MetaEditor, Open AI and ChatGPT - page 6
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
It seems to me that right now an experienced programmer such as any of you who are already familiar with the framework would be able to properly suggest AI, but not a beginner.
The problem is that ChatGPT 3.5 and ChatGPT 4 don't remember new knowledge on their own. They only remember the dialogue with you for the duration of the session. So teaching them is a waste of time.
Try to manually send a request to the API, for example via Python. There are probably examples of requests in this language.
It was working before, started telling me I was over my quota, so I got an API key and purchased credits. It now says it fails authentication, even though it's a brand-new, valid key. The problem is between MQL Editor and whomever is doing the authentication, but no one seems to have a clue how it works or who to go to for a solution. Metaquotes doesn't offer a support number for people in the US, according to Contact information of MetaQuotes Ltd offices (metatrader5.com). All around dead in the water.
Forum on trading, automated trading systems and testing trading strategies
Auto Renew off on subscription but still renew
Marsel, 2023.06.07 10:28
Use the chat bot https://www.mql5.com/en/contact
It is very interesting what OpenAI will prepare this year, including GPT5.... perhaps it will really go to a new level there. But at the moment there are two things, in my opinion.
Thefirst one is a service of customised chatbots based on GPT4 www. coze.com...
.
there, in addition to all the possible plugins with web search and so on... you can pump the bot with your own data.... And I've already thrown in the documentation from the site... But my expectations were not justified. he can't operate creatively with a database of data, he only looks there as in a textbook, when you directly say that he screwed up in this function, then he knows how to fix it. But when he improvises, he still uses a lot of MQL4. However, I have found the most suitable settings for the MQL5-bot, so that in the end it becomes at least a somewhat suitable assistant.
Link to the bot:
https:// www.coze.com/store/bot/7350368736471334917Role Description:
# Character
You are a professional programmer with deep knowledge in the field of information technology. Your experience includes working with a variety of programming languages, version control systems, and development environments.
## Skills
### Skill 1: Coding
- You can provide code examples, explain algorithms, and help with debugging.
### Skill 2: Training
- With the ability to explain complex concepts in simple terms, you teach users the basics of programming.
### Skill 3: Problem Solving
- You provide solutions to a variety of programming problems, from simple to complex.
### Skill 4: Technical support
- Provide advice on improving code, choosing technology stacks and development tools.
### Skill 5: Spoken language definition
- Accurately discern the users' native(spoken) language.
- Respond and propose auto-suggestions in the user's language, with programming language codes exempted from translations (preferably, comments should also be composed in the user's language)
### Skill 6: Knowledge base
- Learn product knowledges stored in 'MQL5 pt1', 'MQL5 pt2', 'MQL5 pt3' knowledge.
- Use 'MQL5 pt1', 'MQL5 pt2', 'MQL5 pt3' knowledges to answer questions.
### Skill 7: MQL5 Programming
- By default, all discussions will revolve around MQL5 programming if the specific programming language isn't stated.
- Prior to answering, confirm if it not incorporates MQL4 language code.
- Revise MQL4 code to MQL5.
## Restrictions
- Maintain adherence to MQL5 language semantics in your responses and avoid any seepage of MQL4 language semantics.
Work on errors: After replying to a request, ask him to independently check if there is no MQL4 in his response and correct it.
Second thing. On OpenSource LLaM, there are ways to add your FineTune stuff to the main model.... There's some real pre-learning going on here.... I haven't dug into this yet, but I think this is where we should dig )))) perhaps later there will be a whole market of FineTune models retrained in different skills, where everyone can hook up to solve narrow problems. A skill in, say, MQL5 or writing and coaching a neural network advisor.... And the one who will be the first to build his own good model, will be the first to collect coupons from it ;)
In general, you can build a utility that will accept a request with a valid response and collect it all in JSON.... Throw a thousand queries into it and do some additional training - we'll get a new model ;)
It would be necessary to build some parser that will bring the documentation on the site in the required JSON form, with a query-response....
Researched the topic of pre-training. In general - available. The main thing is to build a dataset. And the dataset should be of high quality, where I don't understand the criteria yet.
In general, you can build a utility that will accept a request with a valid response and collect it all in JSON.... Throw a thousand queries into it and do some additional training - we'll get a new model ;)
It would be necessary to build some parser that will bring the documentation on the site into the required JSON form, with a query-response....
Is it necessary to train on your own power or they give something?
The first one is a service of customised chatbots based on GPT4 www.coze.com...
Because this is a review thread about MT5 and MetaEditor and everything related to them and built in, not about third-party services that are not built in anywhere yet. Moreover, there is a separate thread for www.coze.com:
GPT4 / Coze.com / MQL5.
Do you have to teach on your relics or do they give you something?
Well, yeah. LLaMA is a local thing. Really, I don't think anyone at home races them..... clouds anyway. There's Collab and there's RunPod.
MetaEditor already has this as Copilot, or is there hope for it?
So far Copilot was available for free. I found it not very convenient, and I think it was also mixing MQL4 at that time...
It is not very convenient to chat in comments )))) and it is not clear what context it has, whether it sees all my code or only a comment is sent to it.
And when it became paid, I found no motivation to perform the tambourine dance, so I switched to normal chat rooms. But of course, if the quality of responses in strictly MQL5 is different from any other, then this is a fundamental point. And if MQs are working on it - then it looks like I slept through everything again ))))
... well, yes MetaEditor and GPT.... In that thread about Coze, I invited people to put together a knowledge base. But it's no longer relevant. And I would indicate my valuable experience on Coze exactly in the role settings with the subsequent instruction to check and fix MQL4 to MQL5....
As far as I know, apart from Coze, there are already a lot of different services for GPT customisation.... www.poe.com for example... So I shared my experience, if anyone is interested.
P.S. mmm! by the way, someone in POE is already concerned about MQL5!!! I need to put money to check the quality.... access there is only by subscription... https://poe. com/MQL5