Why we might need it: New Features -Vector Store in File Search: Now supports up to 10,000 files, a 500x increase from the previous limit, handling chunking, embedding, and quality knowledge retrieval more efficiently. -Tool Choice Forcing: Allows specific tool enforcement per thread to optimize task execution. -Assistant Roll Messages: Enables the inclusion of assistant-originated messages in message histories, enhancing thread management. Improvements Over Version 1 -Enhanced Context Management: Message history tracking now includes options for automatically truncating or setting specific thresholds for the number of messages in context, addressing issues related to token cost inefficiency. -Streamlined SDK Support: Improved support across diverse programming environments like TypeScript and Python, simplifying integration and usage. Search Optimization -File Search Enhancements: Includes improved query handling capabilities with enhanced reranking and user search (query) rewriting for optimized file search. -Chunking and Overlap Strategy: Implements a refined chunking strategy with significant overlap (400 tokens overlap on 800 token chunks) to ensure comprehensive data retrieval without missing context. Multi-Hop Document Reasoning -Complex Query Handling: Breaks down complex queries into simpler, multiple queries that can be handled more effectively to synthesize and return comprehensive responses.

Search Techniques -Hybrid Search Approaches: Utilizes both keyword and semantic search techniques to capture the intent and specifics of queries, not just the literal terms. Re-Ranking Mechanism -Query and Chunk Re-Scoring: After initial retrieval, queries and document chunks are re-ranked and re-scored based on relevance to the query, ensuring that the most pertinent information is prioritized. Chunking Strategy -Limited Chunk Context: Limits the number of chunks in context to 20, optimizing performance without overwhelming the system with too much data. Benefits for Companies -Cost-Effectiveness and Efficiency: By addressing issues related to token cost and response latency, Assistants V2 becomes a more viable option for business applications, reducing operational costs and improving user experience. -Enhanced Customization and Control: Offers businesses more control over the tool's behavior and output, aligning more closely with specific use-case requirements. Cost and Speed Efficiency -Optimized Performance: Reduced latency and cost through more efficient use of tokens and enhanced system responsiveness. Knowledge Base and Control Features -Expanded Document Handling: From a limit of 20 documents to up to 10,000, coupled with advanced control features like max token settings for threads, providing better management of resource consumption and cost. Fine-Tuning Strategy -Use-Case Specificity: You can start using Assistants API v2 with the full capabilities of GPT-4 and then collect the comprehensive data across various threads. This data (all those threads) is then used to fine-tune the smaller GPT-3.5 model, resulting in much cheaper, much faster, and much more targeted model that excludes unnecessary domains (like coding or medical information if your use case is travelling & hotel business), optimizing performance by focusing only to specific needs.

Future Developments -Text Extraction from Images: Planned capabilities for extracting text from images to further enhance the utility and applicability of the API in diverse business scenarios.

@best-army-74344 🦸🏽‍♂️ 💎 🫡 First we need to study, then we need to practice, and then we can start building Botpress templates and real projects for our companies and clients.

You are not real.

How do you know all these things about AI

What school do you go to. Cuase enroll me in. I need this to get clients 🙂

Sorry if im annoying, but i gotta know.

Ill just build a GPT that increases my sentences

Last time we were working on the calcutor project, we were unfimiliar with Chatbots ( I think )

But now, we have increased our skills AND projects by a vast number

🤣 I love this GPT

sry

I look like him

OMW, HES SO FLUFFY

Did you see that at a part of that gif he has 4 arms

Look closely

Morningside AI (Liam Ottleys company) explaining the benefits and why we might need it

i know him lol

i am subbed

same with millions of others..

ps i tried dm'ing you but it would go through, what happened to your profile? Last message i got was the easter one!

Send me a DM

In those many messages, I shared all the instructions needed to use Assistants API v2.

Yes, I disabled DMs on this Discord account some time ago so I can concentrate more on work 🛠️ 🤖 🧑‍💻

man i thought you blocked me, i was heart broken for a second but then i thought he wouldn't do that to me, he love's me! haha!

here are the places v1 is mentioned in the original template @quick-musician-29561 made... without tool: const OPENAI_API_KEY = env.OPENAI_API_KEY; const assistantID = env.ASSISTANT_ID; const BASE_URL = 'https://api.openai.com/v1'; const headers = { 'Content-Type': 'application/json', 'Authorization':

Bearer ${OPENAI_API_KEY}

, 'OpenAI-Beta': 'assistants=v1', }; const retrieveRun = async () => { const statusResponse = await axios.request({ method: 'get', maxBodyLength: Infinity, url:

https://api.openai.com/v1/threads/${session.threadID}/runs/${runId}

, headers: { 'OpenAI-Beta': 'assistants=v1', Authorization:

Bearer ${OPENAI_API_KEY}

, 'Content-Type': 'application/json' } });

when simply changing all the mentions of v1 to v2 it gives this error: 23 May 2024, 11:03:18

[e81fd351c5]    [bot-action]    [flow:Main] [node:Add_Stuff_Here] [card:OpenAI Assistant: Create Thread, Send Messages, Retrieve Responses]:     Error creating thread: { error: 
   { type: 'invalid_request_error',
     code: 'unknown_url',
     message: 'Unknown request URL: POST /v2/threads. Please check the URL for typos, or see the docs at https://platform.openai.com/docs/api-reference/.',
     param: null } }
23 May 2024, 11:03:18
[e81fd351c5]    [bot-action]    [flow:Main] [node:Add_Stuff_Here] [card:OpenAI Assistant: Create Thread, Send Messages, Retrieve Responses]:     Error executing action "inline-card:OpenAI Assistant: Create Thread, Send Messages, Retrieve Responses.js" in flow:Main:node:Add_Stuff_Here
  [Error, Request failed with status code 404

Sorry about that, and you're right! 💯 I have mentioned that I'm reading DMs in another account, and it's in my Discord bio. It's just easier this way; I can focus 100% on building stuff when needed, and also fully concentrate on DMs when I log into that other account. https://cdn.discordapp.com/attachments/1232193993836855347/1243201431331410002/Screenshot_from_2024-05-23_16-54-43.png?ex=66509d0e&is=664f4b8e&hm=f5c8a66b66d0ff8db7d663ead7f87e7292e84156341404f985a320b9c94011e1&

I think @best-army-74344 has fully functional Assistants API v2 code 🔥

I'll update all my old chatbots and tutorials to use v2, for example this one which I already updated https://discord.com/channels/1108396290624213082/1180023587646746715/1244193401214210072

🦸‍♂️ 🛠️ 🫡

I'd like to use it for images as well, but not sure about how to display them correctly