fix(build): cap parallel compilation threads to prevent CPU lockup

Author: Katsoragi

routes/cookbook_helpers.py

@@ -359,9 +359,20 @@ def _user_shell_path_bootstrap() -> list[str]:
     return [
         'ODYSSEUS_USER_SHELL="${SHELL:-}"',
         'if [ -n "$ODYSSEUS_USER_SHELL" ] && [ -x "$ODYSSEUS_USER_SHELL" ]; then',
-        '  ODYSSEUS_USER_PATH="$("$ODYSSEUS_USER_SHELL" -ic \'printf "__ODYSSEUS_PATH__%s\\n" "$PATH"\' 2>/dev/null | sed -n \'s/^__ODYSSEUS_PATH__//p\' | tail -n 1 || true)"',
+        '  case "$ODYSSEUS_USER_SHELL" in',
+        '    *fish*) ODYSSEUS_USER_PATH="$("$ODYSSEUS_USER_SHELL" -c \'printf "__ODYSSEUS_PATH__%s\\n" "$PATH"\' 2>/dev/null | sed -n \'s/^__ODYSSEUS_PATH__//p\' | tail -n 1 || true)" ;;',
+        '    *) ODYSSEUS_USER_PATH="$("$ODYSSEUS_USER_SHELL" -lc \'printf "__ODYSSEUS_PATH__%s\\n" "$PATH"\' 2>/dev/null | sed -n \'s/^__ODYSSEUS_PATH__//p\' | tail -n 1 || true)" ;;',
+        '  esac',
         '  if [ -n "$ODYSSEUS_USER_PATH" ]; then export PATH="$ODYSSEUS_USER_PATH:$PATH"; fi',
         'fi',
+        'if command -v nproc >/dev/null 2>&1; then',
+        '  _n="$(nproc)"',
+        '  export CMAKE_BUILD_PARALLEL_LEVEL="$(( _n > 2 ? _n - 2 : 1 ))"',
+        '  export NPROC="$(( _n > 2 ? _n - 2 : 1 ))"',
+        'else',
+        '  export CMAKE_BUILD_PARALLEL_LEVEL=2',
+        '  export NPROC=2',
+        'fi',
         # Windows can expose python3 as a Microsoft Store App Execution Alias
         # under WindowsApps. Git Bash sees that stub as present, but it exits
         # before running Python. A Windows venv usually has python.exe, not
@@ -378,6 +389,7 @@ def _cached_model_scan_script(model_dirs: list[str] | None = None, add_hf_cache:
     """
     lines = [
         "import json, os, re, shutil, subprocess, urllib.request",
+        f"model_dirs = {model_dirs or []!r}",
         "models = []",
         "seen = set()",
         "BLOCKED_ROOTS = ('/sys', '/proc', '/dev', '/run', '/var/run')",
@@ -484,6 +496,12 @@ def _cached_model_scan_script(model_dirs: list[str] | None = None, add_hf_cache:
         "    # Docker images mount ./data/huggingface at /app/.cache/huggingface.",
         "    # When HOME is /root, expanduser() misses that persisted cache.",
         "    add('/app/.cache/huggingface/hub')",
+        "    for p in model_dirs:",
+        "        try:",
+        "            ep = os.path.expanduser(p)",
+        "            if os.path.isdir(ep) and any(d.startswith('models--') for d in os.listdir(ep)):",
+        "                add(p)",
+        "        except Exception: pass",
         f"    add({add_hf_cache!r})" if add_hf_cache else "",
         "    return candidates",
         "def normalize_model_dir(p):",
@@ -563,12 +581,11 @@ def _cached_model_scan_script(model_dirs: list[str] | None = None, add_hf_cache:
         "            models.append({'repo_id':name,'size_bytes':size_bytes,'nb_files':1,'has_incomplete':False,'path':'ollama','backend':'ollama','is_ollama':True})",
         "        return",
         "for _hf_cache in hf_cache_paths(): scan_hf(_hf_cache)",
+        "for _md in model_dirs: scan_dir(os.path.expanduser(_md))",
         "scan_ollama_api()",
         "scan_ollama()",
+        "print(json.dumps(models))",
     ]
-    for model_dir in model_dirs or []:
-        lines.append(f"scan_dir({model_dir!r})")
-    lines.append("print(json.dumps(models))")
     return "\n".join(lines) + "\n"
 
 

routes/cookbook_routes.py

@@ -2122,7 +2122,7 @@ async def model_serve(request: Request, req: ServeRequest):
                 # else a plain CPU build. nproc is Linux-only — fall back to
                 # `sysctl hw.ncpu` on macOS. (Tip: `brew install llama.cpp` ships
                 # a prebuilt llama-server and skips this whole source build.)
-                runner_lines.append('  NPROC="$(nproc 2>/dev/null || sysctl -n hw.ncpu 2>/dev/null || echo 4)"')
+                runner_lines.append('  _n="$(nproc 2>/dev/null || sysctl -n hw.ncpu 2>/dev/null || echo 4)"; NPROC="$(( _n > 2 ? _n - 2 : 1 ))"')
                 runner_lines.append('  if [ "$(uname -s)" = "Darwin" ]; then')
                 runner_lines.append('    command -v cmake >/dev/null 2>&1 || echo "WARNING: cmake not found — install it with: brew install cmake (or: brew install llama.cpp for a prebuilt llama-server)."')
                 # Start from a clean cache: a prior failed configure (e.g. a CUDA

src/llm_core.py

@@ -2529,6 +2529,11 @@ def _format_routed_content(parts: List[Tuple[str, bool]]) -> List[str]:
                         if data.strip():
                             if data.startswith("{"):
                                 j = json.loads(data)
+                                if "error" in j:
+                                    err = j["error"] or {}
+                                    text = err.get("message") if isinstance(err, dict) else str(err)
+                                    yield f'event: error\ndata: {json.dumps({"status": 400, "text": text})}\n\n'
+                                    return
                                 chunk_model = j.get("model")
                                 if isinstance(chunk_model, str) and chunk_model.strip():
                                     _actual_model = chunk_model.strip()

src/model_context.py

@@ -460,10 +460,10 @@ def _query_context_length(endpoint_url: str, model: str) -> Tuple[int, bool]:
     except Exception as e:
         logger.debug(f"Failed to query context length for {model}: {e}")
 
+    _is_local = is_local_endpoint(endpoint_url)
     # For local/self-hosted endpoints, trust the API value (user set --max-model-len)
     # For cloud APIs, use the larger value (API can report low defaults)
     if api_ctx and known:
-        _is_local = is_local_endpoint(endpoint_url)
         if _is_local and api_ctx < known:
             logger.info(f"Local endpoint reports {api_ctx} for {model} (known max: {known}) — using API value")
             return api_ctx, True
@@ -474,9 +474,21 @@ def _query_context_length(endpoint_url: str, model: str) -> Tuple[int, bool]:
     if api_ctx:
         return api_ctx, True
     if known:
+        if _is_local:
+            import os as _os
+            env_val = _os.environ.get("ODYSSEUS_LOCAL_CONTEXT")
+            try:
+                limit = int(env_val) if env_val else 4096
+            except (ValueError, TypeError):
+                limit = 4096
+            val = min(known, limit)
+            logger.info(f"Using capped known context window for local endpoint {model}: {val} (known max: {known}, cap: {limit})")
+            return val, True
         logger.info(f"Using known context window for {model}: {known}")
         return known, True
 
+    if _is_local:
+        return 4096, False
     return DEFAULT_CONTEXT, False
 
 

static/js/cookbook-hwfit.js

@@ -2098,6 +2098,8 @@ export function _hwfitInit() {
     const selectors = [
       document.getElementById('hwfit-server-select'),
       document.getElementById('hwfit-dl-server'),
+      document.getElementById('hwfit-cache-server'),
+      document.getElementById('hwfit-deps-server'),
     ];
     for (const sel of selectors) {
       if (!sel) continue;