added llm_done_list endpoint

Endle · Sep 1, 2024 · b8116d9 · b8116d9
1 parent 03daaa7
commit b8116d9
Show file tree

Hide file tree

Showing 5 changed files with 54 additions and 11 deletions.
diff --git a/fireSeqSearch_addon/main.js b/fireSeqSearch_addon/main.js
@@ -200,10 +200,8 @@ function createFireSeqDom(count) {
 
 async function appendResultToSearchResult(serverInfo, parsedSearchResult, dom) {
     const firefoxExtensionUserOption = await checkUserOptions();
-
     consoleLogForDebug('Loaded user option: ' + JSON.stringify(firefoxExtensionUserOption));
 
-
     function buildListItems(parsedSearchResult) {
         const hitList = document.createElement("ul");
         for (const record of parsedSearchResult) {
@@ -253,14 +251,29 @@ async function appendResultToSearchResult(serverInfo, parsedSearchResult, dom) {
 }
 
 async function processLlmSummary(serverInfo, parsedSearchResult, dom) {
+    async function keepRetryFetch(url) {
+        while(true) {
+            fetch(url, {
+                signal: AbortSignal.timeout(500)
+            }).then(response => {
+                let text = response.text();
+                console.log(text);
+                return text;
+            });
+            //setTimeout(() => {}, 10000);
+            break;
+        }
+        return "abc";
+    }
     for (const record of parsedSearchResult) {
         // TODO remove hard code port
         const llm_api = "http://127.0.0.1:3030/summarize/" + record.title;
         console.log("llm called");
         console.log(record.title);
-        const response = await fetch(llm_api);
-        const text = await response.text();
-        console.log(text);
+        keepRetryFetch(llm_api).then(response => {
+            console.log("returned");
+            console.log(response);
+        });
     }
 }
 
@@ -318,19 +331,18 @@ function getSearchParameterFromCurrentPage() {
 (function() {
     const searchParameter = getSearchParameterFromCurrentPage();
 
-
     addGlobalStyle(fireSeqSearchScriptCSS);
 
+    console.log("main to invoke");
     //https://gomakethings.com/waiting-for-multiple-all-api-responses-to-complete-with-the-vanilla-js-promise.all-method/
     Promise.all([
         fetch("http://127.0.0.1:3030/server_info"),
         fetch("http://127.0.0.1:3030/query/" + searchParameter)
     ]).then(function (responses) {
+        console.log("main to invoke");
         return Promise.all(responses.map(function (response) {return response.json();}));
     }).then(function (data) {
-        //consoleLogForDebug(data);
         mainProcess(data);
-        //return appendResultToSearchResult(data);
     }).then((_e) => {
         const highlightedItems = document.querySelectorAll('.fireSeqSearchHighlight');
         consoleLogForDebug(highlightedItems);

diff --git a/fire_seq_search_server/src/http_client/endpoints.rs b/fire_seq_search_server/src/http_client/endpoints.rs
@@ -1,5 +1,5 @@
 use std::sync::Arc;
-use log::debug;
+use log::{debug, info};
 
 use crate::query_engine::{QueryEngine, ServerInformation};
 use axum::Json;
@@ -30,6 +30,15 @@ pub async fn summarize(
     Html(r.await)
 }
 
+pub async fn get_llm_done_list(
+    State(engine_arc): State<Arc<QueryEngine>>
+    ) -> Html<String>{
+
+    info!("get list endpoint called");
+    let r = engine_arc.get_llm_done_list();
+    Html(r.await)
+}
+
 pub async fn generate_word_cloud(State(engine_arc): State<Arc<QueryEngine>>)
                                                     -> Html<String> {
     let div_id = "fireSeqSearchWordcloudRawJson";

diff --git a/fire_seq_search_server/src/local_llm/mod.rs b/fire_seq_search_server/src/local_llm/mod.rs
@@ -199,7 +199,7 @@ impl LlmEngine{
     pub async fn call_llm_engine(&self) {
 
         let health = self.health().await.unwrap();
-        if (health.slots_idle == 0) {
+        if health.slots_idle == 0 {
             info!("No valid slot, continue");
             return;
         }
@@ -223,8 +223,9 @@ impl LlmEngine{
         }
         drop(jcache);
 
+        info!("Start summarize job:  {}", &title);
         let summarize_result = self.summarize(&doc.body).await;
-        info!("get summarize result {}", &title);
+        info!("Finished summarize job:  {}", &title);
 
         let mut jcache = self.job_cache.lock().await;//.unwrap();
         next_job = jcache.job_queue.pop_front();
@@ -237,6 +238,16 @@ impl LlmEngine{
         return jcache.done_job.get(title).cloned();
     }
 
+    pub async fn get_llm_done_list(&self) -> String {
+        let mut r = Vec::new();
+        let jcache = self.job_cache.lock().await;
+        for (title, _text) in &jcache.done_job {
+            info!("already done : {}", &title);
+            r.push(title.to_owned());
+        }
+        return r.join("\n");
+    }
+
     pub async fn health(&self) -> Result<HealthCheck, Box<dyn std::error::Error>>  {
         let res = self.client.get(self.endpoint.to_owned() + "/health")
             .send()

diff --git a/fire_seq_search_server/src/main.rs b/fire_seq_search_server/src/main.rs
@@ -95,6 +95,7 @@ async fn main() {
         .route("/server_info", get(endpoints::get_server_info))
         .route("/wordcloud", get(endpoints::generate_word_cloud))
         .route("/summarize/:title", get(endpoints::summarize))
+        .route("/llm_done_list", get(endpoints::get_llm_done_list))
         .with_state(engine_arc.clone());
 
     let listener = tokio::net::TcpListener::bind(&engine_arc.server_info.host)

diff --git a/fire_seq_search_server/src/query_engine/mod.rs b/fire_seq_search_server/src/query_engine/mod.rs
@@ -166,6 +166,16 @@ impl QueryEngine {
             "LLM turned off".to_owned()
         }
     }
+    // TODO should serilize the whole vector
+    pub async fn get_llm_done_list(&self) -> String {
+        if cfg!(feature="llm") {
+            let llm = self.llm.as_ref().unwrap();
+            "Finished doc titles:\n====\n".to_owned()
+                + &llm.get_llm_done_list().await
+        } else {
+            "LLM turned off".to_owned()
+        }
+    }
 }
 
 fn term_preprocess(term:String) -> String {