# Stand-up Notes

## 25-8-2025
### Felix
* Completed first draft of the abstract
* HS genotyping; polishing founders vcf file:
* - sorting xsome order
* - remove duplicate markers

### Johannnes
* Extending the rag system, make it more of an agent:
* - fixing bugs
* - reduce runtime

### Bonface
* CD went down!
* revert it to last commit
* set it up locally (not very easy)
* coordinate with Fred and Aarun for help

## 27-8-2025
### Felix
* updating the abstract: meeting with Kauthar, for more tips and insight
* HS rats genotyping
* still no significant output

### Bonface
* fixed GN3, GN guile, and GN auth
* reproduce GN2 error, not yet fixed though

### Johannes
* LLM transformer is taking too long to run
* change input documentations to observe results

## 29-8-2025
### Felix
* Still on the abstract
* Errors with hs genotyping script

### Johannes
*  Still on the rag system: experimenting with documents as input

### Bonface
* CD down again, tux02 is the calprit
* Running gn2 outside container in meantime

## 1-9-2025
### Felix
* improving the smoothing scripts for hs: building a snakemake pipeline; redrafting the abstract introductory part

### Bonface
* looked on deployment to fix...,
* install gpu drivers for penguine 2 with alex
* it has gpu drivers tesla k8, but hardware installation is important for the meantime

### Alex
* rqlt2/lmdb runs and rebooted
* dumping bxd phenotype dataset for testing

### Johannes
* no big updates
* literature review on rag systems and how they work

## 3-9-2025
### Felix
* Improved the inferring script
* Abstract on hold first

### Johannes
* fix documents using agents: the issue is parallelization
* reviewing fahamu AI (the python part)

### Bonface
* Gn2 finally runs
* Documenting the progress and steps
* Gn-auth is broken; Fred to solve it
* Updated tesla k8 gpu drivers


### Alex
* Rqtl-wrapper for BXD is done
* focusing on cross-information;lmdb

## 5-9-2025
### Felix
* managed to generate haplotype blocks for hs rats

### Johannes
* RAG experimentation script: balg01, running slow; prepare for msc defense

### Alex
* dumping phenotypes to lmdb

### Bonface
* install drivers on tesla k8
* review Johannes scripts
* fixing gn2
* review alex's patches, {pull requests via emails}

* AOB;
* API tokens for RAG and coding

## 10-9-2025
### Felix
* Managed to run gemma and generate plots for the original hs genotype data and hs smoothed genotype data
* Still working on fine-tuning the statistical metrices to suite the plots

### Bonface
* All the GN machines are up and running, expect GN Auth
* Sent documentations to Pjotr on the progress with troubleshooting GN machines

### Alex
* writing queries to move phenotypes matrices from gn2 db to lmdb

### Johannes
* defense

## 17-9-2025
### Felix
* Completed abstract writing
* Working on generating genotype files for hs in gn2 format

### Bonface
* Gn machines are up and running
* Still some issues with manually restarting the container
* Worked with Alex to set up an external bootable ssd for Johannes

### Alex
* Succeeded dumping phenotypes to lmdb
* Been reading literature around phenotypes in gn2
* Assisted Bonz in setting up external bootable ssd for Johannes

### Johannes
* Managed to get access to bux01 server for the llms work

## 19-09-2025
### Felix
* Finalized HS genotypes/haplotypes

### Bonface
* GN machines are running on CD
* troubleshooting assisted by Aaron

### Johannes
* Experimenting with RAG systems

### Alex
* Writing LMDB scripts to dump phenotypes to LMDB from the common gn DB

## [Review] MONTH OCTOBER, 2025
*** Metric wise, plans and goals to achieve

### Bonface
Thoughts: Good start after relocation.

* CD/CI: Make sure that tests/infra is super stable
* Look at suggested forge/guix-bioinformatics upgrades that affect prod
* Support work for GNQA: sane gn-qna(fahamu)/AI deploys; Review Johannes' work; Help with tuning models; play catch up (cover up knowledge gap); figure out how to compare the different models.
* Follow up on my phd

### Alex
Thoughts: very productive

* Rqtl2 lmdb adapter running on production; depends on traction with Rqtl2 upstream by karl Browman to gn-bioinformatics
* Collaborate with Felix on his scripts

### Johannes

Thoughts: Not very productive, barely, looking forward for more improvement

* Experimenting on the qtl data with the AI system being currently built; also in need of optimization
* Comparing LLMs/GNQA agents; getting assistance from Shelby's work

### Felix
* HS genotypes to be fully supported in GN2
* Poster presentation
* PhD: ML objective (ML fundamentals); Complete 1st manuscript draft; Process my student pass