@lzhzzzzwill/cofos 1.1.0 → 1.1.3

package/bin/cofos.js

@@ -172,6 +172,8 @@ function main() {
     HF_HUB_ENABLE_HF_TRANSFER: process.env.HF_HUB_ENABLE_HF_TRANSFER || "0",
     HF_HUB_DOWNLOAD_TIMEOUT: process.env.HF_HUB_DOWNLOAD_TIMEOUT || "120",
     HF_HUB_ETAG_TIMEOUT: process.env.HF_HUB_ETAG_TIMEOUT || "30",
+    HF_HUB_VERBOSITY: process.env.HF_HUB_VERBOSITY || "warning",
+    TRANSFORMERS_VERBOSITY: process.env.TRANSFORMERS_VERBOSITY || "warning",
   };
   const proc = spawn(py, backendArgs, {
     stdio: ["pipe", "pipe", "pipe"],
@@ -184,7 +186,7 @@ function main() {
   proc.stdout.setEncoding("utf-8");
   proc.stderr.setEncoding("utf-8");
 
-  let buf = "", ready = false, generating = false, closed = false;
+  let buf = "", ready = false, generating = false, responseOpen = false, closed = false;
   const inputQ = [];
   const rl = createInterface({ input: process.stdin, terminal: true });
   rl.on("line", (l) => { inputQ.push(l); });
@@ -208,27 +210,35 @@ function main() {
     // dropping it keeps it from being prepended to the next one.
     if (!generating) { buf = ""; return; }
 
+    if (!responseOpen) {
+      const ri = buf.indexOf(RESPONSE);
+      if (ri === -1) {
+        // Keep enough tail to catch a RESPONSE marker split across chunks.
+        const keep = RESPONSE.length - 1;
+        if (buf.length > keep) buf = buf.slice(-keep);
+        return;
+      }
+      buf = buf.slice(ri + RESPONSE.length);
+      if (buf[0] === "\n") buf = buf.slice(1);
+      responseOpen = true;
+    }
+
     // Scan for DONE marker — stream everything before it.
     const di = buf.indexOf(DONE);
     if (di !== -1) {
-      // Flush any pre-DONE content (cleaned of RESPONSE marker).
-      let pre = buf.slice(0, di);
-      const ri = pre.indexOf(RESPONSE);
-      if (ri !== -1) pre = pre.slice(ri + RESPONSE.length);
+      const pre = buf.slice(0, di);
       if (pre) process.stdout.write(pre);
       buf = buf.slice(di + DONE.length);
       if (buf[0] === "\n") buf = buf.slice(1);
       generating = false;
+      responseOpen = false;
       console.log("");
       drain();
     } else {
-      // Stream buffered content, keeping only enough tail to catch a
-      // partial DONE marker.
-      const keep = DONE.length;
+      // Stream buffered content, keeping enough tail to catch a partial DONE marker.
+      const keep = DONE.length - 1;
       if (buf.length > keep) {
-        let out = buf.slice(0, -keep);
-        const ri = out.indexOf(RESPONSE);
-        if (ri !== -1) out = out.slice(ri + RESPONSE.length);
+        const out = buf.slice(0, -keep);
         if (out) process.stdout.write(out);
         buf = buf.slice(-keep);
       }
@@ -297,6 +307,7 @@ function main() {
     }
     if (/^\/(clear)$/.test(input)) {
       mlBuf = [];
+      responseOpen = false;
       proc.stdin.write(input + "\n");
       generating = true;
       return;
@@ -313,6 +324,7 @@ function main() {
     }
 
     // Forward to backend.
+    responseOpen = false;
     generating = true;
     proc.stdin.write(input + "\n");
   }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lzhzzzzwill/cofos",
-  "version": "1.1.0",
+  "version": "1.1.3",
   "description": "COFOS 9B + RAG chat CLI with optional local PDF ingestion.",
   "type": "module",
   "bin": {

package/scripts/chat.py

@@ -24,6 +24,8 @@ os.environ.setdefault("HF_HUB_DISABLE_XET", "1")
 os.environ.setdefault("HF_HUB_ENABLE_HF_TRANSFER", "0")
 os.environ.setdefault("HF_HUB_DOWNLOAD_TIMEOUT", "120")
 os.environ.setdefault("HF_HUB_ETAG_TIMEOUT", "30")
+os.environ.setdefault("HF_HUB_VERBOSITY", "warning")
+os.environ.setdefault("TRANSFORMERS_VERBOSITY", "warning")
 
 HISTORY_FILE = os.path.join(os.path.expanduser("~"), ".cofos_history.jsonl")
 HISTORY_MAX_TURNS = 20
@@ -338,19 +340,35 @@ def ensure_runtime_data(project_root: Path, config: dict) -> None:
         "runtime/bm25/bm25_info.json": bm25_dir,
     }
 
+    missing_items = []
     for remote_path, target_dir in file_map.items():
-        target_dir.mkdir(parents=True, exist_ok=True)
         local_name = remote_path.rsplit("/", 1)[-1]
         local_path = target_dir / local_name
-        if local_path.exists():
-            continue
+        if not local_path.exists():
+            missing_items.append((remote_path, target_dir, local_path))
+
+    total = len(missing_items)
+    if total == 0:
+        print("[chat] Runtime data ready.", file=sys.stderr, flush=True)
+        return
+
+    def progress_line(index: int, remote_path: str) -> str:
+        width = 18
+        filled = int(width * index / total)
+        bar = "#" * filled + "-" * (width - filled)
+        pct = int(100 * index / total)
+        return f"[chat] Runtime data [{index}/{total}] [{bar}] {pct:3d}% {remote_path}"
+
+    for index, (remote_path, target_dir, local_path) in enumerate(missing_items, 1):
+        target_dir.mkdir(parents=True, exist_ok=True)
+        print(progress_line(index, remote_path), file=sys.stderr, flush=True)
         cached = hf_hub_download(
             repo_id=repo,
             repo_type="dataset",
             filename=remote_path,
         )
         shutil.copy2(cached, local_path)
-        print(f"[chat] Downloaded {remote_path} -> {local_path}", file=sys.stderr, flush=True)
+        print(f"[chat]   -> {local_path}", file=sys.stderr, flush=True)
 
     print("[chat] Runtime data ready.", file=sys.stderr, flush=True)
 
@@ -446,6 +464,8 @@ def main() -> None:
     from inference.student_adapter_inference import StudentAdapterInference, StudentRAGContext
 
     logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
+    for noisy_logger in ("httpx", "httpcore", "huggingface_hub"):
+        logging.getLogger(noisy_logger).setLevel(logging.WARNING)
 
     config_path = default_config_path(runtime_root) if args.config == DEFAULT_CONFIG else resolve_project_path(runtime_root, args.config)
     model_path = resolve_project_path(runtime_root, args.model_path)
@@ -466,7 +486,7 @@ def main() -> None:
         pdf_changed = parse_pdfs_if_needed(runtime_root, config, force=args.rebuild_pdf_index)
         build_augmented_bm25_if_needed(runtime_root, config, force=args.rebuild_pdf_index or pdf_changed)
 
-    print("[chat] Loading model (this may take 1-2 min on first run) ...", file=sys.stderr, flush=True)
+    print("[chat] Loading model weights. First run may download 12GB+ and take tens of minutes; cached runs are much faster ...", file=sys.stderr, flush=True)
 
     runner = StudentAdapterInference(
         config=config,