@llama-node/llama-cpp - npm Package Compare versions

Comparing version 0.0.22 to 0.0.23

index.d.ts

@@ -27,4 +27,4 @@ /* tslint:disable */
 export const enum TokenizeResultType {
-  Error = 0,
-  Data = 1
+  Error = 'Error',
+  Data = 'Data'
 }
@@ -40,5 +40,5 @@ export interface TokenizeResult {
 export const enum InferenceResultType {
-  Error = 0,
-  Data = 1,
-  End = 2
+  Error = 'Error',
+  Data = 'Data',
+  End = 'End'
 }
@@ -51,4 +51,4 @@ export interface InferenceResult {
 export const enum EmbeddingResultType {
-  Error = 0,
-  Data = 1
+  Error = 'Error',
+  Data = 'Data'
 }
@@ -61,11 +61,5 @@ export interface EmbeddingResult {
   static load(path: string, params: LlamaContextParams | undefined | null, enableLogger: boolean): LLama
-  getWordEmbedding(input: LlamaInvocation,
-  callback: (result: EmbeddingResult) => void): void
-  tokenize(params: string,
-  nCtx: number,
-  callback: (result:
-  { type: TokenizeResultType, data: number[] }
-  ) => void): void
-  inference(input: LlamaInvocation,
-  callback: (result: InferenceResult) => void): void
+  getWordEmbedding(input: LlamaInvocation, callback: (result: EmbeddingResult) => void): void
+  tokenize(params: string, nCtx: number, callback: (result: TokenizeResult) => void): void
+  inference(input: LlamaInvocation, callback: (result: InferenceResult) => void): void
 }

package.json

 {
   "name": "@llama-node/llama-cpp",
-  "version": "0.0.22",
+  "version": "0.0.23",
   "main": "index.js",
@@ -5,0 +5,0 @@ "types": "index.d.ts",

src/lib.rs

@@ -26,3 +26,3 @@ #![deny(clippy::all)]
 };
-use types::{InferenceResult, TokenizeResult, EmbeddingResult};
+use types::{EmbeddingResult, InferenceResult, TokenizeResult};
 
@@ -68,5 +68,8 @@ #[napi]
 
-    #[napi(ts_args_type = "input: LlamaInvocation,
-        callback: (result: EmbeddingResult) => void")]
-    pub fn get_word_embedding(&self, input: LlamaInvocation, callback: JsFunction) -> Result<()> {
+    #[napi]
+    pub fn get_word_embedding(
+        &self,
+        input: LlamaInvocation,
+        #[napi(ts_arg_type = "(result: EmbeddingResult) => void")] callback: JsFunction,
+    ) -> Result<()> {
         let tsfn: ThreadsafeFunction<EmbeddingResult, ErrorStrategy::Fatal> =
@@ -98,8 +101,9 @@ callback.create_threadsafe_function(0, |ctx| Ok(vec![ctx.value]))?;
 
-    #[napi(ts_args_type = "params: string,
-    nCtx: number,
-    callback: (result:
-      { type: TokenizeResultType, data: number[] }
-    ) => void")]
-    pub fn tokenize(&self, params: String, n_ctx: i32, callback: JsFunction) -> Result<()> {
+    #[napi]
+    pub fn tokenize(
+        &self,
+        params: String,
+        n_ctx: i32,
+        #[napi(ts_arg_type = "(result: TokenizeResult) => void")] callback: JsFunction,
+    ) -> Result<()> {
         let (tokenize_sender, tokenize_receiver) = channel::<TokenizeResult>();
@@ -136,5 +140,8 @@
 
-    #[napi(ts_args_type = "input: LlamaInvocation,
-        callback: (result: InferenceResult) => void")]
-    pub fn inference(&self, input: LlamaInvocation, callback: JsFunction) -> Result<()> {
+    #[napi]
+    pub fn inference(
+        &self,
+        input: LlamaInvocation,
+        #[napi(ts_arg_type = "(result: InferenceResult) => void")] callback: JsFunction,
+    ) -> Result<()> {
         let tsfn: ThreadsafeFunction<InferenceResult, ErrorStrategy::Fatal> =
@@ -141,0 +148,0 @@ callback.create_threadsafe_function(0, |ctx| Ok(vec![ctx.value]))?;

src/llama.rs

@@ -11,3 +11,3 @@ use std::{
     context::{LLamaContext, LlamaContextParams, LlamaInvocation},
-    tokenizer::{embedding_to_output, llama_token_eos, tokenize},
+    tokenizer::{llama_token_eos, tokenize},
     types::{
@@ -126,4 +126,2 @@ EmbeddingResult, EmbeddingResultType, InferenceResult, InferenceResultType, InferenceToken,
 
-        log::info!("hard coded token_eos: {}", token_eos);
-
         // Generate remaining tokens.
@@ -133,2 +131,3 @@ let mut n_remaining = context_params_c.n_ctx - tokenized_input.len() as i32;
         let mut stop_sequence_i = 0;
+        let mut completed = false;
         while n_remaining > 0 {
@@ -140,9 +139,3 @@ let tok = input_ctx.llama_sample(embd.as_slice(), n_used as i32, input);
             if tok == token_eos {
-                sender
-                    .send(InferenceResult {
-                        r#type: InferenceResultType::End,
-                        data: None,
-                        message: None,
-                    })
-                    .unwrap();
+                completed = true;
                 break;
@@ -167,9 +160,3 @@ }
                     if stop_sequence_i >= tokenized_stop_prompt.len() {
-                        sender
-                            .send(InferenceResult {
-                                r#type: InferenceResultType::End,
-                                data: None,
-                                message: None,
-                            })
-                            .unwrap();
+                        completed = true;
                         break;
@@ -201,6 +188,27 @@ }
         }
-        embedding_to_output(
-            input_ctx,
-            &embd[tokenized_input.len()..n_used + 1 - stop_sequence_i],
-        );
+
+        if completed {
+            sender
+                .send(InferenceResult {
+                    r#type: InferenceResultType::Data,
+                    data: Some(InferenceToken {
+                        token: "\n\n<end>\n".to_string(),
+                        completed: true,
+                    }),
+                    message: None,
+                })
+                .unwrap();
+        }
+
+        sender
+            .send(InferenceResult {
+                r#type: InferenceResultType::End,
+                data: None,
+                message: None,
+            })
+            .unwrap();
+        // embedding_to_output(
+        //     input_ctx,
+        //     &embd[tokenized_input.len()..n_used + 1 - stop_sequence_i],
+        // );
     }
@@ -207,0 +215,0 @@ }

src/types.rs

@@ -12,3 +12,3 @@ use crate::context::LlamaInvocation;
 
-#[napi]
+#[napi(string_enum)]
 pub enum TokenizeResultType {
@@ -32,3 +32,3 @@ Error,
 
-#[napi]
+#[napi(string_enum)]
 pub enum InferenceResultType {
@@ -47,3 +47,3 @@ Error,
 
-#[napi]
+#[napi(string_enum)]
 pub enum EmbeddingResultType {
@@ -50,0 +50,0 @@ Error,

New alerts

Shell access

Supply chain risk

This module accesses the system shell. Accessing the system shell increases the risk of executing arbitrary code.

Found 1 instance in 1 package

Filesystem access

Supply chain risk

Accesses the file system, and could potentially read sensitive data.

Found 1 instance in 1 package

Fixed alerts

Shell access

Supply chain risk

This module accesses the system shell. Accessing the system shell increases the risk of executing arbitrary code.

Found 1 instance in 1 package

Filesystem access

Supply chain risk

Accesses the file system, and could potentially read sensitive data.

Found 1 instance in 1 package

Improved metrics

Total package byte prevSize

10709448

0.03%

Worsened metrics

Lines of code

502

-1.18%

No dependency changes