@aj-archipelago/cortex 1.0.0 → 1.0.1

package/LICENSE

@@ -1,10 +1,6 @@
 MIT License
 
-<<<<<<< HEAD
-Copyright (c) 2022 Al Jazeera Media Network
-=======
-Copyright (c) 2023 aj-archipelago
->>>>>>> b78a6667176ebfd09d2c7ed1549d8dc18691c344
+Copyright (c) 2023 Al Jazeera Media Network
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

package/README.md

@@ -305,7 +305,7 @@ export default {
 
 ### Building and Loading Pathways
 
-Pathways are loaded from modules in the `pathways` directory. The pathways are built and loaded to the `config` object using the `buildPathways` function. The `buildPathways` function loads the base pathway, the core pathways, and any custom pathways. It then creates a new object that contains all the pathways and adds it to the pathways property of the config object. The order of loading means that custom pathways will always override any core pathways that Cortext provides. While pathways are designed to be self-contained, you can override some pathway properties - including whether they're even available at all - in the `pathways` section of the config file.
+Pathways are loaded from modules in the `pathways` directory. The pathways are built and loaded to the `config` object using the `buildPathways` function. The `buildPathways` function loads the base pathway, the core pathways, and any custom pathways. It then creates a new object that contains all the pathways and adds it to the pathways property of the config object. The order of loading means that custom pathways will always override any core pathways that Cortex provides. While pathways are designed to be self-contained, you can override some pathway properties - including whether they're even available at all - in the `pathways` section of the config file.
 
 ## Core (Default) Pathways
 
@@ -435,7 +435,7 @@ If you encounter any issues while using Cortex, there are a few things you can d
 If you would like to contribute to Cortex, there are two ways to do so. You can submit issues to the Cortex GitHub repository or submit pull requests with your proposed changes.
 
 ## License
-Cortex is released under the MIT License. See [LICENSE](https://github.com/ALJAZEERAPLUS/cortex/blob/main/LICENSE) for more details.
+Cortex is released under the MIT License. See [LICENSE](https://github.com/aj-archipelago/cortex/blob/main/LICENSE) for more details.
 
 ## API Reference
 Detailed documentation on Cortex's API can be found in the /graphql endpoint of your project. Examples of queries and responses can also be found in the Cortex documentation, along with tips for getting the most out of Cortex.

package/SECURITY.md

@@ -0,0 +1,22 @@
+# Security Policy
+
+## Supported Versions
+
+We take the security of our project seriously. The table below shows the versions of Cortex currently being supported with security updates.
+
+| Version | Supported          |
+| ------- | ------------------ |
+| 1.x.x   | :white_check_mark: |
+
+## Reporting a Vulnerability
+
+If you have discovered a security vulnerability in Cortex, please follow these steps to report it:
+
+1. **Do not** create a public GitHub issue, as this might expose the vulnerability to others.
+2. Please follow the GitHub process for [Privately Reporting a Security Vulnerability](https://docs.github.com/en/code-security/security-advisories/guidance-on-reporting-and-writing/privately-reporting-a-security-vulnerability)
+
+## Disclosure Policy
+
+Cortex follows responsible disclosure practices. Once a vulnerability is confirmed and a fix is developed, we will release a security update and publicly disclose the vulnerability. We will credit the reporter of the vulnerability in the disclosure, unless the reporter wishes to remain anonymous.
+
+We appreciate your help in keeping Cortex secure and your responsible disclosure of any security vulnerabilities you discover.

package/graphql/pathwayResolver.js

@@ -151,7 +151,7 @@ class PathwayResolver {
         }
         const encoded = encode(text);
         if (!this.useInputChunking || encoded.length <= chunkTokenLength) { // no chunking, return as is
-            if (encoded.length >= chunkTokenLength) {
+            if (encoded.length > 0 && encoded.length >= chunkTokenLength) {
                 const warnText = `Truncating long input text. Text length: ${text.length}`;
                 this.warnings.push(warnText);
                 console.warn(warnText);
@@ -189,7 +189,7 @@ class PathwayResolver {
         // the token ratio is the ratio of the total prompt to the result text - both have to be included
         // in computing the max token length
         const promptRatio = this.pathwayPrompter.plugin.getPromptTokenRatio();
-        let chunkMaxTokenLength = promptRatio * this.pathwayPrompter.plugin.getModelMaxTokenLength() - maxPromptTokenLength;
+        let chunkMaxTokenLength = promptRatio * this.pathwayPrompter.plugin.getModelMaxTokenLength() - maxPromptTokenLength - 1;
         
         // if we have to deal with prompts that have both text input
         // and previous result, we need to split the maxChunkToken in half

package/graphql/plugins/localModelPlugin.js

@@ -0,0 +1,23 @@
+// localModelPlugin.js
+import ModelPlugin from './modelPlugin.js';
+import { execFileSync } from 'child_process';
+
+class LocalModelPlugin extends ModelPlugin {
+    constructor(config, pathway) {
+        super(config, pathway);
+    }
+
+    async execute(text, parameters, prompt, pathwayResolver) {
+        const { modelPromptText } = this.getCompiledPrompt(text, parameters, prompt);
+
+        try {
+            const result = execFileSync(executablePath, [text], { encoding: 'utf8' });
+            return result;
+        } catch (error) {
+            console.error('Error running local model:', error);
+            throw error;
+        }
+    }
+}
+
+export default LocalModelPlugin;

package/graphql/plugins/openAiChatPlugin.js

@@ -12,14 +12,14 @@ class OpenAIChatPlugin extends ModelPlugin {
         const { stream } = parameters;
     
         // Define the model's max token length
-        const modelMaxTokenLength = this.getModelMaxTokenLength() * this.getPromptTokenRatio();
+        const modelTargetTokenLength = this.getModelMaxTokenLength() * this.getPromptTokenRatio();
     
         let requestMessages = modelPromptMessages || [{ "role": "user", "content": modelPromptText }];
     
         // Check if the token length exceeds the model's max token length
-        if (tokenLength > modelMaxTokenLength) {
+        if (tokenLength > modelTargetTokenLength) {
             // Remove older messages until the token length is within the model's limit
-            requestMessages = this.truncateMessagesToTargetLength(requestMessages, modelMaxTokenLength);
+            requestMessages = this.truncateMessagesToTargetLength(requestMessages, modelTargetTokenLength);
         }
     
         const requestParameters = {

package/graphql/plugins/openAiCompletionPlugin.js

@@ -13,19 +13,22 @@ class OpenAICompletionPlugin extends ModelPlugin {
         let { modelPromptMessages, modelPromptText, tokenLength } = this.getCompiledPrompt(text, parameters, prompt);
         const { stream } = parameters;
         let modelPromptMessagesML = '';
-        const modelMaxTokenLength = this.getModelMaxTokenLength();
+        // Define the model's max token length
+        const modelTargetTokenLength = this.getModelMaxTokenLength() * this.getPromptTokenRatio();
         let requestParameters = {};
     
         if (modelPromptMessages) {
-            const requestMessages = this.truncateMessagesToTargetLength(modelPromptMessages, modelMaxTokenLength - 1);
+            const minMsg = [{ role: "system", content: "" }];
+            const addAssistantTokens = encode(this.messagesToChatML(minMsg, true).replace(this.messagesToChatML(minMsg, false), '')).length;
+            const requestMessages = this.truncateMessagesToTargetLength(modelPromptMessages, (modelTargetTokenLength - addAssistantTokens));
             modelPromptMessagesML = this.messagesToChatML(requestMessages);
             tokenLength = encode(modelPromptMessagesML).length;
         
-            if (tokenLength >= modelMaxTokenLength) {
-                throw new Error(`The maximum number of tokens for this model is ${modelMaxTokenLength}. Please reduce the number of messages in the prompt.`);
+            if (tokenLength > modelTargetTokenLength) {
+                throw new Error(`Input is too long at ${tokenLength} tokens (this target token length for this pathway is ${modelTargetTokenLength} tokens because the response is expected to take up the rest of the model's max tokens (${this.getModelMaxTokenLength()}). You must reduce the size of the prompt to continue.`);
             }
         
-            const max_tokens = modelMaxTokenLength - tokenLength - 1;
+            const max_tokens = this.getModelMaxTokenLength() - tokenLength;
         
             requestParameters = {
                 prompt: modelPromptMessagesML,
@@ -38,11 +41,11 @@ class OpenAICompletionPlugin extends ModelPlugin {
                 stream
             };
         } else {
-            if (tokenLength >= modelMaxTokenLength) {
-                throw new Error(`The maximum number of tokens for this model is ${modelMaxTokenLength}. Please reduce the length of the prompt.`);
+            if (tokenLength > modelTargetTokenLength) {
+                throw new Error(`Input is too long at ${tokenLength} tokens. The target token length for this pathway is ${modelTargetTokenLength} tokens because the response is expected to take up the rest of the ${this.getModelMaxTokenLength()} tokens that the model can handle. You must reduce the size of the prompt to continue.`);
             }
         
-            const max_tokens = modelMaxTokenLength - tokenLength - 1;
+            const max_tokens = this.getModelMaxTokenLength() - tokenLength;
         
             requestParameters = {
                 prompt: modelPromptText,

package/graphql/plugins/openAiWhisperPlugin.js

@@ -1,4 +1,4 @@
-// OpenAICompletionPlugin.js
+// openAiWhisperPlugin.js
 import ModelPlugin from './modelPlugin.js';
 
 import FormData from 'form-data';

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@aj-archipelago/cortex",
-  "version": "1.0.0",
+  "version": "1.0.1",
   "description": "Cortex is a GraphQL API for AI. It provides a simple, extensible interface for using AI services from OpenAI, Azure and others.",
   "repository": {
     "type": "git",