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
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
//! High-level client API
mod builder;
mod config;
pub use builder::ClientBuilder;
pub use config::{ClientConfig, InferenceConfig};
use crate::error::Result;
use crate::registry::ProviderRegistry;
use crate::types::{GenerateRequest, GenerateResponse, GenerateStream};
#[cfg(feature = "tracing")]
use tracing::Instrument;
#[cfg(feature = "tracing")]
use crate::tracing as gen_ai_tracing;
/// High-level inference client for AI generation
#[derive(Clone)]
pub struct Inference {
registry: ProviderRegistry,
#[allow(dead_code)]
config: ClientConfig,
}
impl Inference {
/// Create a new inference client with default configuration
///
/// Providers are auto-registered from environment variables:
/// - `OPENAI_API_KEY` for OpenAI
/// - `ANTHROPIC_API_KEY` for Anthropic
/// - `GEMINI_API_KEY` for Google Gemini
pub fn new() -> Self {
Self::builder()
.build()
.expect("Failed to build Inference client")
}
/// Create an inference client with custom provider configuration
///
/// # Example
///
/// ```rust,no_run
/// use stakai::{Inference, InferenceConfig};
///
/// let client = Inference::with_config(
/// InferenceConfig::new()
/// .openai("sk-...", None)
/// .anthropic("sk-ant-...", None)
/// .gemini("your-key", None)
/// );
/// ```
pub fn with_config(config: InferenceConfig) -> Result<Self> {
Self::builder().with_inference_config(config).build()
}
/// Create an inference client builder
pub fn builder() -> ClientBuilder {
ClientBuilder::default()
}
/// Generate a response
///
/// When the `tracing` feature is enabled, this operation is automatically
/// traced with [GenAI semantic conventions](https://opentelemetry.io/docs/specs/semconv/gen-ai/gen-ai-spans/).
///
/// # Arguments
///
/// * `request` - Generation request with model identifier (e.g., "gpt-4" or "openai/gpt-4")
///
/// # Example
///
/// ```rust,no_run
/// # use stakai::{Inference, GenerateRequest, Message, Model, Role};
/// # async fn example() -> Result<(), Box<dyn std::error::Error>> {
/// let client = Inference::new();
/// let request = GenerateRequest::new(
/// Model::custom("gpt-4", "openai"),
/// vec![Message::new(Role::User, "Hello!")]
/// );
/// let response = client.generate(&request).await?;
/// # Ok(())
/// # }
/// ```
///
/// # Tracing
///
/// When the `tracing` feature is enabled, spans are automatically emitted with:
/// - `gen_ai.operation.name`: "chat"
/// - `gen_ai.provider.name`: provider name (e.g., "openai", "anthropic")
/// - `gen_ai.request.model`: model identifier
/// - `gen_ai.input.messages`: JSON array of input messages (opt-in)
/// - `gen_ai.output.messages`: JSON array of output messages (opt-in)
/// - `gen_ai.usage.input_tokens`: prompt tokens used
/// - `gen_ai.usage.output_tokens`: completion tokens used
/// - `gen_ai.response.finish_reasons`: array of finish reasons
pub async fn generate(&self, request: &GenerateRequest) -> Result<GenerateResponse> {
#[cfg(feature = "tracing")]
{
let span = tracing::info_span!(
"chat",
"gen_ai.operation.name" = "chat",
"gen_ai.provider.name" = %request.model.provider,
"gen_ai.request.model" = %request.model.id,
"gen_ai.request.temperature" = tracing::field::Empty,
"gen_ai.request.max_tokens" = tracing::field::Empty,
"gen_ai.request.top_p" = tracing::field::Empty,
"gen_ai.request.frequency_penalty" = tracing::field::Empty,
"gen_ai.request.presence_penalty" = tracing::field::Empty,
"gen_ai.input.messages" = tracing::field::Empty,
"gen_ai.output.messages" = tracing::field::Empty,
"gen_ai.tool.definitions" = tracing::field::Empty,
"gen_ai.usage.input_tokens" = tracing::field::Empty,
"gen_ai.usage.output_tokens" = tracing::field::Empty,
// Non-standard: Cache token metrics (not part of OTel GenAI semantic conventions)
"gen_ai.usage.cache_read_input_tokens" = tracing::field::Empty,
"gen_ai.usage.cache_write_input_tokens" = tracing::field::Empty,
"gen_ai.response.finish_reasons" = tracing::field::Empty,
);
// Record optional request parameters
if let Some(t) = request.options.temperature {
span.record("gen_ai.request.temperature", t);
}
if let Some(m) = request.options.max_tokens {
span.record("gen_ai.request.max_tokens", m as i64);
}
if let Some(p) = request.options.top_p {
span.record("gen_ai.request.top_p", p);
}
if let Some(fp) = request.options.frequency_penalty {
span.record("gen_ai.request.frequency_penalty", fp);
}
if let Some(pp) = request.options.presence_penalty {
span.record("gen_ai.request.presence_penalty", pp);
}
// Record custom telemetry metadata and tool definitions
{
let _guard = span.enter();
if let Some(ref metadata) = request.telemetry_metadata {
gen_ai_tracing::record_telemetry_metadata(metadata);
}
if let Some(ref tools) = request.options.tools {
gen_ai_tracing::record_tool_definitions(tools);
}
}
// Clone data needed inside the async block
let messages = request.messages.clone();
return async {
// Record input messages as span attribute
gen_ai_tracing::record_input_messages(&messages);
let response = self.generate_internal(request).await?;
// Record response attributes
tracing::Span::current().record(
"gen_ai.usage.input_tokens",
response.usage.prompt_tokens as i64,
);
tracing::Span::current().record(
"gen_ai.usage.output_tokens",
response.usage.completion_tokens as i64,
);
// Non-standard: Cache token metrics (not part of OTel GenAI semantic conventions)
if let Some(cache_read) = response.usage.cache_read_tokens() {
tracing::Span::current()
.record("gen_ai.usage.cache_read_input_tokens", cache_read as i64);
}
if let Some(cache_write) = response.usage.cache_write_tokens() {
tracing::Span::current()
.record("gen_ai.usage.cache_write_input_tokens", cache_write as i64);
}
// finish_reasons is an array per OTel spec
let finish_reason = format!("{:?}", response.finish_reason.unified);
let finish_reasons_json =
serde_json::to_string(&vec![&finish_reason]).unwrap_or_default();
tracing::Span::current().record(
"gen_ai.response.finish_reasons",
finish_reasons_json.as_str(),
);
// Record response content as span attribute
gen_ai_tracing::record_response_content(&response, &finish_reason);
Ok(response)
}
.instrument(span)
.await;
}
#[cfg(not(feature = "tracing"))]
self.generate_internal(request).await
}
/// Internal generate implementation
async fn generate_internal(&self, request: &GenerateRequest) -> Result<GenerateResponse> {
let provider = self.registry.get_provider(&request.model.provider)?;
provider.generate(request.clone()).await
}
/// Generate a streaming response
///
/// When the `tracing` feature is enabled, the stream is automatically
/// traced with [GenAI semantic conventions](https://opentelemetry.io/docs/specs/semconv/gen-ai/gen-ai-spans/).
/// Token usage is recorded when the stream completes (on the `Finish` event).
///
/// # Arguments
///
/// * `request` - Generation request with model identifier
///
/// # Example
///
/// ```rust,no_run
/// # use stakai::{Inference, GenerateRequest, Message, Model, Role, StreamEvent};
/// # use futures::StreamExt;
/// # async fn example() -> Result<(), Box<dyn std::error::Error>> {
/// let client = Inference::new();
/// let request = GenerateRequest::new(
/// Model::custom("gpt-4", "openai"),
/// vec![Message::new(Role::User, "Count to 5")]
/// );
/// let mut stream = client.stream(&request).await?;
///
/// while let Some(event) = stream.next().await {
/// match event? {
/// StreamEvent::TextDelta { delta, .. } => print!("{}", delta),
/// _ => {}
/// }
/// }
/// # Ok(())
/// # }
/// ```
///
/// # Tracing
///
/// When the `tracing` feature is enabled, spans are automatically emitted with:
/// - `gen_ai.operation.name`: "chat" (streaming is still a chat operation)
/// - `gen_ai.provider.name`: provider name (e.g., "openai", "anthropic")
/// - `gen_ai.request.model`: model identifier
/// - `gen_ai.input.messages`: JSON array of input messages (opt-in)
/// - `gen_ai.output.messages`: JSON array of output messages (opt-in, recorded on finish)
/// - `gen_ai.usage.input_tokens`: prompt tokens (recorded on stream finish)
/// - `gen_ai.usage.output_tokens`: completion tokens (recorded on stream finish)
/// - `gen_ai.response.finish_reasons`: array of finish reasons (recorded on stream finish)
pub async fn stream(&self, request: &GenerateRequest) -> Result<GenerateStream> {
#[cfg(feature = "tracing")]
{
let span = tracing::info_span!(
"chat",
"gen_ai.operation.name" = "chat",
"gen_ai.provider.name" = %request.model.provider,
"gen_ai.request.model" = %request.model.id,
"gen_ai.request.temperature" = tracing::field::Empty,
"gen_ai.request.max_tokens" = tracing::field::Empty,
"gen_ai.request.top_p" = tracing::field::Empty,
"gen_ai.request.frequency_penalty" = tracing::field::Empty,
"gen_ai.request.presence_penalty" = tracing::field::Empty,
"gen_ai.input.messages" = tracing::field::Empty,
"gen_ai.output.messages" = tracing::field::Empty,
"gen_ai.tool.definitions" = tracing::field::Empty,
"gen_ai.usage.input_tokens" = tracing::field::Empty,
"gen_ai.usage.output_tokens" = tracing::field::Empty,
// Non-standard: Cache token metrics (not part of OTel GenAI semantic conventions)
"gen_ai.usage.cache_read_input_tokens" = tracing::field::Empty,
"gen_ai.usage.cache_write_input_tokens" = tracing::field::Empty,
"gen_ai.response.finish_reasons" = tracing::field::Empty,
);
// Record optional request parameters
if let Some(t) = request.options.temperature {
span.record("gen_ai.request.temperature", t);
}
if let Some(m) = request.options.max_tokens {
span.record("gen_ai.request.max_tokens", m as i64);
}
if let Some(p) = request.options.top_p {
span.record("gen_ai.request.top_p", p);
}
if let Some(fp) = request.options.frequency_penalty {
span.record("gen_ai.request.frequency_penalty", fp);
}
if let Some(pp) = request.options.presence_penalty {
span.record("gen_ai.request.presence_penalty", pp);
}
// Record input messages, custom telemetry metadata, and tool definitions
{
let _guard = span.enter();
gen_ai_tracing::record_input_messages(&request.messages);
if let Some(ref metadata) = request.telemetry_metadata {
gen_ai_tracing::record_telemetry_metadata(metadata);
}
if let Some(ref tools) = request.options.tools {
gen_ai_tracing::record_tool_definitions(tools);
}
}
// Create the inner stream, then wrap it with our span
let inner_stream = self.stream_internal(request).await?;
// Return a stream that will record usage and completion when it finishes
Ok(GenerateStream::with_span(Box::pin(inner_stream), span))
}
#[cfg(not(feature = "tracing"))]
self.stream_internal(request).await
}
/// Internal stream implementation
async fn stream_internal(&self, request: &GenerateRequest) -> Result<GenerateStream> {
let provider = self.registry.get_provider(&request.model.provider)?;
provider.stream(request.clone()).await
}
/// Get the provider registry
pub fn registry(&self) -> &ProviderRegistry {
&self.registry
}
}
impl Default for Inference {
fn default() -> Self {
Self::new()
}
}