txt
+```
+
+This analyzes repeated usage, rejects sensitive/noisy terms, asks the configured reviewer model for strict JSON, and persists only compact candidates.
 
 ## Tool Calls