diff --git a/src/llm/mod.rs b/src/llm/mod.rs
index aaa412abd..8e8190f53 100644
--- a/src/llm/mod.rs
+++ b/src/llm/mod.rs
@@ -1,5 +1,7 @@
 use std::borrow::Cow;
 
+use anyhow::Result;
+use async_trait::async_trait;
 use schemars::schema::SchemaObject;
 use serde::{Deserialize, Serialize};
 
@@ -32,5 +34,21 @@ pub struct LlmGenerateResponse {
     pub text: String,
 }
 
-mod client;
-pub use client::LlmClient;
+#[async_trait]
+pub trait LlmGenerationClient: Send + Sync {
+    async fn generate<'req>(
+        &self,
+        request: LlmGenerateRequest<'req>,
+    ) -> Result<LlmGenerateResponse>;
+}
+
+mod ollama;
+
+pub async fn new_llm_generation_client(spec: LlmSpec) -> Result<Box<dyn LlmGenerationClient>> {
+    let client = match spec.api_type {
+        LlmApiType::Ollama => {
+            Box::new(ollama::Client::new(spec).await?) as Box<dyn LlmGenerationClient>
+        }
+    };
+    Ok(client)
+}
diff --git a/src/llm/client.rs b/src/llm/ollama.rs
similarity index 89%
rename from src/llm/client.rs
rename to src/llm/ollama.rs
index 28ada294f..31bb22447 100644
--- a/src/llm/client.rs
+++ b/src/llm/ollama.rs
@@ -1,8 +1,10 @@
+use super::LlmGenerationClient;
 use anyhow::Result;
+use async_trait::async_trait;
 use schemars::schema::SchemaObject;
 use serde::{Deserialize, Serialize};
 
-pub struct LlmClient {
+pub struct Client {
     generate_url: String,
     model: String,
     reqwest_client: reqwest::Client,
@@ -30,7 +32,7 @@ struct OllamaResponse {
 
 const OLLAMA_DEFAULT_ADDRESS: &str = "http://localhost:11434";
 
-impl LlmClient {
+impl Client {
     pub async fn new(spec: super::LlmSpec) -> Result<Self> {
         let address = match &spec.address {
             Some(addr) => addr.trim_end_matches('/'),
@@ -42,10 +44,13 @@ impl LlmClient {
             reqwest_client: reqwest::Client::new(),
         })
     }
+}
 
-    pub async fn generate<'a>(
+#[async_trait]
+impl LlmGenerationClient for Client {
+    async fn generate<'req>(
         &self,
-        request: super::LlmGenerateRequest<'a>,
+        request: super::LlmGenerateRequest<'req>,
     ) -> Result<super::LlmGenerateResponse> {
         let req = OllamaRequest {
             model: &self.model,
diff --git a/src/ops/functions/extract_by_llm.rs b/src/ops/functions/extract_by_llm.rs
index 3b3989dbc..64363ebc1 100644
--- a/src/ops/functions/extract_by_llm.rs
+++ b/src/ops/functions/extract_by_llm.rs
@@ -5,7 +5,9 @@ use schemars::schema::SchemaObject;
 use serde::Serialize;
 
 use crate::base::json_schema::ToJsonSchema;
-use crate::llm::{LlmClient, LlmGenerateRequest, LlmSpec, OutputFormat};
+use crate::llm::{
+    new_llm_generation_client, LlmGenerateRequest, LlmGenerationClient, LlmSpec, OutputFormat,
+};
 use crate::ops::sdk::*;
 
 #[derive(Debug, Clone, Serialize, Deserialize)]
@@ -16,7 +18,7 @@ pub struct Spec {
 }
 
 struct Executor {
-    client: LlmClient,
+    client: Box<dyn LlmGenerationClient>,
     output_json_schema: SchemaObject,
     output_type: EnrichedValueType,
     system_prompt: String,
@@ -41,7 +43,7 @@ Output only the JSON without any additional messages or explanations."
 impl Executor {
     async fn new(spec: Spec) -> Result<Self> {
         Ok(Self {
-            client: LlmClient::new(spec.llm_spec).await?,
+            client: new_llm_generation_client(spec.llm_spec).await?,
             output_json_schema: spec.output_type.to_json_schema(),
             output_type: spec.output_type,
             system_prompt: get_system_prompt(&spec.instruction),