feat: +zhipu and spark support

pilot/model/proxy/llms/spark.py

@@ -4,7 +4,6 @@ import base64
 import hmac
 import hashlib
 import websockets 
-import asyncio
 from datetime import datetime
 from typing import List
 from time import mktime
@@ -73,13 +72,12 @@ def spark_generate_stream(
         },
         "payload": {
             "message": {
-                "text": last_user_input.get("") 
+                "text": last_user_input.get("content") 
             }
         }
     }
 
-    # TODO
+    async_call(request_url, data)
-     
 
 async def async_call(request_url, data):
     async with websockets.connect(request_url) as ws:

pilot/model/proxy/llms/zhipu.py

@@ -1,5 +1,11 @@
-from pilot.model.proxy.llms.proxy_model import ProxyModel
+import os
+import json
+from typing import List
 
+from pilot.model.proxy.llms.proxy_model import ProxyModel
+from pilot.scene.base_message import ModelMessage, ModelMessageRoleType
+
+CHATGLM_DEFAULT_MODEL = "chatglm_pro"
 
 def zhipu_generate_stream(
     model: ProxyModel, tokenizer, params, device, context_len=2048
@@ -9,10 +15,33 @@ def zhipu_generate_stream(
     print(f"Model: {model}, model_params: {model_params}")
 
     proxy_api_key = model_params.proxy_api_key
-    proxy_server_url = model_params.proxy_server_url
+    proxyllm_backend = CHATGLM_DEFAULT_MODEL or model_params.proxyllm_backend 
-    proxyllm_backend = model_params.proxyllm_backend
 
-    if not proxyllm_backend:
+    import zhipuai
-        proxyllm_backend = "chatglm_pro"
+    zhipuai.api_key = proxy_api_key
-    # TODO
+
-    yield "Zhipu LLM was not supported!"
+    history = []
+
+    messages: List[ModelMessage] = params["messages"]
+    # Add history conversation
+
+    for message in messages:
+        if message.role == ModelMessageRoleType.HUMAN:
+            history.append({"role": "user", "content": message.content})
+        elif message.role == ModelMessageRoleType.SYSTEM:
+            history.append({"role": "system", "content": message.content})
+        elif message.role == ModelMessageRoleType.AI:
+            history.append({"role": "assistant", "content": message.content})
+        else:
+            pass
+
+    res = zhipuai.model_api.sse_invoke(
+        model=proxyllm_backend,
+        prompt=history,
+        temperature=params.get("temperature"),
+        top_p=params.get("top_p"),
+        incremental=False,
+    )
+    for r in res.events():
+        if r.event == "add":
+            yield r.data