@James Scott has joined the channel
@Chris Krecicki has joined the channel
@Nicholas McFadden has joined the channel
Hey everyone, should have access to our claude api service, should be in your emails, let me know!
@James Scott what is our rate limiting & cost per hour of transcription. Can you please login and see this.
We need to know how many concurrent api requests we can send @Joe Santana needs this info -- need to know our plan
Rate Limits To ensure a smooth experience for all users, certain operations have per-account usage limits.
ModelUsage limitDescriptionSpeech-to-Text: Pre-recorded audio200 concurrent transcriptionsThe maximum number of API requests you can make simultaneously. If you exceed your usage limit, any API requests that exceed this limit will be added to a queue. Queued transcriptions will be processed automatically as previously submitted transcriptions complete. Speech-to-Text: Universal StreamingDynamic scaling, no limit. The maximum number of additional streams you can open each minute with a 10% ramp per minute. LeMURDefault:** 30 requests per minute
**Your actual LeMUR usage limit may differ from this default value. The default number of LeMUR requests you can make per minute.
If you exceed the limit, the API request will fail with a 429 status code. For more information, see Rate limits.
Nice @Joe Santana @Ryan -- Rate Limits Assmebly.ai Transcription: • Speech-to-Text (Pre-recorded): 200 concurrent transcriptions with automatic queuing for overflow • LeMUR: 30 requests per minute (account-specific), fails with 429 error when exceeded Your Usage Summary: • Total spend: $42.49 • Total hours: 157.38 Cost per hour: $0.27
@Ryan @Joe Santana -- we also had a bug with the transcription app regarding DB permissions on routes. We solved it this morning and I went to Luis, Jackson, Lexi and Nyla -- I will be showing John, Brooke, Ghannon and a few others too that needed it. We are going to save thousands of man hours from listing to calls. Huge huge money and time savings just commenced.
@Ryan @Joe Santana -- we also had a bug with the transcription app regarding DB permissions on routes. We solved it this morning and I went to Luis, Jackson, Lexi and Nyla -- I will be showing John, Brooke, Ghannon and a few others too that needed it. We are going to save thousands of man hours from listing to calls. Huge huge money and time savings just commenced.
@Chris Krecicki @James Scott @Nicholas McFadden, see you on todays call. Today's agenda is looking Tortelligence AI in Sanic ready to ship. I have a friend in insurance sales that asked me about two drugs. @James Scott, be sure we all three have a user setup so everyone can play with it before the meeting.
I have to update the back end data that’s proving to take a while @Ryan but the front facing is looking good and ready
How much time do you need, we can push meeting if necessary. I'd rather have it ready and humming first so what we see is what Marc & Cam & Andy see.
A couple days at most have to run the scripts and make sure all the data is present that’s about it but other than that I’m done still in testing to put the final bells and whistles but other than that we are good
I have my tort news scanner in sanic already. I believe James does too. I have to extend the bas ei his files and I need to merge his views and htl files assuming he also setup blueprints in main
Second things. Dustin says we can allow outsiders into our sanic application and we'd just give them DB permissions to use the tort intelligence tab
Alternative-- you have this toerintelligence domain, are we going to load up an entirely new docker container and point the domian to it
Or are you going to just point the doain to the current sanic appcation located at shieldlegal.dev
So when they visit it'll show toerintelligence.ai as the domain even know were laoding the same site
Also, Dustin hasnt had time, he's wildly busy. But we need to have him run the migration scripts I made to the production DB.
Also, is James using google secrets for his Zillis login info. Does he use any outside DB other than zillis?
Also, is James using google secrets for his Zillis login info. Does he use any outside DB other than zillis?
I had AI summerize what I just said: Current Status: • Both you and James have tort news scanners set up in Sanic • You need to merge work by extending base files and combining James's views/HTML files (assuming he used blueprints) Access & Domain Questions: • Dustin confirmed outsiders can access the Sanic app with DB permissions for the tort intelligence tab • Need to decide between two options for tortintelligence.ai domain: a. Create new Docker container for the domain b. Point domain to existing Sanic app at shieldlegal.dev (same site, different domain display) Pending Tasks: • Dustin needs to run your migration scripts on production DB (he's currently very busy) • Need clarification on James's setup: Is he using Google Secrets for Zillis login? Does he use any external DBs besides Zillis? Main Decision Needed: Whether to create separate infrastructure for the new domain or route it through the existing application.
Were in infra mode right now and merge mode
Start in SL Dev, once Cam approves for commercial beta, use branded domain.
Probably have a website built so we’d use client name (or app) dot Tortelligence.ai
@Ryan ok cool we'll figure out the landing page stuff downline and point a subdomain to sl dev so we can use our abckend still and have the LP on the branded domain
So it seems to be I need to merge what I've got with waht James has. James do you have a repo to your work I can look at?
Does he use google secrets for zillis?
Dustin and Tyson are going to run a migration script to setup the production database here soon too
If you do talk to James before I do tell him to get my his repo link and add me as a user or just zip it up and send it here so I can integrate it and come back with whatever else I need to merge it. I have Dustin/Tyson working on the DB migration from my test DB. Same goes for his Zillis info for logins etc so I can set it up on google secrets. [11:58 AM] Those are the next steps
@Chris Krecicki i can give u the repo link shortly
i will spend tomorrow merging it all -- also your zillis i need to add that info to google secrets and update your code to pull api keys etc from there
if you are interested in peft lora on moe
@Chris Krecicki you might need access let me know but the folder is medilens https://github.com/JamesShield/tortelligence_ai
@James Scott yeah i dont have access to this repo -- pleaase get me access some time today?
I should have this completed for our meeting tomorrow if I can get it today
@Ryan how do I get access to his repo? who is in charge of it?
@Chris Krecicki whats your user and email
cody is my legal name -- i just go by chris in business settings
@Chris Krecicki i invited this email cody@krecicki.com let me know when you get it!
@James Scott boom, done and done thank you
@Chris Krecicki the file is called medilens
going to udpate it to work with google secrets
let me send u the .env file i didnt use secrets
np i am focused on egtting this doen today but give me an hour or two to review how you set this up
any other permissions you need to give? I cannot clone the repo? @James Scott
@James Scott hah this will take a bit 🤣
@Ryan sometimes we see news related to mass sexual abuse outside the US like in France and Spain. Should I figure out a way to exclude these articles if I can or do we do business outside the US? I remember Cameron was in France.
Leave them in for now until we get Cam to use it with some feedback
@Ryan lol im excited to show him -- hope he likes it so far .. give me the next 2 days to get this integrated and tested and on the other side @Joe Santana is getting dustin/tyson/nick (one of them) to get the DB tables live in prod so I can switch out my test db info and we can push it to github and then I will test in prod just to be 100% its all working .. shooting for monday/tuesday assuming we can get these tables migrated
Again thanks for the job @Ryan i am really enjoying it a lot
@Ryan @James Scott I wanted to provide an update on the major integration project completed today.
Project Scope & Deliverables: Successfully integrated the external codebase into our main Sanic application, requiring comprehensive architectural refactoring to align with our existing systems and best practices.
Key Technical Accomplishments: • Restructured application blueprints, routing architecture, and API initialization framework • Migrated configuration management from local environment files to Google Secrets Manager for enhanced security compliance • Implemented conversation state management across multiple chatbot instances with in-memory persistence • Conducted full-stack refactoring including frontend optimization and modular view separation • Updated UI components and CSS styling to integrate with our existing light/dark mode theming system • Reorganized project file structure from a single disorganized folder into proper modular architecture, removing dead files and unused dependencies • Integrated BigQuery with asynchronous processing to maintain application performance and worker stability it was crashing the application workers due to BigQuery load time
Code Organization & Optimization: Restructured the project file architecture for better maintainability and implemented proper separation of concerns. Updated the overall codebase by consolidating files and removing unused components to optimize application performance.
Project Timeline: Dedicated focused development effort from 7:00 AM to 5:45 PM to ensure timely delivery and system stability.
The integration is now complete and fully operational within our development environment. All systems have been tested and verified for performance and reliability locally. The team (Dustin, Tyson or Nick) is going to get the database tables live soon and we can take the whole site live once we test the connection and get Cameron's feedback. See you guys tomorrow for the demo. Thanks James for getting me this at 7am. I am off for the night.
*Thread Reply:* @Chris Krecicki , that’s great work Chris! Thank you so much. Cannot wait to get access. Once we are ready how will we administer users? Want to have money setup for testing and trial.
*Thread Reply:* @Ryan Tyson & Dustin are the keepers of access to anything and everything. So we'll be able to have them add new users and give permissions. That's out of my realm of access. But we did talk about it yesterday. I will review options when we have our meeting today and go over what we talked about.
*Thread Reply:* @Ryan and you're welcome ❤️
*Thread Reply:* We've had a conversation about just cloning our current installation and stripping it of the call recording tools and other things. And we can just clone the repo and the docker container, but that was still being discussed.
*Thread Reply:* That is a conversation for Nick and Dustin because they have ownership over those things
*Thread Reply:* @Ryan Also I have Zero authority over Dustin, NIck & Tyson. So I have to have you push them to finish the database migrations and udpates for: the tort finder and the mass transcription projects. And that is my bottle neck right now is waiting on those two things. For now I think we should just launch it on our current docker container and once you all start selling it and we start getting more users we can split it off.
*Thread Reply:* But I need the DB stuff done or I cannot have it updated on the current sanic app in production because it'll break without the DB setup in prod.
*Thread Reply:* Ok. Let me know when they said they could get it done for us
*Thread Reply:* @Ryan no time frame, only time frame was, when they get time
*Thread Reply:* They've just got a knife to their neck as far as time goes with the integrations things .. from what I heard Sigitch (dicellos integrations team) thought we had 3-4x as many people working on integrations. So I understand why they've been saying "when they get time".
Amazing man!!! I am updating some back end tables shouldn’t interfere with what you going on
@James Scott shouldnt be an issue as long as we dont need to add anything new and its just updating the tables you've already got
The chat bots all have memory now too of the last 20 conversation pairs, we use local memory for this too. So theyre smart now.
They've just got a knife to their neck as far as time goes with the integrations things .. from what I heard Sigitch (dicellos integrations team) thought we had 3-4x as many people working on integrations. So I understand why they've been saying "when they get time".
@Ryan are we having that meeting today or no
@Dustin Surwill has joined the channel
@Ryan You're right. We need some paid news sources we can scrape more real time data from. I'm going to keep working on this scraper using news and search and specific sites like law.com but if we had access to real-time and better sources we could really get good data and this would 10x be better
@James Scott Here is my branch, clone this and only make css changes for now, if you have to do anything technical or need to add new files, just reach out so I don't have to figure out how to integrate it later https://github.com/shield-legal/internal-tools-site/tree/tort-control-panel-new-feature -- you should just send me the jinja or the views files here or via email so you don't have to worry about pushing a branch -- I can just add it manually (the updates) and make sure it works and I can push the branch myself.
@James Scott install uv package manager, run uv init in the cloned repo, then type uv sync, then uv run main.py -- then just wait for it to load everything
@James Scott i do not recommend using pip
@James Scott Also make an effort to make the changes in the ::root for light and dark mode and use var() in your CSS based on --whatever-you-name-it
@James Scott install uv package manager, run uv init in the cloned repo, then type uv sync, then uv run main.py -- then just wait for it to load everything
@James Scott i do not recommend using pip
@James Scott Also make an effort to make the changes in the ::root for light and dark mode and use var() in your CSS based on --whatever-you-name-it
Also @James Scott your Litigation Recommendations -- with some prompt work and some guidelines you could make this at least make stuff up instead of saying "N/A, because this abstract...." -- change the model to a smarter model too than 3.5-turbo. Try the 4.1o-micro
Also @James Scott your Litigation Recommendations -- with some prompt work and some guidelines you could make this at least make stuff up instead of saying "N/A, because this abstract...." -- change the model to a smarter model too than 3.5-turbo.
Also @James Scott your Litigation Recommendations -- with some prompt work and some guidelines you could make this at least make stuff up instead of saying "N/A, because this abstract...." -- change the model to a smarter model too than 3.5-turbo. Try the 4.1o-micro
@Dustin Surwill, meeting with Marc & Cameron on Tuesday with Chris and James, Joe and we need that newtortopportunities for running migration scripts for the Internal Tools. Can you have that ready by EOD Monday, for our Tuesday afternoon meeting.
And @Tyson Green, I need ryan@shield-legal.com added as a User to the Internal Tools.
*Thread Reply:* You need to log in to the website at least once before we can give you access: https://shieldlegal.dev/
You need to log in to the website at least once before we can give you access: https://shieldlegal.dev/
@Ryan @James Scott check this approach out ... local news + our keywords below .. seeing things before anyone else does after syndication. Also now without anymore noise from other law firm marketing, guides, surveys, opinions, Mormon and other torts people are already all over, and more noise, etc. Cam/Rick will see things immediately (assuming we run this daily). See the output on the newtortopportunities database, see the csv. It works absolutely incredible now. Very excited to get this live.
self.search_queries = [
"teacher sexual abuse 2025",
"priest sexual abuse 2025",
"coach sexual abuse 2025",
"doctor sexual abuse 2025",
"covered up sexual abuse 2025",
"failed to report sexual abuse 2025",
"multiple sexual abuse victims 2025",
"widespread sexual abuse 2025",
"pattern of sexual abuse 2025"
]
@Ryan @Chris Krecicki i just updated the code with chris added some key fundamentally to it, and i have the data in a way i am ready to have the end users test it, would provide valuable feedback as well as user testing on how they work and flow with it
Get me a user abd I’ll play with it in our AI meeting tomorrow all meeting. I’ve pinged Dustin on user.
@James Scott @Ryan OK team, we're ready for our demo tomorrow. James, once this goes live. You can pull the main internal-tools application and you wont have to setup a test db, @Dustin Surwill is getting the migration stuff taken care of. You'll never have to build in a stand alone environment too.
@Ryan @Chris Krecicki good with me
@Ryan let me know when @Cameron Rentch wants to do the demo, I am ready to go. I saw it's pretty late today.
@James Scott i just realized you've got a CSS issue on these table. I am sorting out what's wrong. The data is there but the CSS is pushing it off. Other than that everything works.
.scrollable-table {
max-height: 600px;
overflow: auto;
border: 1px solid #dee2e6;
position: relative;
}
but the table is super super slow loading when you sort etc, is this from bigquery? @James Scott
but the table is super super slow loading when you sort etc, is this from bigquery?
but the table is super super slow loading when you sort etc, is this from bigquery? @James Scott
Nice I fixed it -- also updated
.table-header {
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
color: white;
padding: 20px;
margin: 0;
}
Yes I believe it is when I do infinite scroll
@James Scott yeah i have to sort out the scroll .. give me a few -- im getting it fixed 🧙
hell yeah I fixed it
overflow-y: auto; overflow-x: hidden;
i split the overflow up so it didnt get covered with a white box but still scrolls
Thanks guys. Each of you can take a test run in our meeting for demo. And keep it short, sweet and prepare to get users setup
@Ryan I've got two meetings that look related Shield Legal - AI Project Sprint and our AI Demo | Drug & Abuse
@Ryan @James Scott I vote for a team meeting in DR and we get a catamaran for the day and do some corporate snorkeling mastermind
Were obviously in the wrong business .. these rent for $20-50k per week https://www.yachtcharterfleet.com/luxury-charter-yacht-62683/e-supercat.htm
lol, nice boat. James?
Use the first meeting to practice demo, focus on UI and a single example that has no sudden dead end. And don’t show them any code on screen 😆
@Ryan i was going to run the scraper live to show off the code LOL
We lose them on any code views. Just UI and digging into tables and using AI assistant etc.
@James Scott which drug and which clinic trial are we using
i thought that was the purpose to go over that lol
@James Scott where you at? i am in the meeting
@James Scott https://superlinked.com/
https://strandsagents.com/latest/
take a look at examples but it makes our agents and everything so much easier
ah thats cool they've built all the infra for you
looks like you MUST use bedrock/aws for backend?
Ah it has Ollama support https://strandsagents.com/latest/documentation/docs/user-guide/concepts/model-providers/ollama/#basic-usage
@Ryan @James Scott -- i just saw the meeting was changed till tomorrow at 2pm -- which is great, I have to go to the school and make sure the kids get on the bus by 3pm today
talk to you all tomorrow -- i am going to leave around 2pm
@James Scott im literally going to leave my computer running with what we planned out so it just works tomorrow LOL
@Chris Krecicki we should be good lol its works
@Ryan @Chris Krecicki any need to go over our demo again?