diskformer 0.5.4 → 0.6.4

package/ggml/llama.js

@@ -55,9 +55,10 @@ export class Llama {
     return this.request({ type: 'load', opfsName: name, budget, slots, ctx, env }, ['loaded'], (d) => { if (d.type === 'plan' && onPlan) onPlan(d.plan); });
   }
 
-  // A reply through the model's chat template; greedy. Resolves { text, stats }.
-  chat(messages, { n = 512, onPiece } = {}) {
-    return this.request({ type: 'chat', messages, n }, ['done'], (d) => { if (d.type === 'piece' && onPiece) onPiece(d.piece); });
+  // A reply through the model's chat template; greedy. Resolves { text, stats }. thinking: false renders the template
+  // with reasoning off (models that reason first then answer directly); default: the template's own choice.
+  chat(messages, { n = 512, thinking = true, onPiece } = {}) {
+    return this.request({ type: 'chat', messages, n, thinking }, ['done'], (d) => { if (d.type === 'piece' && onPiece) onPiece(d.piece); });
   }
 
   // A continuation of an already formatted prompt; greedy.

package/ggml/worker.mjs

@@ -3,7 +3,7 @@
 // sync access handles, straight into the wasm heap. A GGUF gets into OPFS by download (resumable, from a URL) or by
 // copying a picked file; with opfs: false a picked File is read in place instead (slower).
 //
-// Messages in: load {opfsName | file, budget | slots, ctx, env} · chat {messages, n} · complete {prompt, n} (the gate)
+// Messages in: load {opfsName | file, budget | slots, ctx, env} · chat {messages, n, thinking} · complete {prompt, n} (the gate)
 //   · abort · download {url, name, size} · list · remove {name}
 // Messages out: log · progress · plan · loaded · piece · done · downloaded · models · removed · error
 import createModule from './dist/df-chat.mjs';  // built from ggml/build-wasm, see ggml/dist/BUILD.md
@@ -188,7 +188,7 @@ onmessage = async ({ data }) => {
     } else if (data.type === 'chat') {
       // messages: the whole conversation; prompt: one user turn (the older single-turn form)
       const reply = data.messages
-        ? await mod._df_chat_messages(str(JSON.stringify(data.messages)), data.n ?? 512)
+        ? await mod._df_chat_messages(str(JSON.stringify(data.messages)), data.n ?? 512, data.thinking === false ? 0 : 1)
         : await mod._df_chat(str(data.prompt), data.n ?? 64);
       post('done', { text: text(reply), stats: JSON.parse(text(mod._df_stats())) });
     }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "diskformer",
-  "version": "0.5.4",
+  "version": "0.6.4",
   "description": "Run a model larger than your GPU memory in a browser tab: weights stay in OPFS, WebGPU holds only what each step needs. llama.cpp with paged MoE experts (diskformer/llama), a disk tier, and hand-written MoE engines.",
   "type": "module",
   "license": "MIT",