diff --git a/crates/lsp-ai/src/config.rs b/crates/lsp-ai/src/config.rs
index 89abfa4..30618d4 100644
--- a/crates/lsp-ai/src/config.rs
+++ b/crates/lsp-ai/src/config.rs
@@ -14,12 +14,12 @@ const fn true_default() -> bool {
 }
 
 #[derive(Clone, Debug, Deserialize, Serialize)]
-pub struct PostProcess {
-    pub extractor: Option<String>,
+pub(crate) struct PostProcess {
+    pub(crate) extractor: Option<String>,
     #[serde(default = "true_default")]
-    pub remove_duplicate_start: bool,
+    pub(crate) remove_duplicate_start: bool,
     #[serde(default = "true_default")]
-    pub remove_duplicate_end: bool,
+    pub(crate) remove_duplicate_end: bool,
 }
 
 impl Default for PostProcess {
@@ -34,7 +34,7 @@ impl Default for PostProcess {
 
 #[derive(Debug, Clone, Deserialize)]
 #[serde(tag = "type")]
-pub enum ValidSplitter {
+pub(crate) enum ValidSplitter {
     #[serde(rename = "tree_sitter")]
     TreeSitter(TreeSitter),
     #[serde(rename = "text_splitter")]
@@ -56,11 +56,11 @@ const fn chunk_overlap_default() -> usize {
 }
 
 #[derive(Debug, Clone, Deserialize)]
-pub struct TreeSitter {
+pub(crate) struct TreeSitter {
     #[serde(default = "chunk_size_default")]
-    pub chunk_size: usize,
+    pub(crate) chunk_size: usize,
     #[serde(default = "chunk_overlap_default")]
-    pub chunk_overlap: usize,
+    pub(crate) chunk_overlap: usize,
 }
 
 impl Default for TreeSitter {
@@ -73,39 +73,39 @@ impl Default for TreeSitter {
 }
 
 #[derive(Debug, Clone, Deserialize)]
-pub struct TextSplitter {
+pub(crate) struct TextSplitter {
     #[serde(default = "chunk_size_default")]
-    pub chunk_size: usize,
+    pub(crate) chunk_size: usize,
 }
 
 #[derive(Debug, Clone, Deserialize, Default)]
-pub struct EmbeddingPrefix {
+pub(crate) struct EmbeddingPrefix {
     #[serde(default)]
-    pub storage: String,
+    pub(crate) storage: String,
     #[serde(default)]
-    pub retrieval: String,
+    pub(crate) retrieval: String,
 }
 
 #[derive(Debug, Clone, Deserialize)]
-pub struct OllamaEmbeddingModel {
+pub(crate) struct OllamaEmbeddingModel {
     // The generate endpoint, default: 'http://localhost:11434/api/embeddings'
-    pub endpoint: Option<String>,
+    pub(crate) endpoint: Option<String>,
     // The model name
-    pub model: String,
+    pub(crate) model: String,
     // The prefix to apply to the embeddings
     #[serde(default)]
-    pub prefix: EmbeddingPrefix,
+    pub(crate) prefix: EmbeddingPrefix,
 }
 
 #[derive(Debug, Clone, Deserialize)]
 #[serde(tag = "type")]
-pub enum ValidEmbeddingModel {
+pub(crate) enum ValidEmbeddingModel {
     #[serde(rename = "ollama")]
     Ollama(OllamaEmbeddingModel),
 }
 
 #[derive(Debug, Clone, Copy, Deserialize)]
-pub enum VectorDataType {
+pub(crate) enum VectorDataType {
     #[serde(rename = "f32")]
     F32,
     #[serde(rename = "binary")]
@@ -114,11 +114,11 @@ pub enum VectorDataType {
 
 #[derive(Debug, Clone, Deserialize)]
 pub(crate) struct VectorStore {
-    pub crawl: Option<Crawl>,
+    pub(crate) crawl: Option<Crawl>,
     #[serde(default)]
-    pub splitter: ValidSplitter,
-    pub embedding_model: ValidEmbeddingModel,
-    pub data_type: VectorDataType,
+    pub(crate) splitter: ValidSplitter,
+    pub(crate) embedding_model: ValidEmbeddingModel,
+    pub(crate) data_type: VectorDataType,
 }
 
 #[derive(Debug, Clone, Deserialize)]
@@ -265,20 +265,20 @@ const fn n_ctx_default() -> u32 {
 #[cfg(feature = "llama_cpp")]
 #[derive(Clone, Debug, Deserialize)]
 #[serde(deny_unknown_fields)]
-pub struct LLaMACPP {
+pub(crate) struct LLaMACPP {
     // Which model to use
-    pub repository: Option<String>,
-    pub name: Option<String>,
-    pub file_path: Option<String>,
+    pub(crate) repository: Option<String>,
+    pub(crate) name: Option<String>,
+    pub(crate) file_path: Option<String>,
     // The layers to put on the GPU
     #[serde(default = "n_gpu_layers_default")]
-    pub n_gpu_layers: u32,
+    pub(crate) n_gpu_layers: u32,
     // The context size
     #[serde(default = "n_ctx_default")]
-    pub n_ctx: u32,
+    pub(crate) n_ctx: u32,
     // The maximum requests per second
     #[serde(default = "max_requests_per_second_default")]
-    pub max_requests_per_second: f32,
+    pub(crate) max_requests_per_second: f32,
 }
 
 #[derive(Clone, Debug, Deserialize)]
@@ -349,7 +349,7 @@ pub(crate) struct Completion {
 }
 
 #[derive(Clone, Debug, Deserialize)]
-pub struct Chat {
+pub(crate) struct Chat {
     // The trigger text
     pub(crate) trigger: String,
     // The name to display in the editor
@@ -362,7 +362,7 @@ pub struct Chat {
 }
 
 #[derive(Clone, Debug, Deserialize)]
-pub struct Action {
+pub(crate) struct Action {
     // The name to display in the editor
     pub(crate) action_display_name: String,
     // The model key to use
@@ -395,13 +395,13 @@ pub(crate) struct ValidClientParams {
 }
 
 #[derive(Clone, Debug)]
-pub struct Config {
+pub(crate) struct Config {
     pub(crate) config: ValidConfig,
     pub(crate) client_params: ValidClientParams,
 }
 
 impl Config {
-    pub fn new(mut args: Value) -> Result<Self> {
+    pub(crate) fn new(mut args: Value) -> Result<Self> {
         // Validate that the models specified are there so we can unwrap
         let configuration_args = args
             .as_object_mut()
@@ -422,23 +422,19 @@ impl Config {
     // Helpers for the backends ///////////
     ///////////////////////////////////////
 
-    pub fn get_chats(&self) -> &Vec<Chat> {
+    pub(crate) fn get_chats(&self) -> &Vec<Chat> {
         &self.config.chats
     }
 
-    pub fn get_actions(&self) -> &Vec<Action> {
+    pub(crate) fn get_actions(&self) -> &Vec<Action> {
         &self.config.actions
     }
 
-    pub fn is_completions_enabled(&self) -> bool {
-        self.config.completion.is_some()
-    }
-
-    pub fn get_completions_post_process(&self) -> Option<&PostProcess> {
+    pub(crate) fn get_completions_post_process(&self) -> Option<&PostProcess> {
         self.config.completion.as_ref().map(|x| &x.post_process)
     }
 
-    pub fn get_completion_transformer_max_requests_per_second(&self) -> anyhow::Result<f32> {
+    pub(crate) fn get_completion_transformer_max_requests_per_second(&self) -> anyhow::Result<f32> {
         match &self
             .config
             .models
@@ -470,7 +466,7 @@ impl Config {
 // For teesting use only
 #[cfg(test)]
 impl Config {
-    pub fn default_with_file_store_without_models() -> Self {
+    pub(crate) fn default_with_file_store_without_models() -> Self {
         Self {
             config: ValidConfig {
                 memory: ValidMemoryBackend::FileStore(FileStore { crawl: None }),
diff --git a/crates/lsp-ai/src/crawl.rs b/crates/lsp-ai/src/crawl.rs
index 1a15f07..2341380 100644
--- a/crates/lsp-ai/src/crawl.rs
+++ b/crates/lsp-ai/src/crawl.rs
@@ -22,7 +22,7 @@ impl Crawl {
     }
 
     #[instrument(skip(self, f))]
-    pub fn maybe_do_crawl(
+    pub(crate) fn maybe_do_crawl(
         &mut self,
         triggered_file: Option<String>,
         mut f: impl FnMut(&config::Crawl, &str) -> anyhow::Result<bool>,
diff --git a/crates/lsp-ai/src/custom_requests/generation_stream.rs b/crates/lsp-ai/src/custom_requests/generation_stream.rs
index a09ed30..7bdb1dc 100644
--- a/crates/lsp-ai/src/custom_requests/generation_stream.rs
+++ b/crates/lsp-ai/src/custom_requests/generation_stream.rs
@@ -5,12 +5,12 @@ pub(crate) enum GenerationStream {}
 
 #[derive(Debug, PartialEq, Clone, Deserialize, Serialize)]
 #[serde(rename_all = "camelCase")]
-pub struct GenerationStreamParams {
-    pub partial_result_token: ProgressToken,
+pub(crate) struct GenerationStreamParams {
+    pub(crate) partial_result_token: ProgressToken,
 
     // This field was "mixed-in" from TextDocumentPositionParams
     #[serde(flatten)]
-    pub text_document_position: TextDocumentPositionParams,
+    pub(crate) text_document_position: TextDocumentPositionParams,
 }
 
 #[derive(Debug, PartialEq, Clone, Deserialize, Serialize)]
diff --git a/crates/lsp-ai/src/embedding_models/mod.rs b/crates/lsp-ai/src/embedding_models/mod.rs
index d125848..8f2cb84 100644
--- a/crates/lsp-ai/src/embedding_models/mod.rs
+++ b/crates/lsp-ai/src/embedding_models/mod.rs
@@ -2,7 +2,7 @@ use crate::config::ValidEmbeddingModel;
 
 mod ollama;
 
-pub fn normalize(mut vector: Vec<f32>) -> Vec<f32> {
+fn normalize(mut vector: Vec<f32>) -> Vec<f32> {
     let magnitude = (vector.iter().map(|&x| x * x).sum::<f32>()).sqrt();
 
     if magnitude != 0.0 {
@@ -15,13 +15,13 @@ pub fn normalize(mut vector: Vec<f32>) -> Vec<f32> {
 }
 
 #[derive(Clone, Copy)]
-pub enum EmbeddingPurpose {
+pub(crate) enum EmbeddingPurpose {
     Storage,
     Retrieval,
 }
 
 #[async_trait::async_trait]
-pub trait EmbeddingModel {
+pub(crate) trait EmbeddingModel {
     async fn embed(
         &self,
         batch: Vec<&str>,
diff --git a/crates/lsp-ai/src/embedding_models/ollama.rs b/crates/lsp-ai/src/embedding_models/ollama.rs
index 2a7b362..a1b030d 100644
--- a/crates/lsp-ai/src/embedding_models/ollama.rs
+++ b/crates/lsp-ai/src/embedding_models/ollama.rs
@@ -8,29 +8,29 @@ use crate::config;
 use super::{normalize, EmbeddingModel, EmbeddingPurpose};
 
 #[derive(Deserialize)]
-pub struct Embed {
+pub(crate) struct Embed {
     embedding: Vec<f32>,
 }
 
 #[derive(Deserialize)]
-pub struct EmbedError {
+pub(crate) struct EmbedError {
     error: Value,
 }
 
 #[derive(Deserialize)]
 #[serde(untagged)]
-pub enum EmbedResponse {
+pub(crate) enum EmbedResponse {
     Success(Embed),
     Error(EmbedError),
     Other(HashMap<String, Value>),
 }
 
-pub struct Ollama {
+pub(crate) struct Ollama {
     config: config::OllamaEmbeddingModel,
 }
 
 impl Ollama {
-    pub fn new(config: config::OllamaEmbeddingModel) -> Self {
+    pub(crate) fn new(config: config::OllamaEmbeddingModel) -> Self {
         Self { config }
     }
 }
diff --git a/crates/lsp-ai/src/memory_backends/file_store.rs b/crates/lsp-ai/src/memory_backends/file_store.rs
index 28a9243..92f95ca 100644
--- a/crates/lsp-ai/src/memory_backends/file_store.rs
+++ b/crates/lsp-ai/src/memory_backends/file_store.rs
@@ -28,7 +28,7 @@ impl AdditionalFileStoreParams {
 }
 
 #[derive(Clone)]
-pub struct File {
+pub(crate) struct File {
     rope: Rope,
     tree: Option<Tree>,
 }
@@ -38,11 +38,11 @@ impl File {
         Self { rope, tree }
     }
 
-    pub fn rope(&self) -> &Rope {
+    pub(crate) fn rope(&self) -> &Rope {
         &self.rope
     }
 
-    pub fn tree(&self) -> Option<&Tree> {
+    pub(crate) fn tree(&self) -> Option<&Tree> {
         self.tree.as_ref()
     }
 }
@@ -278,15 +278,15 @@ impl FileStore {
         })
     }
 
-    pub fn file_map(&self) -> &RwLock<HashMap<String, File>> {
+    pub(crate) fn file_map(&self) -> &RwLock<HashMap<String, File>> {
         &self.file_map
     }
 
-    pub fn contains_file(&self, uri: &str) -> bool {
+    pub(crate) fn contains_file(&self, uri: &str) -> bool {
         self.file_map.read().contains_key(uri)
     }
 
-    pub fn position_to_byte(&self, position: &TextDocumentPositionParams) -> anyhow::Result<usize> {
+    pub(crate) fn position_to_byte(&self, position: &TextDocumentPositionParams) -> anyhow::Result<usize> {
         let file_map = self.file_map.read();
         let uri = position.text_document.uri.to_string();
         let file = file_map
@@ -494,7 +494,7 @@ impl MemoryBackend for FileStore {
 // For testing use only
 #[cfg(test)]
 impl FileStore {
-    pub fn default_with_filler_file() -> anyhow::Result<Self> {
+    pub(crate) fn default_with_filler_file() -> anyhow::Result<Self> {
         let config = Config::default_with_file_store_without_models();
         let file_store_config = if let config::ValidMemoryBackend::FileStore(file_store_config) =
             config.config.memory.clone()
diff --git a/crates/lsp-ai/src/memory_backends/mod.rs b/crates/lsp-ai/src/memory_backends/mod.rs
index 126649b..f9dfb8e 100644
--- a/crates/lsp-ai/src/memory_backends/mod.rs
+++ b/crates/lsp-ai/src/memory_backends/mod.rs
@@ -11,7 +11,7 @@ mod postgresml;
 mod vector_store;
 
 #[derive(Clone, Debug)]
-pub enum PromptType {
+pub(crate) enum PromptType {
     ContextAndCode,
     FIM,
 }
@@ -33,20 +33,20 @@ impl From<&Value> for MemoryRunParams {
 }
 
 #[derive(Debug)]
-pub struct ContextAndCodePrompt {
-    pub context: String,
-    pub code: String,
-    pub selected_text: Option<String>,
+pub(crate) struct ContextAndCodePrompt {
+    pub(crate) context: String,
+    pub(crate) code: String,
+    pub(crate) selected_text: Option<String>,
 }
 
 #[derive(Debug)]
-pub struct FIMPrompt {
-    pub prompt: String,
-    pub suffix: String,
+pub(crate) struct FIMPrompt {
+    pub(crate) prompt: String,
+    pub(crate) suffix: String,
 }
 
 #[derive(Debug)]
-pub enum Prompt {
+pub(crate) enum Prompt {
     FIM(FIMPrompt),
     ContextAndCode(ContextAndCodePrompt),
 }
@@ -96,10 +96,7 @@ impl<'a> TryFrom<&'a Prompt> for &'a FIMPrompt {
 }
 
 #[async_trait::async_trait]
-pub trait MemoryBackend {
-    async fn init(&self) -> anyhow::Result<()> {
-        Ok(())
-    }
+pub(crate) trait MemoryBackend {
     fn opened_text_document(&self, params: DidOpenTextDocumentParams) -> anyhow::Result<()>;
     fn code_action_request(
         &self,
@@ -144,7 +141,7 @@ impl TryFrom<Config> for Box<dyn MemoryBackend + Send + Sync> {
 // easier to just pass in a default prompt.
 #[cfg(test)]
 impl Prompt {
-    pub fn default_with_cursor() -> Self {
+    pub(crate) fn default_with_cursor() -> Self {
         Self::ContextAndCode(ContextAndCodePrompt {
             context: r#"def test_context():\n    pass"#.to_string(),
             code: r#"def test_code():\n    <CURSOR>"#.to_string(),
@@ -152,14 +149,14 @@ impl Prompt {
         })
     }
 
-    pub fn default_fim() -> Self {
+    pub(crate) fn default_fim() -> Self {
         Self::FIM(FIMPrompt {
             prompt: r#"def test_context():\n    pass"#.to_string(),
             suffix: r#"def test_code():\n    "#.to_string(),
         })
     }
 
-    pub fn default_without_cursor() -> Self {
+    pub(crate) fn default_without_cursor() -> Self {
         Self::ContextAndCode(ContextAndCodePrompt {
             context: r#"def test_context():\n    pass"#.to_string(),
             code: r#"def test_code():\n    "#.to_string(),
diff --git a/crates/lsp-ai/src/memory_backends/postgresml/mod.rs b/crates/lsp-ai/src/memory_backends/postgresml/mod.rs
index a003d7c..ab3ecd4 100644
--- a/crates/lsp-ai/src/memory_backends/postgresml/mod.rs
+++ b/crates/lsp-ai/src/memory_backends/postgresml/mod.rs
@@ -79,7 +79,7 @@ pub(crate) struct PostgresML {
 
 impl PostgresML {
     #[instrument]
-    pub fn new(
+    pub(crate) fn new(
         mut postgresml_config: config::PostgresML,
         configuration: Config,
     ) -> anyhow::Result<Self> {
diff --git a/crates/lsp-ai/src/memory_backends/vector_store.rs b/crates/lsp-ai/src/memory_backends/vector_store.rs
index 9d6ed2e..f0dc51a 100644
--- a/crates/lsp-ai/src/memory_backends/vector_store.rs
+++ b/crates/lsp-ai/src/memory_backends/vector_store.rs
@@ -329,7 +329,7 @@ impl VectorStoreInner {
     }
 }
 
-pub struct VectorStore {
+pub(crate) struct VectorStore {
     file_store: Arc<FileStore>,
     crawl: Option<Arc<Mutex<Crawl>>>,
     splitter: Arc<Box<dyn Splitter + Send + Sync>>,
@@ -340,7 +340,7 @@ pub struct VectorStore {
 }
 
 impl VectorStore {
-    pub fn new(
+    pub(crate) fn new(
         mut vector_store_config: config::VectorStore,
         config: Config,
     ) -> anyhow::Result<Self> {
@@ -703,7 +703,7 @@ impl MemoryBackend for VectorStore {
         // Get the embedding
         let embedding = self
             .embedding_model
-            .embed(vec![&query], EmbeddingPurpose::Storage)
+            .embed(vec![&query], EmbeddingPurpose::Retrieval)
             .await?
             .into_iter()
             .nth(0)
diff --git a/crates/lsp-ai/src/splitters/mod.rs b/crates/lsp-ai/src/splitters/mod.rs
index 5592586..6e7a56a 100644
--- a/crates/lsp-ai/src/splitters/mod.rs
+++ b/crates/lsp-ai/src/splitters/mod.rs
@@ -6,13 +6,13 @@ mod text_splitter;
 mod tree_sitter;
 
 #[derive(Debug, Serialize)]
-pub struct ByteRange {
-    pub start_byte: usize,
-    pub end_byte: usize,
+pub(crate) struct ByteRange {
+    pub(crate) start_byte: usize,
+    pub(crate) end_byte: usize,
 }
 
 impl ByteRange {
-    pub fn new(start_byte: usize, end_byte: usize) -> Self {
+    pub(crate) fn new(start_byte: usize, end_byte: usize) -> Self {
         Self {
             start_byte,
             end_byte,
@@ -21,9 +21,9 @@ impl ByteRange {
 }
 
 #[derive(Serialize)]
-pub struct Chunk {
-    pub text: String,
-    pub range: ByteRange,
+pub(crate) struct Chunk {
+    pub(crate) text: String,
+    pub(crate) range: ByteRange,
 }
 
 impl Chunk {
@@ -32,7 +32,7 @@ impl Chunk {
     }
 }
 
-pub trait Splitter {
+pub(crate) trait Splitter {
     fn split(&self, file: &File) -> Vec<Chunk>;
     fn split_file_contents(&self, uri: &str, contents: &str) -> Vec<Chunk>;
 
diff --git a/crates/lsp-ai/src/template.rs b/crates/lsp-ai/src/template.rs
index c98634f..bb79fd5 100644
--- a/crates/lsp-ai/src/template.rs
+++ b/crates/lsp-ai/src/template.rs
@@ -11,7 +11,7 @@ fn template_name_from_template_string(template: &str) -> String {
     xxhash_rust::xxh3::xxh3_64(template.as_bytes()).to_string()
 }
 
-pub fn apply_chat_template(
+pub(crate) fn apply_chat_template(
     template: &str,
     chat_messages: Vec<ChatMessage>,
     bos_token: &str,
diff --git a/crates/lsp-ai/src/transformer_backends/anthropic.rs b/crates/lsp-ai/src/transformer_backends/anthropic.rs
index aab123e..0809a31 100644
--- a/crates/lsp-ai/src/transformer_backends/anthropic.rs
+++ b/crates/lsp-ai/src/transformer_backends/anthropic.rs
@@ -57,7 +57,7 @@ struct AnthropicChatMessage {
 }
 
 #[derive(Deserialize, Serialize)]
-pub struct ChatError {
+pub(crate) struct ChatError {
     error: Value,
 }
 
diff --git a/crates/lsp-ai/src/transformer_backends/llama_cpp/mod.rs b/crates/lsp-ai/src/transformer_backends/llama_cpp/mod.rs
index a1e1ac1..850b1ca 100644
--- a/crates/lsp-ai/src/transformer_backends/llama_cpp/mod.rs
+++ b/crates/lsp-ai/src/transformer_backends/llama_cpp/mod.rs
@@ -23,23 +23,23 @@ const fn max_new_tokens_default() -> usize {
 
 // NOTE: We cannot deny unknown fields as the provided parameters may contain other fields relevant to other processes
 #[derive(Debug, Deserialize)]
-pub struct LLaMACPPRunParams {
-    pub fim: Option<FIM>,
+pub(crate) struct LLaMACPPRunParams {
+    pub(crate) fim: Option<FIM>,
     messages: Option<Vec<ChatMessage>>,
     chat_template: Option<String>, // A Jinja template
     chat_format: Option<String>,   // The name of a template in llamacpp
     #[serde(default = "max_new_tokens_default")]
-    pub max_tokens: usize,
+    pub(crate) max_tokens: usize,
     // TODO: Explore other arguments
 }
 
-pub struct LLaMACPP {
+pub(crate) struct LLaMACPP {
     model: Model,
 }
 
 impl LLaMACPP {
     #[instrument]
-    pub fn new(configuration: config::LLaMACPP) -> anyhow::Result<Self> {
+    pub(crate) fn new(configuration: config::LLaMACPP) -> anyhow::Result<Self> {
         let model_path = match (
             &configuration.file_path,
             &configuration.repository,
diff --git a/crates/lsp-ai/src/transformer_backends/llama_cpp/model.rs b/crates/lsp-ai/src/transformer_backends/llama_cpp/model.rs
index 79e87a5..4a024ba 100644
--- a/crates/lsp-ai/src/transformer_backends/llama_cpp/model.rs
+++ b/crates/lsp-ai/src/transformer_backends/llama_cpp/model.rs
@@ -17,14 +17,14 @@ use super::LLaMACPPRunParams;
 
 static BACKEND: Lazy<LlamaBackend> = Lazy::new(|| LlamaBackend::init().unwrap());
 
-pub struct Model {
+pub(crate) struct Model {
     model: LlamaModel,
     n_ctx: NonZeroU32,
 }
 
 impl Model {
     #[instrument]
-    pub fn new(model_path: PathBuf, config: &config::LLaMACPP) -> anyhow::Result<Self> {
+    pub(crate) fn new(model_path: PathBuf, config: &config::LLaMACPP) -> anyhow::Result<Self> {
         // Initialize the model_params
         let model_params = LlamaModelParams::default().with_n_gpu_layers(config.n_gpu_layers);
 
@@ -42,7 +42,7 @@ impl Model {
     }
 
     #[instrument(skip(self))]
-    pub fn complete(&self, prompt: &str, params: LLaMACPPRunParams) -> anyhow::Result<String> {
+    pub(crate) fn complete(&self, prompt: &str, params: LLaMACPPRunParams) -> anyhow::Result<String> {
         info!("Completing with llama.cpp with prompt:\n{prompt}");
 
         // initialize the context
@@ -129,7 +129,7 @@ impl Model {
     }
 
     #[instrument(skip(self))]
-    pub fn apply_chat_template(
+    pub(crate) fn apply_chat_template(
         &self,
         messages: Vec<ChatMessage>,
         template: Option<String>,
@@ -144,13 +144,13 @@ impl Model {
     }
 
     #[instrument(skip(self))]
-    pub fn get_eos_token(&self) -> anyhow::Result<String> {
+    pub(crate) fn get_eos_token(&self) -> anyhow::Result<String> {
         let token = self.model.token_eos();
         Ok(self.model.token_to_str(token, Special::Tokenize)?)
     }
 
     #[instrument(skip(self))]
-    pub fn get_bos_token(&self) -> anyhow::Result<String> {
+    pub(crate) fn get_bos_token(&self) -> anyhow::Result<String> {
         let token = self.model.token_bos();
         Ok(self.model.token_to_str(token, Special::Tokenize)?)
     }
diff --git a/crates/lsp-ai/src/transformer_backends/mod.rs b/crates/lsp-ai/src/transformer_backends/mod.rs
index f962564..4358ac6 100644
--- a/crates/lsp-ai/src/transformer_backends/mod.rs
+++ b/crates/lsp-ai/src/transformer_backends/mod.rs
@@ -19,7 +19,7 @@ mod ollama;
 mod open_ai;
 
 #[async_trait::async_trait]
-pub trait TransformerBackend {
+pub(crate) trait TransformerBackend {
     async fn do_completion(
         &self,
         prompt: &Prompt,
@@ -38,6 +38,7 @@ pub trait TransformerBackend {
         params: Value,
     ) -> anyhow::Result<DoGenerationResponse>;
 
+    #[allow(dead_code)]
     async fn do_generate_stream(
         &self,
         request: &GenerationStreamRequest,
diff --git a/crates/lsp-ai/src/transformer_backends/ollama.rs b/crates/lsp-ai/src/transformer_backends/ollama.rs
index 678b486..befaebc 100644
--- a/crates/lsp-ai/src/transformer_backends/ollama.rs
+++ b/crates/lsp-ai/src/transformer_backends/ollama.rs
@@ -69,7 +69,7 @@ enum OllamaChatResponse {
 
 impl Ollama {
     #[instrument]
-    pub fn new(configuration: config::Ollama) -> Self {
+    pub(crate) fn new(configuration: config::Ollama) -> Self {
         Self { configuration }
     }
 
diff --git a/crates/lsp-ai/src/transformer_backends/open_ai/mod.rs b/crates/lsp-ai/src/transformer_backends/open_ai/mod.rs
index 0f45254..c75b580 100644
--- a/crates/lsp-ai/src/transformer_backends/open_ai/mod.rs
+++ b/crates/lsp-ai/src/transformer_backends/open_ai/mod.rs
@@ -106,7 +106,7 @@ pub(crate) enum OpenAIChatResponse {
 
 impl OpenAI {
     #[instrument]
-    pub fn new(configuration: config::OpenAI) -> Self {
+    pub(crate) fn new(configuration: config::OpenAI) -> Self {
         Self { configuration }
     }
 
diff --git a/crates/lsp-ai/src/transformer_worker.rs b/crates/lsp-ai/src/transformer_worker.rs
index 94f29bd..4e7b420 100644
--- a/crates/lsp-ai/src/transformer_worker.rs
+++ b/crates/lsp-ai/src/transformer_worker.rs
@@ -54,37 +54,37 @@ impl GenerationRequest {
 // The generate stream is not yet ready but we don't want to remove it
 #[allow(dead_code)]
 #[derive(Clone, Debug)]
-pub struct GenerationStreamRequest {
+pub(crate) struct GenerationStreamRequest {
     id: RequestId,
     params: GenerationStreamParams,
 }
 
 impl GenerationStreamRequest {
-    pub fn new(id: RequestId, params: GenerationStreamParams) -> Self {
+    pub(crate) fn new(id: RequestId, params: GenerationStreamParams) -> Self {
         Self { id, params }
     }
 }
 
 #[derive(Clone, Debug)]
-pub struct CodeActionRequest {
+pub(crate) struct CodeActionRequest {
     id: RequestId,
     params: CodeActionParams,
 }
 
 impl CodeActionRequest {
-    pub fn new(id: RequestId, params: CodeActionParams) -> Self {
+    pub(crate) fn new(id: RequestId, params: CodeActionParams) -> Self {
         Self { id, params }
     }
 }
 
 #[derive(Clone, Debug)]
-pub struct CodeActionResolveRequest {
+pub(crate) struct CodeActionResolveRequest {
     id: RequestId,
     params: CodeAction,
 }
 
 impl CodeActionResolveRequest {
-    pub fn new(id: RequestId, params: CodeAction) -> Self {
+    pub(crate) fn new(id: RequestId, params: CodeAction) -> Self {
         Self { id, params }
     }
 }
@@ -112,16 +112,17 @@ impl WorkerRequest {
     }
 }
 
-pub struct DoCompletionResponse {
-    pub insert_text: String,
+pub(crate) struct DoCompletionResponse {
+    pub(crate) insert_text: String,
 }
 
-pub struct DoGenerationResponse {
-    pub generated_text: String,
+pub(crate) struct DoGenerationResponse {
+    pub(crate) generated_text: String,
 }
 
-pub struct DoGenerationStreamResponse {
-    pub generated_text: String,
+#[allow(dead_code)]
+pub(crate) struct DoGenerationStreamResponse {
+    pub(crate) generated_text: String,
 }
 
 fn post_process_start(response: String, front: &str) -> String {
diff --git a/crates/lsp-ai/src/utils.rs b/crates/lsp-ai/src/utils.rs
index b5d6fcd..5aa44f5 100644
--- a/crates/lsp-ai/src/utils.rs
+++ b/crates/lsp-ai/src/utils.rs
@@ -77,7 +77,7 @@ pub(crate) fn parse_tree(
         .with_context(|| format!("parsing tree failed for {uri}"))
 }
 
-pub fn format_file_chunk(uri: &str, excerpt: &str, root_uri: Option<&str>) -> String {
+pub(crate) fn format_file_chunk(uri: &str, excerpt: &str, root_uri: Option<&str>) -> String {
     let path = match root_uri {
         Some(root_uri) => {
             if uri.starts_with(root_uri) {