@aj-archipelago/cortex 1.1.3 → 1.1.4

package/server/plugins/cohereSummarizePlugin.js

@@ -2,8 +2,8 @@
 import ModelPlugin from './modelPlugin.js';
 
 class CohereSummarizePlugin extends ModelPlugin {
-    constructor(config, pathway, modelName, model) {
-        super(config, pathway, modelName, model);
+    constructor(pathway, model) {
+        super(pathway, model);
     }
 
     // Set up parameters specific to the Cohere Summarize API
@@ -23,17 +23,10 @@ class CohereSummarizePlugin extends ModelPlugin {
     }
 
     // Execute the request to the Cohere Summarize API
-    async execute(text, parameters, prompt, pathwayResolver) {
-        const url = this.requestUrl();
+    async execute(text, parameters, prompt, cortexRequest) {
         const requestParameters = this.getRequestParameters(text, parameters, prompt);
-        const { requestId, pathway} = pathwayResolver;
-
-        const data = { ...(this.model.params || {}), ...requestParameters };
-        const params = {};
-        const headers = { 
-            ...this.model.headers || {}
-        };
-        return this.executeRequest(url, data, params, headers, prompt, requestId, pathway);
+        cortexRequest.data = { ...cortexRequest.data, ...requestParameters };
+        return this.executeRequest(cortexRequest);
     }
 
     // Parse the response from the Cohere Summarize API

package/server/plugins/localModelPlugin.js

@@ -5,8 +5,8 @@ import { encode } from 'gpt-3-encoder';
 import logger from '../../lib/logger.js';
 
 class LocalModelPlugin extends ModelPlugin {
-    constructor(config, pathway, modelName, model) {
-        super(config, pathway, modelName, model);
+    constructor(pathway, model) {
+        super(pathway, model);
     }
 
     // if the input starts with a chatML response, just return that
@@ -47,7 +47,7 @@ class LocalModelPlugin extends ModelPlugin {
         };
     }
 
-    async execute(text, parameters, prompt, _pathwayResolver) {
+    async execute(text, parameters, prompt, _cortexRequest) {
         const requestParameters = this.getRequestParameters(text, parameters, prompt);
         const { executablePath, args } = this.model;
         args.push("--prompt", requestParameters.prompt);

package/server/plugins/modelPlugin.js

@@ -1,18 +1,18 @@
 // ModelPlugin.js
 import HandleBars from '../../lib/handleBars.js';
-
-import { request } from '../../lib/request.js';
+import { executeRequest } from '../../lib/requestExecutor.js';
 import { encode } from 'gpt-3-encoder';
 import { getFirstNToken } from '../chunker.js';
 import logger, { obscureUrlParams } from '../../lib/logger.js';
+import { config } from '../../config.js';
 
 const DEFAULT_MAX_TOKENS = 4096;
 const DEFAULT_MAX_RETURN_TOKENS = 256;
 const DEFAULT_PROMPT_TOKEN_RATIO = 0.5;
 
 class ModelPlugin {
-    constructor(config, pathway, modelName, model) {
-        this.modelName = modelName;
+    constructor(pathway, model) {
+        this.modelName = model.name;
         this.model = model;
         this.config = config;
         this.environmentVariables = config.getEnv();
@@ -33,7 +33,6 @@ class ModelPlugin {
 
         this.requestCount = 0;
         this.lastRequestStartTime = new Date();
-        this.shouldCache = config.get('enableCache') && (pathway.enableCache || pathway.temperature == 0);
     }
 
     truncateMessagesToTargetLength(messages, targetTokenLength) {
@@ -220,14 +219,14 @@ class ModelPlugin {
     parseResponse(data) { return data; }
 
     // Default simple logging
-    logRequestStart(url, _data) {
+    logRequestStart() {
         this.requestCount++;
         this.lastRequestStartTime = new Date();
         const logMessage = `>>> [${this.requestId}: ${this.pathwayName}.${this.requestCount}] request`;
         const header = '>'.repeat(logMessage.length);
         logger.info(`${header}`);
         logger.info(`${logMessage}`);
-        logger.info(`>>> Making API request to ${obscureUrlParams(url)}`);
+        logger.info(`>>> Making API request to ${obscureUrlParams(this.url)}`);
     }
 
     logAIRequestFinished() {
@@ -257,15 +256,22 @@ class ModelPlugin {
         prompt && prompt.debugInfo && (prompt.debugInfo += `\n${JSON.stringify(data)}`);
     }
     
-    async executeRequest(url, data, params, headers, prompt, requestId, pathway) {
+    async executeRequest(cortexRequest) {
         try {
-            this.aiRequestStartTime = new Date();
+            const { url, data, pathway, requestId, prompt } = cortexRequest;
+            this.url = url;
             this.requestId = requestId;
-            this.logRequestStart(url, data);
-            const responseData = await request({ url, data, params, headers, cache: this.shouldCache }, this.modelName, this.requestId, pathway);
+            this.pathwayName = pathway.name;
+            this.pathwayPrompt = pathway.prompt;
+
+            cortexRequest.cache = config.get('enableCache') && (pathway.enableCache || pathway.temperature == 0);
+            this.logRequestStart();
+
+            const responseData = await executeRequest(cortexRequest);
             
             if (responseData.error) {
-                throw new Error(`An error was returned from the server: ${JSON.stringify(responseData.error)}`);
+                logger.error(`An error was returned from the server: ${JSON.stringify(responseData.error)}`);
+                throw responseData;
             }
         
             this.logRequestData(data, responseData, prompt);

package/server/plugins/openAiChatExtensionPlugin.js

@@ -2,8 +2,8 @@
 import OpenAIChatPlugin from './openAiChatPlugin.js';
 
 class OpenAIChatExtensionPlugin extends OpenAIChatPlugin {
-    constructor(config, pathway, modelName, model) {
-        super(config, pathway, modelName, model);
+    constructor(pathway, model) {
+        super(pathway, model);
         this.tool = '';
     }
 
@@ -47,9 +47,9 @@ class OpenAIChatExtensionPlugin extends OpenAIChatPlugin {
         return reqParams;
     }
 
-    async execute(text, parameters, prompt, pathwayResolver) {
-        const result = await super.execute(text, parameters, prompt, pathwayResolver);
-        pathwayResolver.tool = this.tool; // add tool info back 
+    async execute(text, parameters, prompt, cortexRequest) {
+        const result = await super.execute(text, parameters, prompt, cortexRequest);
+        cortexRequest.pathwayResolver.tool = this.tool; // add tool info back 
         return result;
     }
 

package/server/plugins/openAiChatPlugin.js

@@ -4,8 +4,8 @@ import { encode } from 'gpt-3-encoder';
 import logger from '../../lib/logger.js';
 
 class OpenAIChatPlugin extends ModelPlugin {
-    constructor(config, pathway, modelName, model) {
-        super(config, pathway, modelName, model);
+    constructor(pathway, model) {
+        super(pathway, model);
     }
 
     // convert to OpenAI messages array format if necessary
@@ -76,16 +76,14 @@ class OpenAIChatPlugin extends ModelPlugin {
         return requestParameters;
     }
 
-    // Execute the request to the OpenAI Chat API
-    async execute(text, parameters, prompt, pathwayResolver) {
-        const url = this.requestUrl(text);
+    // Assemble and execute the request to the OpenAI Chat API
+    async execute(text, parameters, prompt, cortexRequest) {
         const requestParameters = this.getRequestParameters(text, parameters, prompt);
-        const { requestId, pathway} = pathwayResolver;
 
-        const data = { ...(this.model.params || {}), ...requestParameters };
-        const params = {}; // query params
-        const headers = this.model.headers || {};
-        return this.executeRequest(url, data, params, headers, prompt, requestId, pathway);
+        cortexRequest.data = { ...(cortexRequest.data || {}), ...requestParameters };
+        cortexRequest.params = {}; // query params
+
+        return this.executeRequest(cortexRequest);
     }
 
     // Parse the response from the OpenAI Chat API

package/server/plugins/openAiCompletionPlugin.js

@@ -16,8 +16,8 @@ const truncatePromptIfNecessary = (text, textTokenCount, modelMaxTokenCount, tar
 }
 
 class OpenAICompletionPlugin extends ModelPlugin {
-    constructor(config, pathway, modelName, model) {
-        super(config, pathway, modelName, model);
+    constructor(pathway, model) {
+        super(pathway, model);
     }
 
     // Set up parameters specific to the OpenAI Completion API
@@ -76,16 +76,13 @@ class OpenAICompletionPlugin extends ModelPlugin {
     }
 
     // Execute the request to the OpenAI Completion API
-    async execute(text, parameters, prompt, pathwayResolver) {
-        const url = this.requestUrl(text);
-        const requestParameters = this.getRequestParameters(text, parameters, prompt, pathwayResolver);
-        const { requestId, pathway} = pathwayResolver;
-
-        const data = { ...(this.model.params || {}), ...requestParameters };
-        const params = {};
-        const headers = this.model.headers || {};
-        
-        return this.executeRequest(url, data, params, headers, prompt, requestId, pathway);
+    async execute(text, parameters, prompt, cortexRequest) {
+        const requestParameters = this.getRequestParameters(text, parameters, prompt, cortexRequest.pathwayResolver);
+
+        cortexRequest.data = { ...(cortexRequest.data || {}), ...requestParameters };
+        cortexRequest.params = {};
+
+        return this.executeRequest(cortexRequest);
     }
 
     // Parse the response from the OpenAI Completion API

package/server/plugins/openAiDallE3Plugin.js

@@ -1,17 +1,15 @@
-import RequestDurationEstimator from '../../lib/requestDurationEstimator.js';
+import RequestMonitor from '../../lib/requestMonitor.js';
 import ModelPlugin from './modelPlugin.js';
-import { request } from '../../lib/request.js';
 import { publishRequestProgress } from '../../lib/redisSubscription.js';
-import logger from '../../lib/logger.js';
 
-const requestDurationEstimator = new RequestDurationEstimator(10);
+const requestDurationEstimator = new RequestMonitor(10);
 
 /**
  * @description This plugin is for the OpenAI DALL-E 3 model.
  */
 class OpenAIDallE3Plugin extends ModelPlugin {
-    constructor(config, pathway, modelName, model) {
-        super(config, pathway, modelName, model);
+    constructor(pathway, model) {
+        super(pathway, model);
     }
 
     /**
@@ -20,28 +18,13 @@ class OpenAIDallE3Plugin extends ModelPlugin {
      * over a websocket.
      */
 
-    async executeRequest(url, data, params, headers, prompt, requestId, pathway) {
-        try {
-            this.aiRequestStartTime = new Date();
-            this.requestId = requestId;
-            this.logRequestStart(url, data);
-            const responseData = await request({ url, data, params, headers, cache: this.shouldCache }, this.modelName, this.requestId, pathway);
-               
-            this.logRequestData(data, responseData, prompt);
-            return this.parseResponse(responseData);
-        } catch (error) {
-            // Log the error and continue
-            logger.error(error.message || error);
-        }
-    }
-
-    async execute(text, parameters, _, pathwayResolver) {
-        const url = this.requestUrl(text);
-        const data = JSON.stringify({ prompt: text });
+    async execute(text, parameters, _, cortexRequest) {
+        const { pathwayResolver } = cortexRequest;
+        cortexRequest.data = JSON.stringify({ prompt: text });
 
-        const { requestId, pathway } = pathwayResolver;
+        const { requestId } = pathwayResolver;
 
-        const makeRequest = () => this.executeRequest(url, data, {}, this.model.headers, {}, requestId, pathway);
+        const makeRequest = () => this.executeRequest(cortexRequest);
 
         if (!parameters.async) {
             // synchronous request
@@ -49,9 +32,9 @@ class OpenAIDallE3Plugin extends ModelPlugin {
         }
         else {
             // async request
-            requestDurationEstimator.startRequest(requestId);
+            const callid = requestDurationEstimator.startCall();
             const requestPromise = makeRequest();
-            this.#sendRequestUpdates(requestId, requestPromise);
+            this.#sendRequestUpdates(requestId, requestPromise, callid);
         }
     }
 
@@ -62,7 +45,7 @@ class OpenAIDallE3Plugin extends ModelPlugin {
      * @param {*} requestPromise 
      * @returns 
      */
-    async #sendRequestUpdates(requestId, requestPromise) {
+    async #sendRequestUpdates(requestId, requestPromise, callid) {
         let state = { status: "pending" };
         let attemptCount = 0;
         let data = null;
@@ -83,7 +66,7 @@ class OpenAIDallE3Plugin extends ModelPlugin {
             };
 
             state.status = status;
-            requestDurationEstimator.endRequest();
+            requestDurationEstimator.endCall(callid);
             publishRequestProgress(requestProgress);
         }
 
@@ -91,7 +74,7 @@ class OpenAIDallE3Plugin extends ModelPlugin {
         // the percent complete
         do {
             let progress =
-                requestDurationEstimator.calculatePercentComplete();
+                requestDurationEstimator.calculatePercentComplete(callid);
 
                 await publishRequestProgress({
                     requestId,

package/server/plugins/openAiEmbeddingsPlugin.js

@@ -2,8 +2,8 @@
 import ModelPlugin from './modelPlugin.js';
 
 class OpenAiEmbeddingsPlugin extends ModelPlugin {
-    constructor(config, pathway, modelName, model) {
-        super(config, pathway, modelName, model);
+    constructor(pathway, model) {
+        super(pathway, model);
     }
 
     getRequestParameters(text, parameters, prompt) {
@@ -17,16 +17,13 @@ class OpenAiEmbeddingsPlugin extends ModelPlugin {
         return requestParameters;
     }
 
-    async execute(text, parameters, prompt, pathwayResolver) {
-        const { requestId, pathway} = pathwayResolver;
+    async execute(text, parameters, prompt, cortexRequest) {
         const requestParameters = this.getRequestParameters(text, parameters, prompt);
-        const url = this.requestUrl();
 
-        const data = requestParameters.data || {};
-        const params = requestParameters.params || {};
-        const headers = this.model.headers || {};
+        cortexRequest.data = requestParameters.data || {};
+        cortexRequest.params = requestParameters.params || {};
 
-        return this.executeRequest(url, data, params, headers, prompt, requestId, pathway);
+        return this.executeRequest(cortexRequest);
     }
 
     parseResponse(data) {

package/server/plugins/openAiImagePlugin.js

@@ -1,28 +1,32 @@
 // OpenAIImagePlugin.js
 import ModelPlugin from './modelPlugin.js';
 import axios from 'axios';
-import RequestDurationEstimator from '../../lib/requestDurationEstimator.js';
+import RequestMonitor from '../../lib/requestMonitor.js';
 import { publishRequestProgress } from '../../lib/redisSubscription.js';
 import logger from '../../lib/logger.js';
 
-const requestDurationEstimator = new RequestDurationEstimator(10);
+const requestDurationEstimator = new RequestMonitor(10);
 
 class OpenAIImagePlugin extends ModelPlugin {
-    constructor(config, pathway, modelName, model) {
-        super(config, pathway, modelName, model);
+    constructor(pathway, model) {
+        super(pathway, model);
     }
 
     // Implement the method to call the DALL-E API
-    async execute(text, parameters, _, pathwayResolver) {
-        const url = this.requestUrl(text);
-        const data = JSON.stringify({ prompt: text });
+    async execute(text, parameters, _, cortexRequest) {
+        const { pathwayResolver } = cortexRequest;
+        cortexRequest.data = JSON.stringify({ prompt: text });
 
         let id;
-        const { requestId, pathway } = pathwayResolver;
+        const { requestId } = pathwayResolver;
+
+        let callid;
 
         try {
-            requestDurationEstimator.startRequest(requestId);
-            id = (await this.executeRequest(url, data, {}, { ...this.model.headers }, {}, requestId, pathway))?.id;
+            callid = requestDurationEstimator.startCall();
+            await this.executeRequest(cortexRequest);
+            id = cortexRequest.requestId;
+
         } catch (error) {
             const errMsg = `Error generating image: ${error?.message || error}`;
             logger.error(errMsg);
@@ -30,14 +34,14 @@ class OpenAIImagePlugin extends ModelPlugin {
         }
 
         if (!parameters.async) {
-            return await this.getStatus(text, id, requestId);
+            return await this.getStatus(text, id, requestId, callid);
         }
         else {
-            this.getStatus(text, id, requestId);
+            this.getStatus(text, id, requestId, callid);
         }
     }
 
-    async getStatus(text, id, requestId) {
+    async getStatus(text, id, requestId, callid) {
         // get the post URL which is used to send the request
         const url = this.requestUrl(text);
 
@@ -51,7 +55,7 @@ class OpenAIImagePlugin extends ModelPlugin {
             const response = (await axios.get(statusUrl, { cache: false, headers: { ...this.model.headers } })).data;
             status = response.status;
             let progress = 
-                requestDurationEstimator.calculatePercentComplete();
+                requestDurationEstimator.calculatePercentComplete(callid);
 
             if (status === "succeeded") {
                 progress = 1;
@@ -66,7 +70,7 @@ class OpenAIImagePlugin extends ModelPlugin {
             });
 
             if (status === "succeeded") {
-                requestDurationEstimator.endRequest();
+                requestDurationEstimator.endCall(callid);
                 break;
             }
             // sleep for 5 seconds