Phase 1: 去 Gemini，AI 改走 OpenAI 兼容 provider（默认硅基流动）

- 新增 modules/ai_provider.py：OpenAI兼容抽象层，文本+视觉容错链+600px压缩+懒加载，
任何 OpenAI 兼容供应商(SiliconFlow/OpenAI/DeepSeek/本地vLLM)都能接
- chat.py 去 google-generativeai/SAFETY_SETTINGS/_safe_text，7个函数走 provider，保留 .ilang prompt 与全部行为
- config.py：Gemini → AI_API_KEY/AI_BASE_URL/AI_MODEL/AI_VISION_MODELS(env)
- bot.py 启动AI测试改走 provider；requirements 换 openai+Pillow
- 实测：中英双语聊天 + 判spam(广告→True/闲聊→False)均正确

bot.py

@@ -487,14 +487,10 @@ def main():
         # Polling mode: run AI test first, then start
         async def _test_ai():
             try:
-                from modules.chat import model, _safe_text
-                r = await model.generate_content_async("Say hi in one word. JSON: {\"intent\":\"chat\",\"reply\":\"hi\"}", safety_settings=[
-                    {"category": "HARM_CATEGORY_HARASSMENT", "threshold": "BLOCK_NONE"},
-                    {"category": "HARM_CATEGORY_HATE_SPEECH", "threshold": "BLOCK_NONE"},
-                    {"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT", "threshold": "BLOCK_NONE"},
-                    {"category": "HARM_CATEGORY_DANGEROUS_CONTENT", "threshold": "BLOCK_NONE"},
-                ])
-                text = _safe_text(r)
+                from modules import ai_provider
+                text = await ai_provider.generate_text(
+                    "Say hi in one word. JSON: {\"intent\":\"chat\",\"reply\":\"hi\"}",
+                    max_tokens=50)
                 if text:
                     logger.info("AI startup test OK: " + text[:100])
                 else:

config.py

@@ -3,9 +3,16 @@ import os
 # Telegram Bot
 BOT_TOKEN = os.environ.get("BOT_TOKEN", "")
 
-# Gemini AI
-GEMINI_API_KEY = os.environ.get("GEMINI_API_KEY", "")
-GEMINI_MODEL = os.environ.get("GEMINI_MODEL", "gemini-2.5-flash")
+# AI provider — OpenAI-compatible (SiliconFlow / OpenAI / DeepSeek / local vLLM / any relay)
+AI_API_KEY = os.environ.get("AI_API_KEY", "")
+AI_BASE_URL = os.environ.get("AI_BASE_URL", "https://api.siliconflow.cn/v1")
+AI_MODEL = os.environ.get("AI_MODEL", "deepseek-ai/DeepSeek-V4-Flash")
+AI_VISION_MODELS = os.environ.get(
+    "AI_VISION_MODELS",
+    "Qwen/Qwen3-VL-30B-A3B-Instruct,Qwen/Qwen3-VL-32B-Instruct,Qwen/Qwen3-VL-8B-Instruct",
+)
+AI_AUDIO_MODEL = os.environ.get("AI_AUDIO_MODEL", "Qwen/Qwen3-Omni-30B-A3B-Instruct")
+AI_IMAGE_MAX_WIDTH = int(os.environ.get("AI_IMAGE_MAX_WIDTH", "600"))
 
 # Database
 DB_PATH = os.environ.get("DB_PATH", "/data/bot.db")

modules/ai_provider.py

@@ -0,0 +1,150 @@
+"""
+AI provider layer — OpenAI-compatible.
+
+TelegramGuard talks to any OpenAI-compatible chat API: SiliconFlow (default),
+OpenAI, DeepSeek, a local vLLM/Ollama, or any relay. Business code never imports
+a vendor SDK — it calls generate_text / generate_vision / generate_audio here.
+
+Config (all via env, see config.py):
+  AI_API_KEY        — your key
+  AI_BASE_URL       — endpoint (default https://api.siliconflow.cn/v1)
+  AI_MODEL          — text model
+  AI_VISION_MODELS  — comma-separated vision fallback chain (first fails -> next)
+  AI_AUDIO_MODEL    — audio-capable model (optional; voice degrades gracefully)
+  AI_IMAGE_MAX_WIDTH — downscale images before upload (default 600)
+
+Design: lazy client init (no network at import, no crash on missing key),
+images downscaled before upload, vision falls back down the model chain.
+"""
+
+import base64
+import io
+import logging
+
+import config
+
+logger = logging.getLogger(__name__)
+
+try:
+    from openai import AsyncOpenAI
+except ImportError:
+    AsyncOpenAI = None
+
+try:
+    from PIL import Image
+except ImportError:
+    Image = None
+
+
+class AIError(Exception):
+    pass
+
+
+_client = None
+
+
+def _get_client():
+    global _client
+    if AsyncOpenAI is None:
+        raise AIError("openai package not installed (pip install openai)")
+    if _client is None:
+        key = getattr(config, "AI_API_KEY", "")
+        base = getattr(config, "AI_BASE_URL", "https://api.siliconflow.cn/v1")
+        if not key:
+            raise AIError("AI_API_KEY not set")
+        _client = AsyncOpenAI(api_key=key, base_url=base, timeout=90)
+    return _client
+
+
+def _text_model():
+    return getattr(config, "AI_MODEL", "deepseek-ai/DeepSeek-V4-Flash")
+
+
+def _vision_models():
+    raw = getattr(config, "AI_VISION_MODELS", "")
+    if raw:
+        return [m.strip() for m in raw.split(",") if m.strip()]
+    return [
+        "Qwen/Qwen3-VL-30B-A3B-Instruct",
+        "Qwen/Qwen3-VL-32B-Instruct",
+        "Qwen/Qwen3-VL-8B-Instruct",
+    ]
+
+
+def _audio_model():
+    return getattr(config, "AI_AUDIO_MODEL", "Qwen/Qwen3-Omni-30B-A3B-Instruct")
+
+
+def _compress(image_bytes):
+    max_w = getattr(config, "AI_IMAGE_MAX_WIDTH", 600)
+    if Image is None or not max_w:
+        return image_bytes
+    try:
+        img = Image.open(io.BytesIO(image_bytes))
+        if img.mode not in ("RGB", "L"):
+            img = img.convert("RGB")
+        if img.width > max_w:
+            h = max(1, int(img.height * max_w / img.width))
+            img = img.resize((max_w, h))
+        out = io.BytesIO()
+        img.save(out, format="JPEG", quality=85)
+        return out.getvalue()
+    except Exception as e:
+        logger.warning("image compress failed, using original: " + str(e))
+        return image_bytes
+
+
+def _image_part(image_bytes):
+    b64 = base64.b64encode(_compress(image_bytes)).decode()
+    return {"type": "image_url", "image_url": {"url": "data:image/jpeg;base64," + b64}}
+
+
+async def _chat(models, messages, max_tokens, temperature):
+    client = _get_client()
+    if isinstance(models, str):
+        models = [models]
+    last_err = None
+    for m in models:
+        try:
+            resp = await client.chat.completions.create(
+                model=m, messages=messages, max_tokens=max_tokens, temperature=temperature
+            )
+            txt = (resp.choices[0].message.content or "").strip()
+            if txt:
+                return txt
+            last_err = AIError("empty response")
+        except Exception as e:
+            last_err = e
+            logger.warning("[AI] model=%s failed: %s", m, e)
+    raise AIError("all models failed: " + str(last_err))
+
+
+async def generate_text(prompt, system=None, max_tokens=800, temperature=0.7):
+    msgs = []
+    if system:
+        msgs.append({"role": "system", "content": system})
+    msgs.append({"role": "user", "content": prompt})
+    return await _chat(_text_model(), msgs, max_tokens, temperature)
+
+
+async def generate_vision(prompt, image_bytes, system=None, max_tokens=800, temperature=0.4):
+    msgs = []
+    if system:
+        msgs.append({"role": "system", "content": system})
+    msgs.append({"role": "user", "content": [
+        {"type": "text", "text": prompt},
+        _image_part(image_bytes),
+    ]})
+    return await _chat(_vision_models(), msgs, max_tokens, temperature)
+
+
+async def generate_audio(prompt, audio_bytes, fmt="ogg", system=None, max_tokens=800, temperature=0.7):
+    b64 = base64.b64encode(audio_bytes).decode()
+    msgs = []
+    if system:
+        msgs.append({"role": "system", "content": system})
+    msgs.append({"role": "user", "content": [
+        {"type": "text", "text": prompt},
+        {"type": "input_audio", "input_audio": {"data": b64, "format": fmt}},
+    ]})
+    return await _chat(_audio_model(), msgs, max_tokens, temperature)

modules/chat.py

@@ -2,20 +2,12 @@ import json
 import logging
 import random
 import os
-import google.generativeai as genai
+
+from modules import ai_provider
 import config
 
 logger = logging.getLogger(__name__)
 
-genai.configure(api_key=config.GEMINI_API_KEY)
-
-# Relax Gemini safety filters
-SAFETY_SETTINGS = [
-    {"category": "HARM_CATEGORY_HARASSMENT", "threshold": "BLOCK_NONE"},
-    {"category": "HARM_CATEGORY_HATE_SPEECH", "threshold": "BLOCK_NONE"},
-    {"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT", "threshold": "BLOCK_NONE"},
-    {"category": "HARM_CATEGORY_DANGEROUS_CONTENT", "threshold": "BLOCK_NONE"},
-]
 
 # Load prompts from .ilang files (prompts/ if exists, else prompts_demo/)
 def _load_prompt(name):
@@ -29,6 +21,7 @@ def _load_prompt(name):
     logger.warning("Prompt not found: " + name)
     return ""
 
+
 SYSTEM_PROMPT = _load_prompt("persona.ilang")
 ANTISPAM_TEXT_PROMPT = _load_prompt("antispam.ilang")
 VISION_PROMPT = _load_prompt("vision.ilang")
@@ -44,16 +37,6 @@ GROUP_WELCOME = (
     "Just give me admin permissions (delete messages + ban users)."
 )
 
-model = genai.GenerativeModel(
-    config.GEMINI_MODEL,
-    system_instruction=SYSTEM_PROMPT,
-    safety_settings=SAFETY_SETTINGS
-)
-vision_model = genai.GenerativeModel(
-    config.GEMINI_MODEL,
-    safety_settings=SAFETY_SETTINGS
-)
-
 
 def _parse(raw):
     if not raw:
@@ -97,24 +80,6 @@ def _ctx(history, info):
     return "\n".join(parts)
 
 
-def _safe_text(response):
-    """Safely extract text from Gemini response — r.text throws ValueError when blocked."""
-    try:
-        if response.text:
-            return response.text.strip()
-    except (ValueError, AttributeError):
-        pass
-    # Try extracting from candidates
-    try:
-        if response.candidates:
-            for c in response.candidates:
-                if hasattr(c, 'content') and c.content and c.content.parts:
-                    return c.content.parts[0].text.strip()
-    except Exception:
-        pass
-    return ""
-
-
 def _deflect():
     lines = [
         "That's a tough one. What else can I help with?",
@@ -128,19 +93,8 @@ async def ai_text(text, history=None, context_info=""):
     try:
         c = _ctx(history, context_info)
         prompt = c + "\nuser: " + text if c else "user: " + text
-        r = await model.generate_content_async(prompt, safety_settings=SAFETY_SETTINGS)
-        raw = _safe_text(r)
+        raw = await ai_provider.generate_text(prompt, system=SYSTEM_PROMPT)
         if not raw:
-            feedback = ""
-            if hasattr(r, 'prompt_feedback'):
-                feedback = str(r.prompt_feedback)
-            if hasattr(r, 'candidates') and r.candidates:
-                for cand in r.candidates:
-                    if hasattr(cand, 'finish_reason'):
-                        feedback += " finish:" + str(cand.finish_reason)
-                    if hasattr(cand, 'safety_ratings'):
-                        feedback += " safety:" + str(cand.safety_ratings)
-            logger.warning("AI empty response. feedback=" + feedback + " prompt_len=" + str(len(prompt)))
             return ("chat", None, _deflect())
         logger.info("AI raw[" + str(len(raw)) + "]: " + raw[:200])
         return _parse(raw)
@@ -155,8 +109,8 @@ async def ai_vision(image_bytes, caption="", history=None, context_info=""):
         prompt = VISION_PROMPT + "\n" + c
         if caption:
             prompt += "\nuser: " + caption
-        r = await vision_model.generate_content_async([prompt, {"mime_type": "image/jpeg", "data": image_bytes}], safety_settings=SAFETY_SETTINGS)
-        return _parse(_safe_text(r))
+        raw = await ai_provider.generate_vision(prompt, image_bytes, system=SYSTEM_PROMPT)
+        return _parse(raw)
     except Exception as e:
         logger.warning("AI vision: " + str(e))
         return ("chat", None, "Couldn't read that image. Try another one?")
@@ -165,9 +119,12 @@ async def ai_vision(image_bytes, caption="", history=None, context_info=""):
 async def ai_voice(audio_bytes, mime_type="audio/ogg", history=None, context_info=""):
     try:
         c = _ctx(history, context_info)
-        prompt = SYSTEM_PROMPT + "\n" + c + "\nUser sent a voice message:"
-        r = await vision_model.generate_content_async([prompt, {"mime_type": mime_type, "data": audio_bytes}], safety_settings=SAFETY_SETTINGS)
-        return _parse(_safe_text(r))
+        prompt = c + "\nUser sent a voice message:" if c else "User sent a voice message:"
+        fmt = "ogg"
+        if "/" in mime_type:
+            fmt = mime_type.split("/", 1)[1].split(";")[0] or "ogg"
+        raw = await ai_provider.generate_audio(prompt, audio_bytes, fmt=fmt, system=SYSTEM_PROMPT)
+        return _parse(raw)
     except Exception as e:
         logger.warning("AI voice: " + str(e))
         return ("chat", None, "Didn't catch that. Try again or type it out.")
@@ -176,9 +133,8 @@ async def ai_voice(audio_bytes, mime_type="audio/ogg", history=None, context_inf
 async def ai_judge_group_message(text):
     try:
         prompt = ANTISPAM_TEXT_PROMPT + "\n\nMessage content: " + text[:1000]
-        r = await vision_model.generate_content_async(prompt, safety_settings=SAFETY_SETTINGS)
-        result = (_safe_text(r) or "ok").lower()
-        return "spam" in result
+        raw = await ai_provider.generate_text(prompt, max_tokens=8, temperature=0.0)
+        return "spam" in (raw or "ok").lower()
     except Exception:
         return False
 
@@ -188,9 +144,8 @@ async def ai_judge_group_image(image_bytes, caption=""):
         prompt = ANTISPAM_TEXT_PROMPT + "\n\nJudge this image. Reply ONLY: spam or ok."
         if caption:
             prompt += "\nCaption: " + caption[:500]
-        r = await vision_model.generate_content_async([prompt, {"mime_type": "image/jpeg", "data": image_bytes}], safety_settings=SAFETY_SETTINGS)
-        result = (_safe_text(r) or "ok").lower()
-        return "spam" in result
+        raw = await ai_provider.generate_vision(prompt, image_bytes, max_tokens=8, temperature=0.0)
+        return "spam" in (raw or "ok").lower()
     except Exception:
         return False
 
@@ -203,8 +158,8 @@ async def ai_group_vision(image_bytes, caption="", history=None):
             prompt += "\nuser: " + caption
         else:
             prompt += "\nuser: [shared an image]"
-        r = await vision_model.generate_content_async([prompt, {"mime_type": "image/jpeg", "data": image_bytes}], safety_settings=SAFETY_SETTINGS)
-        raw = _safe_text(r)
+        raw = await ai_provider.generate_vision(prompt, image_bytes, system=SYSTEM_PROMPT)
+        raw = (raw or "").strip()
         if not raw:
             return _deflect()
         intent, device, reply = _parse(raw)
@@ -219,8 +174,8 @@ async def ai_group_reply(text, history=None):
     try:
         ctx = _ctx(history, "GROUP_CHAT: You were @mentioned in a group. Reply directly, 1-2 sentences.")
         prompt = ctx + "\nuser: " + text
-        r = await model.generate_content_async(prompt, safety_settings=SAFETY_SETTINGS)
-        raw = _safe_text(r)
+        raw = await ai_provider.generate_text(prompt, system=SYSTEM_PROMPT)
+        raw = (raw or "").strip()
         if not raw:
             return _deflect()
         intent, device, reply = _parse(raw)

requirements.txt

@@ -1,3 +1,4 @@
 python-telegram-bot[job-queue]>=21.0
-google-generativeai==0.8.4
+openai>=1.40
+Pillow>=10.0
 aiosqlite>=0.20.0