Add files via upload
This commit is contained in:
1 parent
5d8d214e9b
commit
fb46b5e6a1
6 files changed
+129
-22
No files matched your search
@@ -11,13 +11,22 @@ import threading
|
||||
from pathlib import Path
|
||||
from typing import Optional
|
||||
|
||||
from playwright.async_api import (
|
||||
async_playwright,
|
||||
BrowserContext,
|
||||
Page,
|
||||
Playwright,
|
||||
TimeoutError as PlaywrightTimeout,
|
||||
)
|
||||
try:
|
||||
from playwright.async_api import (
|
||||
async_playwright,
|
||||
BrowserContext,
|
||||
Page,
|
||||
Playwright,
|
||||
TimeoutError as PlaywrightTimeout,
|
||||
)
|
||||
_PLAYWRIGHT_OK = True
|
||||
except ImportError:
|
||||
_PLAYWRIGHT_OK = False
|
||||
async_playwright = None # type: ignore[assignment]
|
||||
BrowserContext = None # type: ignore[assignment]
|
||||
Page = None # type: ignore[assignment]
|
||||
Playwright = None # type: ignore[assignment]
|
||||
PlaywrightTimeout = Exception # type: ignore[assignment,misc]
|
||||
_OS = platform.system() # "Windows" | "Darwin" | "Linux"
|
||||
|
||||
def _normalize_url(url: str) -> str:
|
||||
@@ -807,6 +816,14 @@ def browser_control(
|
||||
player=None,
|
||||
session_memory=None,
|
||||
) -> str:
|
||||
if not _PLAYWRIGHT_OK:
|
||||
msg = (
|
||||
"Playwright is not installed.\n"
|
||||
"Fix: pip install playwright && python -m playwright install chromium"
|
||||
)
|
||||
_log(player, msg)
|
||||
return msg
|
||||
|
||||
params = parameters or {}
|
||||
action = params.get("action", "").lower().strip()
|
||||
browser = params.get("browser", "").lower().strip() or None
|
||||
|
||||
@@ -9,9 +9,6 @@ from pathlib import Path
|
||||
from datetime import datetime
|
||||
from urllib.parse import quote_plus
|
||||
|
||||
import pyautogui
|
||||
import numpy as np
|
||||
|
||||
try:
|
||||
import requests
|
||||
_REQUESTS_OK = True
|
||||
|
||||
+35
-1
@@ -22,7 +22,7 @@ import subprocess
|
||||
import sys
|
||||
import time
|
||||
from pathlib import Path
|
||||
from typing import Generator
|
||||
from typing import Callable, Generator
|
||||
|
||||
import requests
|
||||
|
||||
@@ -184,6 +184,40 @@ def warmup_model(system_prompt: str | None = None) -> bool:
|
||||
return False
|
||||
|
||||
|
||||
def check_model_available(log: Callable | None = None) -> bool:
|
||||
"""
|
||||
Returns True if the configured model is already pulled in Ollama.
|
||||
Logs an actionable warning (to console + optional UI callback) if not.
|
||||
Always returns True for non-Ollama providers (cannot inspect their model list).
|
||||
"""
|
||||
if get_llm_provider() != "ollama":
|
||||
return True
|
||||
|
||||
url, model = get_llm_settings()
|
||||
try:
|
||||
resp = requests.get(f"{url}/api/tags", timeout=5)
|
||||
resp.raise_for_status()
|
||||
pulled = [m.get("name", "") for m in resp.json().get("models", [])]
|
||||
model_base = model.split(":")[0]
|
||||
found = any(
|
||||
m == model or m == model_base or m.startswith(model_base + ":")
|
||||
for m in pulled
|
||||
)
|
||||
if not found:
|
||||
available = ", ".join(pulled) if pulled else "none"
|
||||
warn = (
|
||||
f"WRN: Model '{model}' is not pulled in Ollama.\n"
|
||||
f" Available: {available}\n"
|
||||
f" Fix: ollama pull {model}"
|
||||
)
|
||||
print(warn)
|
||||
if log:
|
||||
log(f"WRN: '{model}' not found — run: ollama pull {model}")
|
||||
return found
|
||||
except Exception:
|
||||
return True # Ollama might still be starting up; non-blocking
|
||||
|
||||
|
||||
def get_llm_settings() -> tuple[str, str]:
|
||||
"""Returns (base_url, model_name)."""
|
||||
cfg = _load_config()
|
||||
|
||||
+17
-4
@@ -25,14 +25,27 @@ class WhisperSTT:
|
||||
try:
|
||||
self._model = WhisperModel(model_name, device=device, compute_type=compute)
|
||||
except Exception as _first_err:
|
||||
# Offline flag set but model not cached yet → download once, then offline forever
|
||||
# Offline flag set but model not cached yet → clear flags and download once.
|
||||
# Keywords cover multiple huggingface_hub error message variants across versions.
|
||||
_e = str(_first_err).lower()
|
||||
if any(k in _e for k in ("offline", "not found", "cache", "localentry", "does not exist")):
|
||||
print(f"[STT] '{model_name}' not cached — downloading (internet required for first run)…")
|
||||
_offline_keywords = (
|
||||
"offline", "not found", "cache", "localentry",
|
||||
"does not exist", "outgoing", "local_files_only",
|
||||
)
|
||||
if any(k in _e for k in _offline_keywords):
|
||||
print(f"[STT] Whisper '{model_name}' not in local cache — downloading (one-time, internet required)…")
|
||||
os.environ.pop("HF_HUB_OFFLINE", None)
|
||||
os.environ.pop("TRANSFORMERS_OFFLINE", None)
|
||||
os.environ.pop("HF_DATASETS_OFFLINE", None)
|
||||
self._model = WhisperModel(model_name, device=device, compute_type=compute)
|
||||
try:
|
||||
self._model = WhisperModel(model_name, device=device, compute_type=compute)
|
||||
except Exception as _dl_err:
|
||||
raise RuntimeError(
|
||||
f"Whisper '{model_name}' model download failed.\n"
|
||||
f"Internet access is required the first time to download the speech model (~75–290 MB).\n"
|
||||
f"After the first download it runs fully offline.\n"
|
||||
f"Details: {_dl_err}"
|
||||
) from _dl_err
|
||||
else:
|
||||
raise
|
||||
|
||||
|
||||
+18
-4
@@ -273,14 +273,28 @@ class KokoroTTSEngine:
|
||||
try:
|
||||
self._pipeline = _create_pipeline()
|
||||
except Exception as _first_err:
|
||||
# Offline flag set but model not cached yet → download once
|
||||
# Offline flag set but model not cached yet → clear flags and download once.
|
||||
# Keywords cover multiple huggingface_hub error message variants across versions.
|
||||
_e = str(_first_err).lower()
|
||||
if any(k in _e for k in ("offline", "not found", "cache", "localentry", "does not exist")):
|
||||
print("[TTS] Kokoro model not cached — downloading (internet required for first run)…")
|
||||
_offline_keywords = (
|
||||
"offline", "not found", "cache", "localentry",
|
||||
"does not exist", "outgoing", "local_files_only",
|
||||
)
|
||||
if any(k in _e for k in _offline_keywords):
|
||||
print("[TTS] Kokoro model not in local cache — downloading (one-time, internet required)…")
|
||||
os.environ.pop("HF_HUB_OFFLINE", None)
|
||||
os.environ.pop("TRANSFORMERS_OFFLINE", None)
|
||||
os.environ.pop("HF_DATASETS_OFFLINE", None)
|
||||
self._pipeline = _create_pipeline()
|
||||
try:
|
||||
self._pipeline = _create_pipeline()
|
||||
except Exception as _dl_err:
|
||||
raise RuntimeError(
|
||||
f"Kokoro model download failed.\n"
|
||||
f"Internet access is required the first time to download the voice model (~330 MB).\n"
|
||||
f"After the first download it runs fully offline.\n"
|
||||
f"Tip: Switch to EdgeTTS (free, no download) in the Configure panel if offline.\n"
|
||||
f"Details: {_dl_err}"
|
||||
) from _dl_err
|
||||
else:
|
||||
raise
|
||||
|
||||
|
||||
@@ -1138,6 +1138,11 @@ class JarvisLocal:
|
||||
except Exception as e:
|
||||
print(f"[STT-Whisper] Mic error: {e}")
|
||||
traceback.print_exc()
|
||||
self.ui.write_log(
|
||||
f"ERR: Microphone — {e}\n"
|
||||
" Check: microphone plugged in, OS privacy permissions granted, "
|
||||
"correct audio device selected in system settings."
|
||||
)
|
||||
|
||||
def _listen_vosk(self) -> None:
|
||||
"""Mic → Vosk streaming → LLM loop."""
|
||||
@@ -1172,6 +1177,11 @@ class JarvisLocal:
|
||||
except Exception as e:
|
||||
print(f"[STT-Vosk] Mic error: {e}")
|
||||
traceback.print_exc()
|
||||
self.ui.write_log(
|
||||
f"ERR: Microphone — {e}\n"
|
||||
" Check: microphone plugged in, OS privacy permissions granted, "
|
||||
"correct audio device selected in system settings."
|
||||
)
|
||||
|
||||
# ------------------------------------------------------------------
|
||||
# Text command loop (UI input box)
|
||||
@@ -1203,10 +1213,11 @@ class JarvisLocal:
|
||||
self.ui.on_reconfigure = self.reconfigure
|
||||
|
||||
# ── Ollama ────────────────────────────────────────────────────
|
||||
from core.llm_client import ensure_ollama_running, warmup_model
|
||||
from core.llm_client import ensure_ollama_running, warmup_model, check_model_available
|
||||
self.ui.write_log("SYS: Checking Ollama…")
|
||||
if ensure_ollama_running():
|
||||
self.ui.write_log("SYS: Ollama OK.")
|
||||
check_model_available(log=self.ui.write_log)
|
||||
else:
|
||||
self.ui.write_log("ERR: Ollama unavailable — run: ollama serve")
|
||||
|
||||
@@ -1230,8 +1241,29 @@ class JarvisLocal:
|
||||
# static prefix → Ollama reuses cached KV → first token <1 s
|
||||
# instead of the ~17 s it takes to re-evaluate 300+ tokens cold.
|
||||
static_prompt = _load_system_prompt()
|
||||
warmup_model(system_prompt=static_prompt)
|
||||
self.ui.write_log("SYS: LLM ready.")
|
||||
ok = warmup_model(system_prompt=static_prompt)
|
||||
if ok:
|
||||
self.ui.write_log("SYS: LLM ready.")
|
||||
else:
|
||||
# warmup() returned False — check if model is missing
|
||||
from core.llm_client import check_model_available
|
||||
if not check_model_available(log=self.ui.write_log):
|
||||
pass # warning already written by check_model_available
|
||||
else:
|
||||
self.ui.write_log("WRN: LLM warmup failed — is Ollama running?")
|
||||
# GPU usage tip (cross-platform, non-blocking)
|
||||
try:
|
||||
import torch
|
||||
if torch.cuda.is_available():
|
||||
self.ui.write_log(f"SYS: GPU detected — {torch.cuda.get_device_name(0)}")
|
||||
else:
|
||||
self.ui.write_log(
|
||||
"WRN: CUDA not detected — LLM running on CPU (slow). "
|
||||
"For GPU speed install CUDA PyTorch: "
|
||||
"pip install torch --index-url https://download.pytorch.org/whl/cu118"
|
||||
)
|
||||
except Exception:
|
||||
pass # torch not installed or not applicable
|
||||
self.ui.mark_startup_ready("llm")
|
||||
except Exception as e:
|
||||
self.ui.write_log(f"ERR: LLM warmup — {e}")
|
||||
|
||||
Reference in new issue
Block a user