Spaces:
Sleeping
Sleeping
Update agent_service.py
Browse files- agent_service.py +8 -10
agent_service.py
CHANGED
|
@@ -248,27 +248,25 @@ Example:
|
|
| 248 |
|
| 249 |
async def _call_llm(self, messages: List[Dict]) -> str:
|
| 250 |
"""
|
| 251 |
-
Call HuggingFace LLM directly using
|
| 252 |
"""
|
| 253 |
try:
|
| 254 |
from huggingface_hub import AsyncInferenceClient
|
| 255 |
|
| 256 |
-
# Build prompt from messages
|
| 257 |
-
prompt = self._messages_to_prompt(messages)
|
| 258 |
-
|
| 259 |
# Create async client
|
| 260 |
client = AsyncInferenceClient(token=self.hf_token)
|
| 261 |
|
| 262 |
-
# Call HF API
|
| 263 |
response_text = ""
|
| 264 |
-
async for
|
| 265 |
-
|
| 266 |
-
model="
|
| 267 |
-
|
| 268 |
temperature=0.7,
|
| 269 |
stream=True
|
| 270 |
):
|
| 271 |
-
|
|
|
|
| 272 |
|
| 273 |
return response_text
|
| 274 |
except Exception as e:
|
|
|
|
| 248 |
|
| 249 |
async def _call_llm(self, messages: List[Dict]) -> str:
|
| 250 |
"""
|
| 251 |
+
Call HuggingFace LLM directly using chat_completion (conversational)
|
| 252 |
"""
|
| 253 |
try:
|
| 254 |
from huggingface_hub import AsyncInferenceClient
|
| 255 |
|
|
|
|
|
|
|
|
|
|
| 256 |
# Create async client
|
| 257 |
client = AsyncInferenceClient(token=self.hf_token)
|
| 258 |
|
| 259 |
+
# Call HF API with chat completion (conversational)
|
| 260 |
response_text = ""
|
| 261 |
+
async for message in await client.chat_completion(
|
| 262 |
+
messages=messages, # Use messages directly
|
| 263 |
+
model="meta-llama/Llama-3.3-70B-Instruct",
|
| 264 |
+
max_tokens=512,
|
| 265 |
temperature=0.7,
|
| 266 |
stream=True
|
| 267 |
):
|
| 268 |
+
if message.choices and message.choices[0].delta.content:
|
| 269 |
+
response_text += message.choices[0].delta.content
|
| 270 |
|
| 271 |
return response_text
|
| 272 |
except Exception as e:
|