working example

Cargo.toml

@@ -4,4 +4,4 @@ resolver = "1"
 members = [
     "orca",
     "examples/*"
-]
+]

examples/pdf/Cargo.toml

@@ -11,3 +11,5 @@ anyhow = "1.0.75"
 tokio = { version = "1.12.0", features = ["full"] }
 clap = "4.4.7"
 serde_json = "1.0.108"
+env_logger = "0.10.0"
+rayon = "1.8.0"

examples/pdf/src/main.rs

@@ -10,11 +10,13 @@ use orca::chains::Chain;
 use orca::llm::bert::Bert;
 use orca::llm::openai::OpenAI;
 use orca::llm::Embedding;
-use orca::prompt;
 use orca::qdrant::Qdrant;
 use orca::qdrant::Value;
+use orca::record::pdf;
 use orca::record::pdf::Pdf;
 use orca::record::Spin;
+use orca::{prompt, prompts};
+use rayon::prelude::*;
 use serde_json::json;
 
 #[derive(Parser, Debug)]
@@ -38,30 +40,30 @@ struct Args {
 async fn main() -> Result<()> {
     let args = Args::parse();
 
-    // print pwd
-    println!("pwd: {:?}", std::env::current_dir()?);
+    // init logger
+    env_logger::init();
 
     let collection = if let Some(col) = args.collection {
         col
     } else {
         args.file.split("/").last().unwrap().split(".").next().unwrap().to_string()
     };
 
-    let pdf_records = Pdf::from_file(&args.file, false).spin()?.split(1000);
+    let pdf_records = Pdf::from_file(&args.file, false).spin()?.split(500);
     let bert = Bert::new().build_model_and_tokenizer().await?;
 
     let qdrant = Qdrant::new("localhost", 6334);
     if qdrant.create_collection(&collection, 384).await.is_ok() {
         let mut embeddings = Vec::new();
         for record in &pdf_records {
             let embedding = bert.generate_embedding(prompt!(record)).await?;
-            embeddings.push(embedding.get_embedding()?);
+            embeddings.push(embedding.to_vec()?);
         }
-        qdrant.insert_many(&collection, embeddings, pdf_records).await?;
+        qdrant.insert_many(&collection, embeddings.clone(), pdf_records).await?;
     }
 
     let query_embedding = bert.generate_embedding(prompt!(args.prompt)).await?;
-    let result = qdrant.search(&collection, query_embedding.get_embedding()?, 5, None).await?;
+    let result = qdrant.search(&collection, query_embedding.to_vec()?.clone(), 5, None).await?;
 
     let context = json!({
         "user_prompt": args.prompt,
@@ -76,26 +78,19 @@ async fn main() -> Result<()> {
             .collect::<Vec<String>>()
     });
 
-    println!("Context: {:#?}", context);
-
     let prompt_for_model = r#"
     {{#chat}}
         {{#system}}
-        You are a highly advanced assistant. You receive a prompt from a user and relevant excerpts extracted from a PDF.
-        You then answer truthfully to the best of your ability. If you do not know the answer, your response is "I don't know".
+        You are a highly advanced assistant. You receive a prompt from a user and relevant excerpts extracted from a PDF. You then answer truthfully to the best of your ability. If you do not know the answer, your response is I don't know.
         {{/system}}
-        
         {{#user}}
-        '{{user_prompt}}'.
+        {{user_prompt}}
         {{/user}}
-        
         {{#system}}
         Based on the retrieved information from the PDF, here are the relevant excerpts:
-        
         {{#each payloads}}
         {{this}}
         {{/each}}
-        
         Please provide a comprehensive answer to the user's question, integrating insights from these excerpts and your general knowledge.
         {{/system}}
     {{/chat}}