29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249 | class Ollama(CustomLLM):
"""Ollama LLM.
Visit https://ollama.com/ to download and install Ollama.
Run `ollama serve` to start a server.
Run `ollama pull <name>` to download a model to run.
Examples:
`pip install llama-index-llms-ollama`
```python
from llama_index.llms.ollama import Ollama
llm = Ollama(model="llama2", request_timeout=60.0)
response = llm.complete("What is the capital of France?")
print(response)
```
"""
base_url: str = Field(
default="http://localhost:11434",
description="Base url the model is hosted under.",
)
model: str = Field(description="The Ollama model to use.")
temperature: float = Field(
default=0.75,
description="The temperature to use for sampling.",
gte=0.0,
lte=1.0,
)
context_window: int = Field(
default=DEFAULT_CONTEXT_WINDOW,
description="The maximum number of context tokens for the model.",
gt=0,
)
request_timeout: float = Field(
default=DEFAULT_REQUEST_TIMEOUT,
description="The timeout for making http request to Ollama API server",
)
prompt_key: str = Field(
default="prompt", description="The key to use for the prompt in API calls."
)
additional_kwargs: Dict[str, Any] = Field(
default_factory=dict,
description="Additional model parameters for the Ollama API.",
)
@classmethod
def class_name(cls) -> str:
return "Ollama_llm"
@property
def metadata(self) -> LLMMetadata:
"""LLM metadata."""
return LLMMetadata(
context_window=self.context_window,
num_output=DEFAULT_NUM_OUTPUTS,
model_name=self.model,
is_chat_model=True, # Ollama supports chat API for all models
)
@property
def _model_kwargs(self) -> Dict[str, Any]:
base_kwargs = {
"temperature": self.temperature,
"num_ctx": self.context_window,
}
return {
**base_kwargs,
**self.additional_kwargs,
}
@llm_chat_callback()
def chat(self, messages: Sequence[ChatMessage], **kwargs: Any) -> ChatResponse:
payload = {
"model": self.model,
"messages": [
{
"role": message.role.value,
"content": message.content,
**message.additional_kwargs,
}
for message in messages
],
"options": self._model_kwargs,
"stream": False,
**kwargs,
}
with httpx.Client(timeout=Timeout(self.request_timeout)) as client:
response = client.post(
url=f"{self.base_url}/api/chat",
json=payload,
)
response.raise_for_status()
raw = response.json()
message = raw["message"]
return ChatResponse(
message=ChatMessage(
content=message.get("content"),
role=MessageRole(message.get("role")),
additional_kwargs=get_additional_kwargs(
message, ("content", "role")
),
),
raw=raw,
additional_kwargs=get_additional_kwargs(raw, ("message",)),
)
@llm_chat_callback()
def stream_chat(
self, messages: Sequence[ChatMessage], **kwargs: Any
) -> ChatResponseGen:
payload = {
"model": self.model,
"messages": [
{
"role": message.role.value,
"content": message.content,
**message.additional_kwargs,
}
for message in messages
],
"options": self._model_kwargs,
"stream": True,
**kwargs,
}
with httpx.Client(timeout=Timeout(self.request_timeout)) as client:
with client.stream(
method="POST",
url=f"{self.base_url}/api/chat",
json=payload,
) as response:
response.raise_for_status()
text = ""
for line in response.iter_lines():
if line:
chunk = json.loads(line)
if "done" in chunk and chunk["done"]:
break
message = chunk["message"]
delta = message.get("content")
text += delta
yield ChatResponse(
message=ChatMessage(
content=text,
role=MessageRole(message.get("role")),
additional_kwargs=get_additional_kwargs(
message, ("content", "role")
),
),
delta=delta,
raw=chunk,
additional_kwargs=get_additional_kwargs(
chunk, ("message",)
),
)
@llm_completion_callback()
def complete(
self, prompt: str, formatted: bool = False, **kwargs: Any
) -> CompletionResponse:
payload = {
self.prompt_key: prompt,
"model": self.model,
"options": self._model_kwargs,
"stream": False,
**kwargs,
}
with httpx.Client(timeout=Timeout(self.request_timeout)) as client:
response = client.post(
url=f"{self.base_url}/api/generate",
json=payload,
)
response.raise_for_status()
raw = response.json()
text = raw.get("response")
return CompletionResponse(
text=text,
raw=raw,
additional_kwargs=get_additional_kwargs(raw, ("response",)),
)
@llm_completion_callback()
def stream_complete(
self, prompt: str, formatted: bool = False, **kwargs: Any
) -> CompletionResponseGen:
payload = {
self.prompt_key: prompt,
"model": self.model,
"options": self._model_kwargs,
"stream": True,
**kwargs,
}
with httpx.Client(timeout=Timeout(self.request_timeout)) as client:
with client.stream(
method="POST",
url=f"{self.base_url}/api/generate",
json=payload,
) as response:
response.raise_for_status()
text = ""
for line in response.iter_lines():
if line:
chunk = json.loads(line)
delta = chunk.get("response")
text += delta
yield CompletionResponse(
delta=delta,
text=text,
raw=chunk,
additional_kwargs=get_additional_kwargs(
chunk, ("response",)
),
)
|