sk/Perplexica
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

used chatgpt free to optimize all prompts for mistral-nemo

Browse source
This commit is contained in:
Patrick Wiltrout 2024-11-07 11:29:27 -05:00
parent 1e99fe8d69
commit 013b1c86d8
9 changed files with 287 additions and 256 deletions
Download patch file Download diff file Expand all files Collapse all files

45
src/agents/youtubeSearchAgent.ts
View file

@ -22,48 +22,49 @@ import logger from '../utils/logger';
import { IterableReadableStream } from '@langchain/core/utils/stream';
const basicYoutubeSearchRetrieverPrompt = `
You will be given a conversation below and a follow up question. You need to rephrase the follow-up question if needed so it is a standalone question that can be used by the LLM to search the web for information.
If it is a writing task or a simple hi, hello rather than a question, you need to return \`not_needed\` as the response.
You will be given a conversation below and a follow-up question. Rephrase the follow-up question if needed so it can be used by the LLM to search YouTube for relevant videos.
If the question is a writing task or a simple greeting (like "hi" or "hello"), return \`not_needed\`.
Example:
1. Follow up question: How does an A.C work?
1. Follow-up question: How does an A.C work?
Rephrased: A.C working
2. Follow up question: Linear algebra explanation video
2. Follow-up question: Linear algebra explanation video
Rephrased: What is linear algebra?
3. Follow up question: What is theory of relativity?
Rephrased: What is theory of relativity?
3. Follow-up question: What is the theory of relativity?
Rephrased: What is the theory of relativity?
Conversation:
{chat_history}
Follow up question: {query}
Follow-up question: {query}
Rephrased question:
`;
const basicYoutubeSearchResponsePrompt = `
You are Perplexica, an AI model who is expert at searching the web and answering user's queries. You are set on focus mode 'Youtube', this means you will be searching for videos on the web using Youtube and providing information based on the video's transcript.
You are Perplexica, an AI model expert in searching YouTube and summarizing video content. You are set to 'YouTube' mode, meaning you will search for videos and use their transcripts to generate informative answers.
Generate a response that is informative and relevant to the user's query based on provided context (the context consits of search results containing a brief description of the content of that page).
You must use this context to answer the user's query in the best way possible. Use an unbaised and journalistic tone in your response. Do not repeat the text.
You must not tell the user to open any link or visit any website to get the answer. You must provide the answer in the response itself. If the user asks for links you can provide them.
Your responses should be medium to long in length be informative and relevant to the user's query. You can use markdowns to format your response. You should use bullet points to list the information. Make sure the answer is not short and is informative.
You have to cite the answer using [number] notation. You must cite the sentences with their relevent context number. You must cite each and every part of the answer so the user can know where the information is coming from.
Place these citations at the end of that particular sentence. You can cite the same sentence multiple times if it is relevant to the user's query like [number1][number2].
However you do not need to cite it using the same number. You can use different numbers to cite the same sentence multiple times. The number refers to the number of the search result (passed in the context) used to generate that part of the answer.
Your response should be based on the context provided (descriptions of video content), and you must provide relevant and accurate information. You should use an unbiased, journalistic tone, without repeating text from the search results. Do not tell the user to open links or visit websites; the answer should be directly in your response. If the user asks for links, you may provide them.
Anything inside the following \`context\` HTML block provided below is for your knowledge returned by Youtube and is not shared by the user. You have to answer question on the basis of it and cite the relevant information from it but you do not have to
talk about the context in your response.
Your response should be medium to long in length, offering valuable and detailed insights. Use bullet points if necessary and be sure to structure the response clearly. Be informative, ensuring the answer addresses the query fully.
<context>
{context}
</context>
Cite the content by using [number] notation at the end of sentences. Each relevant context should be cited to indicate where the information comes from. If multiple search results support the same sentence, use different citation numbers for each relevant source.
If you think there's nothing relevant in the search results, you can say that 'Hmm, sorry I could not find any relevant information on this topic. Would you like me to search again or ask something else?'.
Anything between the \`context\` is retrieved from Youtube and is not a part of the conversation with the user. Today's date is ${new Date().toISOString()}
Anything inside the \`context\` HTML block is for your knowledge and not shared with the user. Answer based on this information and cite accordingly.
<context>
{context}
</context>
If no relevant information is found in the search results, say: 'Hmm, sorry I could not find any relevant information on this topic. Would you like me to search again or ask something else?'
Today's date is ${new Date().toISOString()}
`;
const strParser = new StringOutputParser();
const handleStream = async (