- Added optional chaining (?.) in smart-chat-model, smart-embed-mod…

…el, and smart-entities - Updated Azure adapter API version to latest preview - Improved error handling in `nearest` method of SmartEntities - Added concurrency prevention in `process_embed_queue` method - Fixed potential division by zero in token calculation
brianpetro · Feb 4, 2025 · 8219c54 · 8219c54
1 parent b052047
commit 8219c54
Show file tree

Hide file tree

Showing 7 changed files with 99 additions and 77 deletions.
diff --git a/smart-chat-model/adapters/_api.js b/smart-chat-model/adapters/_api.js
@@ -442,7 +442,7 @@ export class SmartChatModelRequestAdapter {
       body.tool_choice = this.tool_choice;
     }
     // special handling for o1 models
-    if(this.model.startsWith('o1-')){
+    if(this.model?.startsWith('o1-')){
       body.messages = body.messages.filter(m => m.role !== 'system'); // remove system messages (not supported by o1 models)
       delete body.temperature; // not supported by o1 models
     }

diff --git a/smart-chat-model/adapters/azure.js b/smart-chat-model/adapters/azure.js
@@ -21,7 +21,7 @@ export class SmartChatModelAzureAdapter extends SmartChatModelOpenaiAdapter {
     api_key_header: "api-key",
     azure_resource_name: "",
     azure_deployment_name: "",
-    azure_api_version: "2023-05-15",
+    azure_api_version: "2024-10-01-preview",
     default_model: "gpt-35-turbo",
     signup_url: "https://learn.microsoft.com/azure/cognitive-services/openai/quickstart?tabs=command-line",
     models_endpoint: "https://{azure_resource_name}.openai.azure.com/openai/deployments?api-version={azure_api_version}",
@@ -49,12 +49,13 @@ export class SmartChatModelAzureAdapter extends SmartChatModelOpenaiAdapter {
       "[CHAT_ADAPTER].azure_api_version": {
         name: 'Azure API Version',
         type: "text",
-        description: "The API version for Azure OpenAI (e.g. '2023-05-15').",
-        default: "2023-05-15",
+        description: "The API version for Azure OpenAI (e.g. '2024-10-01-preview').",
+        default: "2024-10-01-preview",
       },
     };
   }
 
+
   /**
    * Build the endpoint dynamically based on Azure settings.
    * Example:

diff --git a/smart-embed-model/adapters/transformers.js b/smart-embed-model/adapters/transformers.js
@@ -158,7 +158,7 @@ export class SmartEmbedTransformersAdapter extends SmartEmbedAdapter {
       });
     } catch (err) {
       console.error("error_processing_batch", err);
-      this.pipeline.dispose();
+      this.pipeline?.dispose();
       this.pipeline = null;
       await this.load();
       return Promise.all(batch_inputs.map(async (item) => {

diff --git a/smart-embed-model/connectors/transformers_iframe.js b/smart-embed-model/connectors/transformers_iframe.js
diff --git a/smart-embed-model/connectors/transformers_worker.js b/smart-embed-model/connectors/transformers_worker.js
@@ -277,10 +277,6 @@ var SmartModel = class {
    */
   process_settings_config(_settings_config, prefix = null) {
     return Object.entries(_settings_config).reduce((acc, [key, val]) => {
-      if (val.conditional) {
-        if (!val.conditional(this)) return acc;
-        delete val.conditional;
-      }
       const new_key = (prefix ? prefix + "." : "") + this.process_setting_key(key);
       acc[new_key] = val;
       return acc;
@@ -816,7 +812,7 @@ var SmartEmbedTransformersAdapter = class extends SmartEmbedAdapter {
       });
     } catch (err) {
       console.error("error_processing_batch", err);
-      this.pipeline.dispose();
+      this.pipeline?.dispose();
       this.pipeline = null;
       await this.load();
       return Promise.all(batch_inputs.map(async (item) => {

diff --git a/smart-entities/adapters/default.js b/smart-entities/adapters/default.js
@@ -9,6 +9,7 @@ import { EntitiesVectorAdapter, EntityVectorAdapter } from "./_adapter.js";
 import { cos_sim } from "../utils/cos_sim.js";
 import { results_acc, furthest_acc } from "../utils/results_acc.js";
 import { sort_by_score_ascending, sort_by_score_descending } from "../utils/sort_by_score.js";
+
 /**
  * @class DefaultEntitiesVectorAdapter
  * @extends EntitiesVectorAdapter
@@ -20,8 +21,15 @@ import { sort_by_score_ascending, sort_by_score_descending } from "../utils/sort
 export class DefaultEntitiesVectorAdapter extends EntitiesVectorAdapter {
   constructor(collection) {
     super(collection);
+    /**
+     * Prevents concurrency in process_embed_queue().
+     * @type {boolean}
+     * @private
+     */
+    this._is_processing_embed_queue = false;
     this._reset_embed_queue_stats();
   }
+
   /**
    * Find the nearest entities to the given vector.
    * @async
@@ -94,84 +102,96 @@ export class DefaultEntitiesVectorAdapter extends EntitiesVectorAdapter {
 
   /**
    * Process a queue of entities waiting to be embedded.
-   * Typically, this will call embed_batch in batches and update entities.
+   * Prevents multiple concurrent runs by using `_is_processing_embed_queue`.
    * @async
    * @returns {Promise<void>}
    */
   async process_embed_queue() {
-    const embed_queue = this.collection.embed_queue;
-    // Reset stats as in SmartEntities
-    this._reset_embed_queue_stats();
-
-    if (this.collection.embed_model_key === "None") {
-      console.log(`Smart Connections: No active embedding model for ${this.collection.collection_key}, skipping embedding`);
-      return;
-    }
-
-    if (!this.collection.embed_model) {
-      console.log(`Smart Connections: No active embedding model for ${this.collection.collection_key}, skipping embedding`);
+    if (this._is_processing_embed_queue) {
+      console.log("process_embed_queue is already running, skipping concurrent call.");
       return;
     }
+    this._is_processing_embed_queue = true;
 
-    const datetime_start = new Date();
-    if (!embed_queue.length) {
-      return console.log(`Smart Connections: No items in ${this.collection.collection_key} embed queue`);
-    }
+    try {
+      const embed_queue = this.collection.embed_queue;
+      // Reset stats as in SmartEntities
+      this._reset_embed_queue_stats();
+
+      if (this.collection.embed_model_key === "None") {
+        console.log(`Smart Connections: No active embedding model for ${this.collection.collection_key}, skipping embedding`);
+        return;
+      }
 
-    console.log(`Time spent getting embed queue: ${(new Date()).getTime() - datetime_start.getTime()}ms`);
-    console.log(`Processing ${this.collection.collection_key} embed queue: ${embed_queue.length} items`);
+      if (!this.collection.embed_model) {
+        console.log(`Smart Connections: No active embedding model for ${this.collection.collection_key}, skipping embedding`);
+        return;
+      }
 
-    // Process in batches according to embed_model.batch_size
-    for (let i = 0; i < embed_queue.length; i += this.collection.embed_model.batch_size) {
-      if (this.is_queue_halted) {
-        this.is_queue_halted = false; // reset halt after break
-        break;
+      const datetime_start = new Date();
+      if (!embed_queue.length) {
+        console.log(`Smart Connections: No items in ${this.collection.collection_key} embed queue`);
+        return;
       }
-      const batch = embed_queue.slice(i, i + this.collection.embed_model.batch_size);
 
-      // Prepare input
-      await Promise.all(batch.map(item => item.get_embed_input()));
+      console.log(`Time spent getting embed queue: ${(new Date()).getTime() - datetime_start.getTime()}ms`);
+      console.log(`Processing ${this.collection.collection_key} embed queue: ${embed_queue.length} items`);
 
-      // Embed batch
-      try {
-        const start_time = Date.now();
-        await this.embed_batch(batch);
-        this.total_time += Date.now() - start_time;
-      } catch (e) {
-        if (e && e.message && e.message.includes("API key not set")) {
-          this.halt_embed_queue_processing(`API key not set for ${this.collection.embed_model_key}\nPlease set the API key in the settings.`);
+      // Process in batches according to embed_model.batch_size
+      for (let i = 0; i < embed_queue.length; i += this.collection.embed_model.batch_size) {
+        if (this.is_queue_halted) {
+          this.is_queue_halted = false; // reset halt after break
+          break;
         }
-        console.error(e);
-        console.error(`Error processing ${this.collection.collection_key} embed queue: ` + JSON.stringify((e || {}), null, 2));
-      }
+        const batch = embed_queue.slice(i, i + this.collection.embed_model.batch_size);
+
+        // Prepare input
+        await Promise.all(batch.map(item => item.get_embed_input()));
 
-      // Update hash and stats
-      batch.forEach(item => {
-        item.embed_hash = item.read_hash;
-        item._queue_save = true;
-      });
-      this.embedded_total += batch.length;
-      this.total_tokens += batch.reduce((acc, item) => acc + (item.tokens || 0), 0);
+        // Embed batch
+        try {
+          const start_time = Date.now();
+          await this.embed_batch(batch);
+          this.total_time += Date.now() - start_time;
+        } catch (e) {
+          if (e && e.message && e.message.includes("API key not set")) {
+            this.halt_embed_queue_processing(`API key not set for ${this.collection.embed_model_key}\nPlease set the API key in the settings.`);
+          }
+          console.error(e);
+          console.error(`Error processing ${this.collection.collection_key} embed queue: ` + JSON.stringify((e || {}), null, 2));
+        }
 
-      // Show progress notice every ~100 items
-      this._show_embed_progress_notice(embed_queue.length);
+        // Update hash and stats
+        batch.forEach(item => {
+          item.embed_hash = item.read_hash;
+          item._queue_save = true;
+        });
+        this.embedded_total += batch.length;
+        this.total_tokens += batch.reduce((acc, item) => acc + (item.tokens || 0), 0);
 
-      // Process save queue every 1000 items
-      if (this.embedded_total - this.last_save_total > 1000) {
-        this.last_save_total = this.embedded_total;
-        await this.collection.process_save_queue();
-        if(this.collection.block_collection) {
-          console.log(`Saving ${this.collection.block_collection.collection_key} block collection`);
-          await this.collection.block_collection.process_save_queue();
+        // Show progress notice every ~100 items
+        this._show_embed_progress_notice(embed_queue.length);
+
+        // Process save queue every 1000 items
+        if (this.embedded_total - this.last_save_total > 1000) {
+          this.last_save_total = this.embedded_total;
+          await this.collection.process_save_queue();
+          if(this.collection.block_collection) {
+            console.log(`Saving ${this.collection.block_collection.collection_key} block collection`);
+            await this.collection.block_collection.process_save_queue();
+          }
         }
       }
-    }
 
-    // Show completion notice
-    this._show_embed_completion_notice(embed_queue.length);
-    await this.collection.process_save_queue();
-    if(this.collection.block_collection) {
-      await this.collection.block_collection.process_save_queue();
+      // Show completion notice
+      this._show_embed_completion_notice(embed_queue.length);
+      await this.collection.process_save_queue();
+      if(this.collection.block_collection) {
+        await this.collection.block_collection.process_save_queue();
+      }
+    } finally {
+      // Always clear the concurrency flag, even on errors or halts
+      this._is_processing_embed_queue = false;
     }
   }
 
@@ -190,9 +210,9 @@ export class DefaultEntitiesVectorAdapter extends EntitiesVectorAdapter {
       model_name: this.collection.embed_model_key
     });
   }
+
   /**
    * Displays the embedding completion notice.
-
    * @private
    * @returns {void}
    */
@@ -204,9 +224,9 @@ export class DefaultEntitiesVectorAdapter extends EntitiesVectorAdapter {
       model_name: this.collection.embed_model_key
     });
   }
+
   /**
    * Halts the embed queue processing.
-
    * @param {string|null} msg - Optional message.
    */
   halt_embed_queue_processing(msg=null) {
@@ -220,9 +240,9 @@ export class DefaultEntitiesVectorAdapter extends EntitiesVectorAdapter {
       model_name: this.collection.embed_model_key
     });
   }
+
   /**
    * Resumes the embed queue processing after a delay.
-
    * @param {number} [delay=0] - The delay in milliseconds before resuming.
    * @returns {void}
    */
@@ -234,15 +254,17 @@ export class DefaultEntitiesVectorAdapter extends EntitiesVectorAdapter {
       this.process_embed_queue();
     }, delay);
   }
+
   /**
    * Calculates the number of tokens processed per second.
    * @private
    * @returns {number} Tokens per second.
    */
   _calculate_embed_tokens_per_second() {
     const elapsed_time = this.total_time / 1000;
-    return Math.round(this.total_tokens / elapsed_time);
+    return Math.round(this.total_tokens / (elapsed_time || 1));
   }
+
   /**
    * Resets the statistics related to embed queue processing.
    * @private
@@ -317,10 +339,9 @@ export class DefaultEntityVectorAdapter extends EntityVectorAdapter {
     }
     this.item.data.embeddings[this.item.embed_model_key].vec = vec;
   }
-
 }
 
 export default {
   collection: DefaultEntitiesVectorAdapter,
   item: DefaultEntityVectorAdapter
-};
+};
diff --git a/smart-entities/smart_entities.js b/smart-entities/smart_entities.js
@@ -154,10 +154,14 @@ export class SmartEntities extends Collection {
    * @returns {Promise<Array<{item:Object, score:number}>>} An array of result objects with score and item.
    */
   async nearest(vec, filter = {}) {
-    if (!vec) return console.warn("nearest: no vec");
+    if (!vec) {
+      console.warn("nearest: no vec");
+      return [];
+    }
     return await this.entities_vector_adapter.nearest(vec, filter);
   }
 
+
   /**
    * Finds the furthest entities from a vector using the default adapter.
    * @async