Further cleanup

src/agent/execution/chains/qa_inference_chain.rs

@@ -54,7 +54,7 @@ impl AgentManager {
         // Inference the agent's LLM with the prompt. If it has an answer, the chain
         // is finished and so just return the answer response as a cleaned String
         let response_json = self.inference_agent(agent.clone(), filled_prompt).await?;
-        if let Ok(answer_str) = self.extract_inference_json_response(response_json, "answer") {
+        if let Ok(answer_str) = self.extract_inference_json_response(response_json.clone(), "answer") {
             let cleaned_answer = Self::ending_stripper(&answer_str);
             println!("QA Chain Final Answer: {:?}", cleaned_answer);
             return Ok(cleaned_answer);

src/agent/execution/job_execution_core.rs

@@ -157,13 +157,9 @@ impl AgentManager {
             // TODO: later we should able to grab errors and return them to the user
             let new_scope_entries = match agent_found.clone() {
                 Some(agent) => {
-                    let resp = AgentManager::process_files_inbox(
-                        self.db.clone(),
-                        agent,
-                        job_message.files_inbox.clone(),
-                        profile,
-                    )
-                    .await?;
+                    let resp = self
+                        .process_files_inbox(self.db.clone(), agent, job_message.files_inbox.clone(), profile)
+                        .await?;
                     resp
                 }
                 None => {
@@ -201,6 +197,7 @@ impl AgentManager {
     /// If save_to_db_directly == true, the files will save to the DB and be returned as `DBScopeEntry`s.
     /// Else, the files will be returned as `LocalScopeEntry`s and thus held inside.
     pub async fn process_files_inbox(
+        &self,
         db: Arc<Mutex<ShinkaiDB>>,
         agent: Arc<Mutex<Agent>>,
         files_inbox: String,
@@ -209,55 +206,48 @@ impl AgentManager {
         let _bert_process = BertCPPProcess::start(); // Gets killed if out of scope
         let mut shinkai_db = db.lock().await;
         let files_result = shinkai_db.get_all_files_from_inbox(files_inbox.clone());
-
         // Check if there was an error getting the files
         let files = match files_result {
             Ok(files) => files,
             Err(e) => return Err(AgentError::ShinkaiDB(e)),
         };
-
-        let mut files_map: HashMap<String, LocalScopeEntry> = HashMap::new();
-
         // Create the RemoteEmbeddingGenerator instance
         let generator = Arc::new(RemoteEmbeddingGenerator::new_default());
+        let mut files_map: HashMap<String, LocalScopeEntry> = HashMap::new();
 
+        // Start processing the files
         for (filename, content) in files.into_iter() {
             eprintln!("Iterating over file: {}", filename);
             if filename.ends_with(".pdf") {
                 eprintln!("Processing PDF file: {}", filename);
+                // Prepare description
                 let pdf_overview = FileParser::parse_pdf_for_keywords_and_description(&content, 3, 200)?;
-
-                let agent_clone = agent.clone();
                 let grouped_text_list_clone = pdf_overview.grouped_text_list.clone();
-                let description_response = tokio::spawn(async move {
-                    let mut agent = agent_clone.lock().await;
-                    let prompt = JobPromptGenerator::simple_doc_description(grouped_text_list_clone);
-                    agent.inference(prompt).await
-                })
-                .await?;
-
-                // TODO: Maybe add: "\nKeywords: keywords_generated_by_RAKE"?
-                eprintln!("description_response: {:?}", description_response);
+                let prompt = JobPromptGenerator::simple_doc_description(grouped_text_list_clone);
+                let description_response = self
+                    .inference_agent_and_extract(agent.clone(), prompt, "answer")
+                    .await?;
 
                 let vrsource = Self::create_vrsource(
                     &filename,
                     SourceFileType::Document(SourceDocumentType::Pdf),
                     Some(pdf_overview.blake3_hash),
                 );
-                eprintln!("vrsource: {:?}", vrsource);
                 let doc = FileParser::parse_pdf(
                     &content,
                     150,
                     &*generator,
                     &filename,
-                    Some(&"".to_string()),
+                    Some(&description_response),
                     vrsource,
                     &vec![],
                 )?;
 
+                // TODO: Maybe add: "\nKeywords: keywords_generated_by_RAKE"?
+                eprintln!("description_response: {:?}", description_response);
+                eprintln!("vrsource: {:?}", vrsource);
+
                 let resource = BaseVectorResource::from(doc.clone());
-                // eprintln!("resource: {:?}", resource);
-                eprintln!("profile: {:?}", profile);
                 shinkai_db.init_profile_resource_router(&profile)?;
                 shinkai_db.save_resource(resource, &profile).unwrap();
 

src/agent/execution/job_prompts.rs

@@ -194,11 +194,13 @@ impl JobPromptGenerator {
                 r#"You are an autoregressive language model that has been fine-tuned with instruction-tuning and RLHF. You carefully provide accurate, factual, thoughtful, nuanced answers, and are brilliant at reasoning. If you think there might not be a correct answer, you say so. Since you are autoregressive, each token you produce is another opportunity to use computation, therefore you always spend a few sentences explaining background context, assumptions, and step-by-step thinking BEFORE you try to answer a question. Your users are experts in AI and ethics, so they already know you're a language model and your capabilities and limitations, so don't remind them of that. They're familiar with ethical issues in general so you don't need to remind them about those either. Don't be verbose in your answers, but do provide details and examples where it might help the explanation."#.to_string(),
                 SubPromptType::System,
             );
+
+        prompt.add_content(format!("Here are segements from a document:"), SubPromptType::User);
         for chunk in chunks {
             prompt.add_content(format!("{}", chunk), SubPromptType::User);
         }
         prompt.add_content(
-            format!("Take a deep breath and then throughtouly summarize the previous document, make it as explanatory as possible."),
+            format!("Take a deep breath and thoroughly summarize the previous segments, making it as explanatory as possible."),
             SubPromptType::User,
         );
         prompt.add_ebnf(String::from(r#""{" "answer" ":" string "}""#), SubPromptType::System);