Add: Berget AI provider support

Author: kristoferlund

CHANGELOG.md

@@ -7,6 +7,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Added
+
+- **Berget provider** - Added support for Berget transcription service with Swedish-optimized KB Whisper Large model
+
 ## 0.0.7 - 2026-02-05
 
 ### Added

README.md

@@ -50,6 +50,9 @@ ostt supports multiple AI transcription providers. Bring your own API key and ch
 - **groq-whisper-large-v3** - High accuracy processing
 - **groq-whisper-large-v3-turbo** - Fastest transcription speed
 
+### Berget
+- **berget-whisper-kb-large** - KB Whisper Large (Swedish optimized)
+
 Configure your preferred provider and model using `ostt auth`.
 
 ## Installation

src/transcription/api/berget.rs

@@ -0,0 +1,126 @@
+//! Berget API implementation.
+//!
+//! Handles transcription requests to Berget's OpenAI-compatible Whisper API using multipart form data.
+
+use std::path::Path;
+use serde::Deserialize;
+
+use super::TranscriptionConfig;
+
+/// Berget API response wrapper
+#[derive(Debug, Deserialize)]
+struct BergetResponse {
+    text: String,
+}
+
+/// Berget API error response
+#[derive(Debug, Deserialize)]
+#[allow(dead_code)]
+struct BergetErrorResponse {
+    code: String,
+    error: String,
+    #[serde(default)]
+    details: Option<String>,
+}
+
+/// Transcribes an audio file using Berget's Whisper API.
+///
+/// Uses multipart form data with bearer token authentication.
+/// Berget provides an OpenAI-compatible API endpoint.
+///
+/// Keywords are passed as the `prompt` parameter to guide transcription context.
+pub async fn transcribe(
+    config: &TranscriptionConfig,
+    audio_path: &Path,
+) -> anyhow::Result<String> {
+    let audio_data = std::fs::read(audio_path).map_err(|e| {
+        anyhow::anyhow!("Failed to read audio file: {e}")
+    })?;
+
+    let client = reqwest::Client::new();
+
+    let file_name = audio_path
+        .file_name()
+        .unwrap_or_default()
+        .to_string_lossy()
+        .to_string();
+
+    let file_part = reqwest::multipart::Part::bytes(audio_data)
+        .file_name(file_name.clone())
+        .mime_str("audio/mpeg")
+        .map_err(|e| anyhow::anyhow!("Failed to create file part for upload: {e}"))?;
+
+    let mut form = reqwest::multipart::Form::new()
+        .part("file", file_part)
+        .text("model", config.model.api_model_name().to_string());
+
+    // Debug log: Log the API call details (without the audio data)
+    let mut debug_params = vec![
+        format!("model={}", config.model.api_model_name()),
+    ];
+
+    // Add keywords as prompt for better transcription context
+    if !config.keywords.is_empty() {
+        let prompt = config.keywords.join(", ");
+        form = form.text("prompt", prompt.clone());
+        debug_params.push(format!("prompt={prompt}"));
+        tracing::debug!("Keywords used as prompt for Berget model: {:?}", config.keywords);
+    }
+
+    let endpoint = config.model.endpoint();
+
+    tracing::debug!(
+        "Berget API Call:\n  URL: {}\n  Method: POST\n  Headers:\n    Authorization: Bearer <redacted>\n    Content-Type: multipart/form-data\n  Body parameters: {}",
+        endpoint,
+        debug_params.join("\n    ")
+    );
+
+    let response = match client
+        .post(endpoint)
+        .bearer_auth(&config.api_key)
+        .multipart(form)
+        .send()
+        .await
+    {
+        Ok(resp) => resp,
+        Err(e) => {
+            let error_msg = if e.is_connect() {
+                "Failed to connect to Berget API server. Check your internet connection.".to_string()
+            } else if e.is_timeout() {
+                "Request to Berget timed out. The API server is not responding.".to_string()
+            } else if e.to_string().contains("builder") {
+                format!("Failed to build Berget API request: {e}. This may be a configuration error.")
+            } else {
+                format!("Berget network error: {e}")
+            };
+            return Err(anyhow::anyhow!(error_msg));
+        }
+    };
+
+    if !response.status().is_success() {
+        let status = response.status();
+        
+        // Parse the JSON error response - all errors follow the same structure
+        let error_message = response
+            .json::<BergetErrorResponse>()
+            .await
+            .map(|e| e.error)
+            .unwrap_or_else(|_| format!("HTTP {status}"));
+
+        return Err(anyhow::anyhow!("Berget API error: {error_message}"));
+    }
+
+    let berget_response: BergetResponse = response
+        .json()
+        .await
+        .map_err(|e| anyhow::anyhow!("Failed to parse Berget response: {e}"))?;
+
+    // Debug log: Log the full response for debugging
+    tracing::debug!(
+        "Berget API Response:\n  Status: Success\n  Transcription length: {} characters\n  Full response: {:#?}",
+        berget_response.text.len(),
+        berget_response
+    );
+
+    Ok(berget_response.text.trim().to_string())
+}

src/transcription/api/mod.rs

@@ -8,6 +8,7 @@ mod openai;
 mod deepgram;
 mod deepinfra;
 mod groq;
+mod berget;
 
 use serde::Deserialize;
 use std::path::Path;
@@ -86,6 +87,9 @@ pub async fn transcribe(
         TranscriptionProvider::Groq => {
             groq::transcribe(config, audio_path).await
         }
+        TranscriptionProvider::Berget => {
+            berget::transcribe(config, audio_path).await
+        }
     }?;
 
     Ok(result)

src/transcription/model.rs

@@ -28,6 +28,8 @@ pub enum TranscriptionModel {
     GroqWhisperLargeV3,
     /// Groq Whisper Large V3 Turbo model (faster)
     GroqWhisperLargeV3Turbo,
+    /// Berget Whisper KB Large model (Swedish optimized)
+    BergetWhisperKBLarge,
 }
 
 impl TranscriptionModel {
@@ -44,6 +46,7 @@ impl TranscriptionModel {
             | TranscriptionModel::DeepInfraWhisperBase => TranscriptionProvider::DeepInfra,
             TranscriptionModel::GroqWhisperLargeV3
             | TranscriptionModel::GroqWhisperLargeV3Turbo => TranscriptionProvider::Groq,
+            TranscriptionModel::BergetWhisperKBLarge => TranscriptionProvider::Berget,
         }
     }
 
@@ -59,6 +62,7 @@ impl TranscriptionModel {
             TranscriptionModel::DeepInfraWhisperBase => "deepinfra-whisper-base",
             TranscriptionModel::GroqWhisperLargeV3 => "groq-whisper-large-v3",
             TranscriptionModel::GroqWhisperLargeV3Turbo => "groq-whisper-large-v3-turbo",
+            TranscriptionModel::BergetWhisperKBLarge => "berget-whisper-kb-large",
         }
     }
 
@@ -74,6 +78,7 @@ impl TranscriptionModel {
             TranscriptionModel::DeepInfraWhisperBase => "Whisper Base (fast, lightweight)",
             TranscriptionModel::GroqWhisperLargeV3 => "Whisper Large V3 (high accuracy)",
             TranscriptionModel::GroqWhisperLargeV3Turbo => "Whisper Large V3 Turbo (fastest)",
+            TranscriptionModel::BergetWhisperKBLarge => "KB Whisper Large (Swedish optimized)",
         }
     }
 
@@ -89,7 +94,12 @@ impl TranscriptionModel {
             TranscriptionModel::DeepInfraWhisperLargeV3
             | TranscriptionModel::DeepInfraWhisperBase => "https://api.deepinfra.com/v1/inference",
             TranscriptionModel::GroqWhisperLargeV3
-            | TranscriptionModel::GroqWhisperLargeV3Turbo => "https://api.groq.com/openai/v1/audio/transcriptions",
+            | TranscriptionModel::GroqWhisperLargeV3Turbo => {
+                "https://api.groq.com/openai/v1/audio/transcriptions"
+            }
+            TranscriptionModel::BergetWhisperKBLarge => {
+                "https://api.berget.ai/v1/audio/transcriptions"
+            }
         }
     }
 
@@ -105,6 +115,7 @@ impl TranscriptionModel {
             TranscriptionModel::DeepInfraWhisperBase => "openai/whisper-base",
             TranscriptionModel::GroqWhisperLargeV3 => "whisper-large-v3",
             TranscriptionModel::GroqWhisperLargeV3Turbo => "whisper-large-v3-turbo",
+            TranscriptionModel::BergetWhisperKBLarge => "KBLab/kb-whisper-large",
         }
     }
 
@@ -120,6 +131,7 @@ impl TranscriptionModel {
             "deepinfra-whisper-base" => Some(TranscriptionModel::DeepInfraWhisperBase),
             "groq-whisper-large-v3" => Some(TranscriptionModel::GroqWhisperLargeV3),
             "groq-whisper-large-v3-turbo" => Some(TranscriptionModel::GroqWhisperLargeV3Turbo),
+            "berget-whisper-kb-large" => Some(TranscriptionModel::BergetWhisperKBLarge),
             _ => None,
         }
     }
@@ -136,6 +148,7 @@ impl TranscriptionModel {
             TranscriptionModel::DeepInfraWhisperBase,
             TranscriptionModel::GroqWhisperLargeV3,
             TranscriptionModel::GroqWhisperLargeV3Turbo,
+            TranscriptionModel::BergetWhisperKBLarge,
         ]
     }
 

src/transcription/provider.rs

@@ -12,6 +12,7 @@ pub enum TranscriptionProvider {
     Deepgram,
     DeepInfra,
     Groq,
+    Berget,
 }
 
 impl TranscriptionProvider {
@@ -21,6 +22,7 @@ impl TranscriptionProvider {
             TranscriptionProvider::Deepgram => "deepgram",
             TranscriptionProvider::DeepInfra => "deepinfra",
             TranscriptionProvider::Groq => "groq",
+            TranscriptionProvider::Berget => "berget",
         }
     }
 
@@ -30,6 +32,7 @@ impl TranscriptionProvider {
             TranscriptionProvider::Deepgram => "Deepgram",
             TranscriptionProvider::DeepInfra => "DeepInfra",
             TranscriptionProvider::Groq => "Groq",
+            TranscriptionProvider::Berget => "Berget",
         }
     }
 
@@ -39,6 +42,7 @@ impl TranscriptionProvider {
             "deepgram" => Some(TranscriptionProvider::Deepgram),
             "deepinfra" => Some(TranscriptionProvider::DeepInfra),
             "groq" => Some(TranscriptionProvider::Groq),
+            "berget" => Some(TranscriptionProvider::Berget),
             _ => None,
         }
     }
@@ -49,6 +53,7 @@ impl TranscriptionProvider {
             TranscriptionProvider::Deepgram,
             TranscriptionProvider::DeepInfra,
             TranscriptionProvider::Groq,
+            TranscriptionProvider::Berget,
         ]
     }
 }