Read Hugging Face Hub token from cache if not provided (#814)

Author: alvarobartt

README.md

@@ -249,7 +249,7 @@ Options:
           [env: DENSE_PATH=]
 
       --hf-token <HF_TOKEN>
-          Your Hugging Face Hub token
+          Your Hugging Face Hub token. If neither `--hf-token` nor `HF_TOKEN` is set, the token will be read from the `$HF_HOME/token` path, if it exists. This ensures access to private or gated models, and allows for a more permissive rate limiting
 
           [env: HF_TOKEN=]
 
@@ -297,6 +297,8 @@ Options:
           [env: JSON_OUTPUT=]
 
       --disable-spans
+          Whether or not to include the log trace through spans
+
           [env: DISABLE_SPANS=]
 
       --otlp-endpoint <OTLP_ENDPOINT>

docs/source/en/cli_arguments.md

@@ -134,7 +134,7 @@ Options:
           [env: DENSE_PATH=]
 
       --hf-token <HF_TOKEN>
-          Your Hugging Face Hub token
+          Your Hugging Face Hub token. If neither `--hf-token` nor `HF_TOKEN` is set, the token will be read from the `$HF_HOME/token` path, if it exists. This ensures access to private or gated models, and allows for a more permissive rate limiting
 
           [env: HF_TOKEN=]
 
@@ -182,6 +182,8 @@ Options:
           [env: JSON_OUTPUT=]
 
       --disable-spans
+          Whether or not to include the log trace through spans
+
           [env: DISABLE_SPANS=]
 
       --otlp-endpoint <OTLP_ENDPOINT>

router/src/lib.rs

@@ -73,14 +73,18 @@ pub async fn run(
         // Using a local model
         (model_id_path.to_path_buf(), None)
     } else {
-        let mut builder = ApiBuilder::from_env()
-            .with_progress(false)
-            .with_token(hf_token);
+        let mut builder = ApiBuilder::from_env().with_progress(false);
 
         if let Some(cache_dir) = huggingface_hub_cache {
             builder = builder.with_cache_dir(cache_dir.into());
         }
 
+        // NOTE: Only set the `token` if it's not None, otherwise leave it as default so that the
+        // token from the cache location is pulled instead, if exists
+        if hf_token.is_some() {
+            builder = builder.with_token(hf_token);
+        }
+
         if let Ok(origin) = std::env::var("HF_HUB_USER_AGENT_ORIGIN") {
             builder = builder.with_user_agent("origin", origin.as_str());
         }

router/src/main.rs

@@ -1,9 +1,10 @@
 use anyhow::Result;
 use clap::Parser;
 use opentelemetry::global;
-use text_embeddings_backend::DType;
 use veil::Redact;
 
+use text_embeddings_backend::DType;
+
 #[cfg(not(target_os = "linux"))]
 #[global_allocator]
 static GLOBAL: mimalloc::MiMalloc = mimalloc::MiMalloc;
@@ -132,7 +133,9 @@ struct Args {
     #[redact(partial)]
     hf_api_token: Option<String>,
 
-    /// Your Hugging Face Hub token
+    /// Your Hugging Face Hub token. If neither `--hf-token` nor `HF_TOKEN` is set, the token
+    /// will be read from the `$HF_HOME/token` path, if it exists. This ensures access to private
+    /// or gated models, and allows for a more permissive rate limiting.
     #[clap(long, env, conflicts_with = "hf_api_token")]
     #[redact(partial)]
     hf_token: Option<String>,
@@ -171,7 +174,7 @@ struct Args {
     #[clap(long, env)]
     json_output: bool,
 
-    // Whether or not to include the log trace through spans
+    /// Whether or not to include the log trace through spans
     #[clap(long, env)]
     disable_spans: bool,