plainllm 1.2.0

A plain & simple LLM client
Documentation 
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

use super::tool_registry::ToolRegistry;
use super::types::FunctionCall;
use serde_json::Value;
use std::sync::Arc;

//////////////////////////////
// 3. The "PlainLLM" and LLMOptions
//////////////////////////////

/// Event handlers to which the user can optionally subscribe
#[derive(Default)]
pub struct LLMEventHandlers {
    // Called whenever a new token chunk is received (in streaming mode).
    pub on_token: Option<Arc<dyn Fn(&str) + Send + Sync>>,

    // If your model returns <think>…</think> reasoning, this gets called whenever
    // we've found the opening <think>
    pub on_start_thinking: Option<Arc<dyn Fn() + Send + Sync>>,

    // If your model returns <think>…</think> reasoning, this gets called whenever
    // we've found the closing </think>
    pub on_stop_thinking: Option<Arc<dyn Fn() + Send + Sync>>,

    // If your model returns <think>…</think> reasoning, you can intercept it here.
    pub on_thinking: Option<Arc<dyn Fn(&str) + Send + Sync>>,

    // Called when the model requests a tool call (if you have a ToolRegistry).
    pub on_tool_call: Option<Arc<dyn Fn(&FunctionCall) + Send + Sync>>,

    // Called after a tool call is executed, returning a success or error.
    pub on_tool_result: Option<Arc<dyn Fn(&FunctionCall, &Result<Value, String>) + Send + Sync>>,
}

/// Builder for LLM call options. Everything is optional.
#[derive(Default)]
pub struct LLMOptions<'a> {
    pub streaming: bool,
    pub tools: Option<&'a ToolRegistry>,
    pub temperature: Option<f32>,
    pub top_p: Option<f32>,
    pub min_p: Option<f32>,
    pub top_k: Option<u32>,
    pub max_tokens: Option<u32>,
    pub stop: Option<Vec<String>>,
    pub presence_penalty: Option<f32>,
    pub frequency_penalty: Option<f32>,
    pub repeat_penalty: Option<f32>,
    pub context_overflow_policy: Option<String>,
    pub event_handlers: LLMEventHandlers,
}

impl<'a> LLMOptions<'a> {
    pub fn new() -> Self {
        Self::default()
    }

    pub fn streaming(mut self, streaming: bool) -> Self {
        self.streaming = streaming;
        self
    }

    pub fn with_tools(mut self, tools: &'a ToolRegistry) -> Self {
        self.tools = Some(tools);
        self
    }

    pub fn temperature(mut self, temperature: f32) -> Self {
        self.temperature = Some(temperature);
        self
    }

    pub fn top_p(mut self, top_p: f32) -> Self {
        self.top_p = Some(top_p);
        self
    }

    pub fn min_p(mut self, min_p: f32) -> Self {
        self.top_p = Some(min_p);
        self
    }

    pub fn top_k(mut self, top_k: u32) -> Self {
        self.top_k = Some(top_k);
        self
    }

    pub fn max_tokens(mut self, max_tokens: u32) -> Self {
        self.max_tokens = Some(max_tokens);
        self
    }

    pub fn stop(mut self, stop: Vec<String>) -> Self {
        self.stop = Some(stop);
        self
    }

    pub fn presence_penalty(mut self, penalty: f32) -> Self {
        self.presence_penalty = Some(penalty);
        self
    }

    pub fn repeat_penalty(mut self, repeat_penalty: f32) -> Self {
        self.repeat_penalty = Some(repeat_penalty);
        self
    }

    pub fn context_overflow_policy(mut self, context_overflow_policy: &str) -> Self {
        self.context_overflow_policy = Some(context_overflow_policy.to_string());
        self
    }

    pub fn frequency_penalty(mut self, penalty: f32) -> Self {
        self.frequency_penalty = Some(penalty);
        self
    }

    pub fn on_token<F>(mut self, callback: F) -> Self
    where
        F: Fn(&str) + Send + Sync + 'static,
    {
        self.event_handlers.on_token = Some(Arc::new(callback));
        self
    }

    pub fn on_start_thinking<F>(mut self, callback: F) -> Self
    where
        F: Fn() + Send + Sync + 'static,
    {
        self.event_handlers.on_start_thinking = Some(Arc::new(callback));
        self
    }

    pub fn on_stop_thinking<F>(mut self, callback: F) -> Self
    where
        F: Fn() + Send + Sync + 'static,
    {
        self.event_handlers.on_stop_thinking = Some(Arc::new(callback));
        self
    }

    pub fn on_thinking<F>(mut self, callback: F) -> Self
    where
        F: Fn(&str) + Send + Sync + 'static,
    {
        self.event_handlers.on_thinking = Some(Arc::new(callback));
        self
    }

    pub fn on_tool_call<F>(mut self, callback: F) -> Self
    where
        F: Fn(&FunctionCall) + Send + Sync + 'static,
    {
        self.event_handlers.on_tool_call = Some(Arc::new(callback));
        self
    }

    pub fn on_tool_result<F>(mut self, callback: F) -> Self
    where
        F: Fn(&FunctionCall, &Result<Value, String>) + Send + Sync + 'static,
    {
        self.event_handlers.on_tool_result = Some(Arc::new(callback));
        self
    }
}