Add support for configuring max_completion_tokens for OpenAI

Related to db94227

Author: spantaleev

docs/sample-provider-configs/openai-o1.yml

@@ -6,9 +6,9 @@ text_generation:
   prompt: null
   temperature: 1.0
   # o1 models do not support max_response_tokens.
-  # They use `max_completion_tokens` as an alternative,
-  # but we don't support it yet (see https://github.com/64bit/async-openai/issues/272).
+  # They use `max_completion_tokens` as an alternative
   max_response_tokens: null
+  max_completion_tokens: 16384
   max_context_tokens: 128000
 speech_to_text:
   model_id: whisper-1

etc/app/config.yml.dist

@@ -80,6 +80,8 @@ agents:
     #       prompt: "You are a brief, but helpful bot called {{ baibot_name }} powered by the {{ baibot_model_id }} model. The date/time of this conversation's start is: {{ baibot_conversation_start_time_utc }}."
     #       temperature: 1.0
     #       max_response_tokens: 16384
+    #       # Reasoning models need to use `max_completion_tokens` instead of `max_response_tokens`.
+    #       max_completion_tokens: ~
     #       max_context_tokens: 128000
     #     speech_to_text:
     #       model_id: whisper-1

src/agent/provider/openai/config.rs

@@ -58,6 +58,9 @@ pub struct TextGenerationConfig {
     #[serde(default)]
     pub max_response_tokens: Option<u32>,
 
+    #[serde(default)]
+    pub max_completion_tokens: Option<u32>,
+
     #[serde(default)]
     pub max_context_tokens: u32,
 }
@@ -69,6 +72,7 @@ impl Default for TextGenerationConfig {
             prompt: Some(default_prompt().to_owned()),
             temperature: super::super::default_temperature(),
             max_response_tokens: Some(16_384),
+            max_completion_tokens: None,
             max_context_tokens: 128_000,
         }
     }

src/agent/provider/openai/controller.rs

@@ -144,6 +144,10 @@ impl ControllerTrait for Controller {
             request_builder.max_tokens(max_response_tokens);
         }
 
+        if let Some(max_completion_tokens) = text_generation_config.max_completion_tokens {
+            request_builder.max_completion_tokens(max_completion_tokens);
+        }
+
         let request = request_builder.build()?;
 
         if let Ok(request_as_json) = serde_json::to_string(&request) {

src/agent/provider/openai_compat/config.rs

@@ -93,6 +93,7 @@ impl TryInto<OpenAITextGenerationConfig> for TextGenerationConfig {
             prompt: self.prompt,
             temperature: self.temperature,
             max_response_tokens: self.max_response_tokens,
+            max_completion_tokens: None,
             max_context_tokens: self.max_context_tokens,
         })
     }