Added all the prompt contents in one file only

Author: preeesha

ai-assistant/src/core/prompt/contents/diagram.ts

@@ -1,18 +1,38 @@
-import { PROMPT_DIAGRAM_COMMAND } from "./contents/diagram";
-import { PROMPT_DOCUMENT_COMMAND } from "./contents/document";
-import { PROMPT_EXTRACT_DB_KEYWORDS } from "./contents/extractDBKeywords";
-import { PROMPT_STYLEGUIDE_COMMAND } from "./contents/styleguide";
-import { PROMPT_TRANSLATE_COMMAND } from "./contents/translate";
-import { PROMPT_WHY_USED_COMMAND } from "./contents/whyUsed";
 import { Prompt } from "./prompt";
 
 export namespace PromptFactory {
     export function makeDBKeywordQueryPrompt(query: string): Prompt {
         const prompt = new Prompt();
-        prompt.pushSystem(PROMPT_EXTRACT_DB_KEYWORDS);
-        prompt.pushUser(
-            `Hey I have this query, can you please extract the possible keywords from it? Please answer in <ANSWER_START>keyword1, keyword2<ANSWER_END> format only and don't say literally anything else.\n\nHere's my query:\n${query}`
-        );
+        prompt.pushSystem(`
+            You are an expert in understanding and answering questions of user.
+
+            ---
+            INPUT: User's text query in either natural language or code format.
+            ---
+            RULES:
+            1. Extract the possible keywords from the user's query.
+            2. If you are unable to extract the keywords, return an empty array.
+            3. Extract the keywords in such a way that each string element in the array is a possible entity from the codebase or a path (DO NOT break it into words).
+            4. STRICTLY, do not make anything other than the answer to the user's query.
+            ---
+            EXAMPLE:
+            1. INPUT: "Find the codebase for the user query CRC_TABLE in the main.ts"
+            OUTPUT: <ANSWER>CRC_TABLE, main.ts</ANSWER>
+            2. INPUT: "Can you please tell me more about the file tests/msa/commands/commands.spec.ts?"
+            OUTPUT: <ANSWER>tests/msa/commands/commands.spec.ts</ANSWER>
+            3. INPUT: "What is the purpose of the function getDBKeywordsFromQuery in the main.ts?"
+            OUTPUT: <ANSWER>getDBKeywordsFromQuery, main.ts</ANSWER>
+            4. INPUT: "Can you please tell me more about the file tests/commands.spec.ts?"
+            OUTPUT: <ANSWER>tests/commands.spec.ts</ANSWER>
+
+            OUTPUT FORMAT: <ANSWER>keyword1,keyword2,full/path/1,full/path/2</ANSWER>
+        `);
+        prompt.pushUser(`
+            Hey I have this query, can you please extract the possible keywords from it? Please answer in <ANSWER_START>keyword1, keyword2<ANSWER_END> format only and don't say literally anything else.
+            
+            Here's my query:
+            ${query}
+        `);
 
         return prompt;
     }
@@ -28,7 +48,14 @@ export namespace PromptFactory {
             5. If you are unable to answer the question, you must tell the user that you are unable to answer the question.
         `);
         prompt.pushUser(
-            `Hey I have a the following codebase in between the tags <CODEBASE_START> and <CODEBASE_END>. Can you please answer the following query?\n\n${query} \n\nHere's the codebase:\n<CODEBASE_START>\n${codebase}\n<CODEBASE_END>`
+            `Hey I have a the following codebase in between the tags <CODEBASE_START> and <CODEBASE_END>. Can you please answer the following query?
+            
+            ${query} 
+            
+            Here's the codebase:
+            <CODEBASE_START>
+            ${codebase}
+            <CODEBASE_END>`
         );
 
         return prompt;
@@ -37,7 +64,17 @@ export namespace PromptFactory {
     export function makeDiagramPrompt(codebase: string, query: string): Prompt {
         const prompt = new Prompt();
 
-        prompt.pushSystem(PROMPT_DIAGRAM_COMMAND);
+        prompt.pushSystem(`
+            You are an expert in understanding and answering questions of user when given a proper context of the codebase.
+            You provide mermaid 8.9.0 based graph and sequence diagrams which enhance the user's understanding of the codebase. These diagrams has a special quality that they are never off the mark and always render properly. The diagram are never other than the information provided in the codebase.
+
+            EXPECTED OUTPUT:
+            <ANSWER>
+            - You provide mermaid 8.9.0 based graph and sequence diagrams only.
+            - The aim is to make it easy for the user to understand the flow & overall working.
+            - The output must not have any kind of errors and must render properly.
+            </ANSWER>
+        `);
         prompt.pushAssistant(
             "Sure, I will strictly follow my instructions. I will provide the answer in a valid PLAIN TEXT only. I won't use parentheses at all even if they are required."
         );
@@ -61,33 +98,36 @@ export namespace PromptFactory {
         query: string
     ): Prompt {
         const prompt = new Prompt();
-        prompt.pushSystem(PROMPT_DOCUMENT_COMMAND);
-        prompt.pushUser(
-            `<CODEBASE_START>\n${codebase}\n<CODEBASE_END>\n\nTarget Entity: ${query}`
-        );
-
-        return prompt;
-    }
+        prompt.pushSystem(`
+            You are an expert in understanding and generating JSDoc documentation for other developers when given a proper context of the codebase.
 
-    export function makeWhyUsedPrompt(codebase: string, query: string): Prompt {
-        const prompt = new Prompt();
-        prompt.pushSystem(PROMPT_WHY_USED_COMMAND);
-        prompt.pushUser(
-            `Hey can you explain why this \`${query}\` entity is used in the following codebase? Here's the codebase:\n\n<CODEBASE_START>\n${codebase}\n<CODEBASE_END>`
-        );
+            INPUT: Inter-related entities from a huge codebase in JSON format, target entity to generate documentation for & number of example usages to provide.
 
-        return prompt;
-    }
+            EXPECTED OUTPUT:
+            <ANSWER_START>
+                <JSDOC_START>
+                    - Generate a short JSDoc documentation for the target entity explaining its purpose and usage.
+                    - Generate a comprehensive JSDoc documentation for the target entity explaining its purpose, usage, and parameters in @description, @param, @returns, @throws sections respectively.
+                    - (IF EXISTS) Explain the edge cases and exceptions the target entity might throw or face in the @throws section.
+                    - (ONLY IF POSSIBLE & RELEVANT) Provide different example usages of the target entity in the codebase.
+                <JSDOC_END>
+                <EXPLANATION_START>
+                    - Provide an additional comprehensive explanation of the target entity with proper reasoning.
+                <EXPLANATION_END>
+            <ANSWER_END>
 
-    export function makeStyleguidePrompt(
-        codebase: string,
-        styleguides: string
-    ): Prompt {
-        const prompt = new Prompt();
-        prompt.pushSystem(PROMPT_STYLEGUIDE_COMMAND);
-        prompt.pushUser(
-            `<STYLEGUIDES_START>${styleguides}<STYLEGUIDES_END>\n<CODEBASE_START>${codebase}<CODEBASE_END>`
-        );
+            RULES:
+            - STRICTLY, do not make anything other than the answer to the user's query.
+            - DON'T REPEAT THE EXAMPLES.
+            - Do not provide any kind of diagram or visualization in the output.
+        `);
+        prompt.pushUser(`
+            <CODEBASE_START>
+            ${codebase}
+            <CODEBASE_END>
+            
+            Target Entity: ${query}
+        `);
 
         return prompt;
     }
@@ -126,7 +166,13 @@ export namespace PromptFactory {
             - The output MUST BE IN ONLY AND ONLY MARKDOWN.
         `);
         prompt.pushUser(
-            `Hey, can you suggest multiple fixes for the target entity? To help you with the context I have provided the codebase of the entities it uses and the target entity. You don't need to worry about the codebase, just focus on the target entity.\n\n<CODEBASE_START>\n${codebase}\n<CODEBASE_END>\n<TARGET_ENTITY_START>${targetEntity}\n<TARGET_ENTITY_END>.`
+            `Hey, can you suggest multiple fixes for the target entity? To help you with the context I have provided the codebase of the entities it uses and the target entity. You don't need to worry about the codebase, just focus on the target entity.
+            
+            <CODEBASE_START>
+            ${codebase}
+            <CODEBASE_END>
+            <TARGET_ENTITY_START>${targetEntity}
+            <TARGET_ENTITY_END>.`
         );
 
         return prompt;
@@ -139,9 +185,70 @@ export namespace PromptFactory {
     ): Prompt {
         const prompt = new Prompt();
 
-        prompt.pushSystem(PROMPT_TRANSLATE_COMMAND);
+        prompt.pushSystem(`
+            You are an expert in understanding various programming languages and specialized in typescript and javascript.
+
+            INPUT: Inter-related entities from a huge codebase in JSON format, target entity to translate, target entity & the language to translate to.
+
+            TASK: Based on the context (codebase) (external entities it uses) provided, translate the target entity to the language provided by the user.
+
+            EXPECTED OUTPUT: code in the target language not in markdown format.
+
+            RULES:
+            - STRICTLY, do not make anything other than the answer to the user's query.
+            - DO NOT REPEAT THE TRANSLATION MULTIPLE TIMES.
+            - Do not provide any kind of diagram or visualization in the output.
+            - The output MUST BE IN ONLY AND ONLY STRING.
+        `);
+        prompt.pushUser(`
+            Hey, can you translate the following codebase in TypeScript to the ${targetLanguage}? I have provided you with other entities as well on which my target entity depends. Here you go:
+
+            <CODEBASE_START>
+            ${codebase}
+            <CODEBASE_END>
+
+            <TARGET_ENTITY_START>${targetEntity}
+            <TARGET_ENTITY_END>
+        `);
+
+        return prompt;
+    }
+
+    export function makeWhyUsedPrompt(codebase: string, query: string): Prompt {
+        const prompt = new Prompt();
+        prompt.pushSystem(`
+            You are an expert in understanding and answering questions of user when given a proper context of the codebase.
+
+            INPUT: User's text query
+
+            EXPECTED OUTPUT:
+            <ANSWER>
+            <EXPLANATION>
+                - Provide an additional comprehensive explanation in markdown list format.
+                - NEVER NEVER NEVER explain the entity itself or it's working.
+                - ALWAYS explain where and why it's used in the codebase with due reasoning and mention of the file and line number.
+            </EXPLANATION>
+            <DIAGRAM>
+                - You only provide flowchart or sequence diagram in the mermaid 8.9.0 format.
+                - The diagram must be clear and understandable for the user. The aim is to make it easy for the user to understand the flow & overall working.
+                - The output must not have any kind of errors and must render properly.
+            </DIAGRAM>
+            </ANSWER>
+
+            RULES:
+            - NEVER NEVER NEVER explain the entity itself or it's working.
+            - Don't tell me how to use that entity in the codebase.
+            - STRICTLY, do not make anything other than the answer to the user's query.
+            - If that entity is used multiple times then provide the reasoning for each usage separately.
+            - DON'T REPEAT THE USAGES OF THE ENTITY MULTIPLE TIMES.
+            - The output MUST BE IN ONLY AND ONLY IN THE ABOVE SPECIFIED FORMAT.
+        `);
         prompt.pushUser(
-            `Hey, can you translate the following codebase in TypeScript to the ${targetLanguage}? I have provided you with other entities as well on which my target entity depends. Here you go:\n\n<CODEBASE_START>\n${codebase}\n<CODEBASE_END>\n<TARGET_ENTITY_START>${targetEntity}\n<TARGET_ENTITY_END>`
+            `Hey can you explain why this \`${query}\` entity is used in the following codebase? Here's the codebase:
+            
+            <CODEBASE_START>
+            ${codebase}
+            <CODEBASE_END>`
         );
 
         return prompt;

ai-assistant/src/core/prompt/contents/document.ts

@@ -15,7 +15,8 @@ export class Prompt {
         return content
             .split("\n")
             .map((x) => x.trim())
-            .join("\n");
+            .join("\n")
+            .trim();
     }
 
     pushSystem(content: string) {