jnana

This is the spiritual successor of StructBioReasoner, wherein we have condensed the execution details into a set of skills that are empowered with the hypothesis tournament framework of Jnana and the hierarchical memory system of Pando.

Run with Loam

This repository is run through Hermes with the checked-in Loam lifecycle plugin. Install the hermes CLI, configure its model provider and credentials, then launch from the repository root:

.loam/run-hermes

The wrapper changes to the repository root, enables project-plugin discovery, activates Loam, and uses the isolated Hermes home at .hermes/.loam-home. It enables the loam-hooks plugin and invokes the pre-LLM hook every five turns by default.

Set LOAM_PRE_LLM_EVERY to change that interval; for example, run it every three turns:

LOAM_PRE_LLM_EVERY=3 .loam/run-hermes

The hook integration and its runtime state are under .loam/. See .loam/README.md for lifecycle-hook behavior and optional global Hermes installation.

Run the GC/codon survey

The standalone survey downloads 12 public NCBI nucleotide records, computes genomic GC, GC3, and synonymous codon usage, then writes derived artifacts to the selected directory. It requires Python 3 with Biopython, NumPy, SciPy, and Matplotlib:

python3 -m pip install biopython numpy scipy matplotlib
python3 experiments/codon_usage_gc_survey.py \
  --out-dir .loam/state/codon-usage-gc-survey

The output directory receives genome_summary.csv, codon_usage.csv, gc_vs_gc3.png, statistics.json, provenance.json, and REPORT.md. The default path is ignored local runtime state; use another writable directory to retain results elsewhere.