1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
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
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
use anyhow::{Error, Result, anyhow};
use async_openai::{
Client,
config::Config,
types::{
ChatCompletionRequestMessageContentPartTextArgs, ChatCompletionRequestUserMessageArgs,
CreateChatCompletionRequest, CreateChatCompletionRequestArgs, CreateChatCompletionResponse,
ResponseFormat,
},
};
use tokio::runtime::Runtime;
use crate::message_list::{Message, MessageList, Role};
/// Implement this for various LLM API standards
pub trait IsLLM {
/// Provides access to the client instance.
fn access_client(&self) -> &Client<impl Config>;
/// Provides access to the model identifier.
fn access_model(&self) -> &str;
}
/// Represent an object that has a system prompt
pub trait SystemPrompt {
/// Get the system prompt
fn get_system_prompt(&self) -> String;
}
/// Implement this for context management
pub trait Context {
/// Update the context
fn push(&mut self, role: Role, content: &str) -> Result<(), Error> {
match role {
Role::User => self
.get_context_mut()
.push(Message::new(Role::User, content.to_string())),
Role::Assistant => {
self.get_context_mut()
.push(Message::new(Role::Assistant, content.to_string()));
}
Role::System => {
self.get_context_mut()
.push(Message::new(Role::System, content.to_string()));
}
_ => return Err(anyhow!("Unsupported role")),
}
Ok(())
}
/// Get access right to read and write the context
fn get_context_mut(&mut self) -> &mut MessageList;
/// Get a copy of the context
fn get_context(&self) -> MessageList;
}
pub trait GenerateJSON
where
Self: IsLLM,
{
/// Generates JSON response from the LLM based on the provided prompt.
///
/// # Arguments
///
/// * `prompt` - A string slice that holds the prompt to be sent to the LLM.
/// * `target` - A string slice that holds the data to be sent to the LLM to generate a json.
///
/// # Returns
///
/// * `Result<String, Error>` - A result containing the JSON response as a string or an error.
fn generate_json(&self, task: &impl SystemPrompt, target: &str) -> Result<String, Error> {
let runtime = tokio::runtime::Runtime::new()?;
let result: String = runtime.block_on(async {
let request = CreateChatCompletionRequestArgs::default()
.model(&self.access_model().to_string())
.response_format(ResponseFormat::JsonObject)
.messages(vec![
ChatCompletionRequestUserMessageArgs::default()
.content(vec![
ChatCompletionRequestMessageContentPartTextArgs::default()
.text(
task.get_system_prompt()
+ "\nThis is the basis for generating a json:\n"
+ target,
)
.build()?
.into(),
])
.build()?
.into(),
])
.build()?;
let response: CreateChatCompletionResponse =
match self.access_client().chat().create(request.clone()).await {
std::result::Result::Ok(response) => response,
Err(e) => {
anyhow::bail!("Failed to execute function: {}", e);
}
};
if let Some(content) = response.choices[0].clone().message.content {
return Ok(content);
}
return Err(anyhow!("No response is retrieved from the LLM"));
})?;
Ok(result)
}
/// Generates JSON response from the LLM based on the provided context.
///
/// # Arguments
///
/// * `context` - A collection of `ChatCompletionRequestMessage` instances that provide the context to be sent to the LLM.
///
/// # Returns
///
/// * `Result<String, Error>` - A result containing the JSON response as a string or an error.
fn generate_json_with_context<T>(&self, task: &T) -> Result<String, Error>
where
T: SystemPrompt + Context,
{
let runtime: Runtime = tokio::runtime::Runtime::new()?;
let result: String = runtime.block_on(async {
let request: CreateChatCompletionRequest = CreateChatCompletionRequestArgs::default()
.model(&self.access_model().to_string())
.response_format(ResponseFormat::JsonObject)
.messages(task.get_context())
.build()?;
let response: CreateChatCompletionResponse =
match self.access_client().chat().create(request.clone()).await {
std::result::Result::Ok(response) => response,
Err(e) => {
anyhow::bail!("Failed to execute function: {}", e);
}
};
if let Some(content) = response.choices[0].clone().message.content {
return Ok(content);
}
return Err(anyhow!("No response is retrieved from the LLM"));
})?;
Ok(result)
}
}
pub trait AsyncGenerateJSON
where
Self: IsLLM,
{
/// Asynchronously generates JSON response from the LLM based on the provided prompt.
///
/// This is the asynchronous version of `generate_json` that can be used in async contexts.
///
/// # Arguments
///
/// * `task` - An implementation of `SystemPrompt` containing schema and instructions.
/// * `target` - A string slice that holds the data to be sent to the LLM to generate a json.
///
/// # Returns
///
/// * `Result<String, Error>` - A result containing the JSON response as a string or an error.
///
/// # Example
///
/// ```
/// use secretary::{openai::OpenAILLM, tasks::basic_task::BasicTask, traits::AsyncGenerateJSON};
/// use serde::{Deserialize, Serialize};
///
/// #[derive(Debug, Serialize, Deserialize)]
/// struct MyData {
/// field: String,
/// }
///
/// #[tokio::main]
/// async fn main() -> anyhow::Result<()> {
/// let llm = OpenAILLM::new("api_base", "api_key", "model")?;
/// let task = BasicTask::new(
/// MyData { field: "Description for field".to_string() },
/// vec!["Extract data from the text".to_string()],
/// );
///
/// let result = llm.async_generate_json(&task, "Some text with info").await?;
/// println!("{}", result);
/// Ok(())
/// }
/// ```
async fn async_generate_json(
&self,
task: &impl SystemPrompt,
target: &str,
) -> Result<String, Error> {
let request = CreateChatCompletionRequestArgs::default()
.model(&self.access_model().to_string())
.response_format(ResponseFormat::JsonObject)
.messages(vec![
ChatCompletionRequestUserMessageArgs::default()
.content(vec![
ChatCompletionRequestMessageContentPartTextArgs::default()
.text(
task.get_system_prompt()
+ "\nThis is the basis for generating a json:\n"
+ target,
)
.build()?
.into(),
])
.build()?
.into(),
])
.build()?;
let response: CreateChatCompletionResponse =
match self.access_client().chat().create(request.clone()).await {
std::result::Result::Ok(response) => response,
Err(e) => {
anyhow::bail!("Failed to execute function: {}", e);
}
};
if let Some(content) = response.choices[0].clone().message.content {
return Ok(content);
}
return Err(anyhow!("No response is retrieved from the LLM"));
}
/// Asynchronously generates JSON response from the LLM based on the provided context.
///
/// This is the asynchronous version of `generate_json_with_context` that enables
/// context-aware conversations in async contexts.
///
/// # Arguments
///
/// * `task` - An implementation of both `SystemPrompt` and `Context` traits that provides
/// both the schema definition and conversation history.
///
/// # Returns
///
/// * `Result<String, Error>` - A result containing the JSON response as a string or an error.
///
/// # Example
///
/// ```
/// use secretary::{
/// message_list::Role,
/// openai::OpenAILLM,
/// tasks::basic_task::BasicTask,
/// traits::{AsyncGenerateJSON, Context},
/// };
/// use serde::{Deserialize, Serialize};
///
/// #[derive(Debug, Serialize, Deserialize)]
/// struct MyData {
/// field: String,
/// }
///
/// #[tokio::main]
/// async fn main() -> anyhow::Result<()> {
/// let llm = OpenAILLM::new("api_base", "api_key", "model")?;
/// let mut task = BasicTask::new(
/// MyData { field: "Description for field".to_string() },
/// vec!["Extract data from the text".to_string()],
/// );
///
/// // Add messages to the conversation context
/// task.push(Role::User, "Here's my first message")?;
///
/// // Generate response with context
/// let result = llm.async_generate_json_with_context(&task).await?;
/// println!("{}", result);
/// Ok(())
/// }
/// ```
async fn async_generate_json_with_context<T>(&self, task: &T) -> Result<String, Error>
where
T: SystemPrompt + Context,
{
let request: CreateChatCompletionRequest = CreateChatCompletionRequestArgs::default()
.model(&self.access_model().to_string())
.response_format(ResponseFormat::JsonObject)
.messages(task.get_context())
.build()?;
let response: CreateChatCompletionResponse =
match self.access_client().chat().create(request.clone()).await {
std::result::Result::Ok(response) => response,
Err(e) => {
anyhow::bail!("Failed to execute function: {}", e);
}
};
if let Some(content) = response.choices[0].clone().message.content {
return Ok(content);
}
return Err(anyhow!("No response is retrieved from the LLM"));
}
}