Figure: Admin Interface Overview. The pipeline provides a web-based administrative interface for orchestrating large-scale phenotype prediction experiments. The interface enables systematic evaluation of multiple LLM models across diverse species cohorts, with real-time progress monitoring and automated job scheduling. Users can dynamically expand the model selection via OpenRouter integration (dropdown shown), track prediction status for each species-model combination, and access detailed extraction statistics including species processed, success rates, and failure counts for individual jobs.
A. Processing Dashboard
Phenotype Prediction Jobs
Species × model combinations currently monitored in the admin panel
Template v07 · Phenotype baseline prompt
Tracks trait predictions per species file against the current model roster.
| Species Files |
gpt-4o-mini
|
llama-3.1-70b
|
gemma-2-27b
|
Add Model from OpenRouter
llama-3.1-8b
llama-3.3-70b
llama-3-70b
|
|---|---|---|---|---|
|
species_WA.txt
|
3,876/3,876
✓3,876
Complete
|
3,872/3,876
✓3,871
✗1
Running…
|
Queued
⌛ 3,876 prompts
|
|
|
species_LA.txt
|
3,876/3,876
✓3,873
✗3
Failed
|
3,876/3,876
✓3,876
Complete
|
3,120/3,876
✓3,118
✗2
Running…
|
|
|
marine_bacteria.txt
|
3,876/3,876
✓3,876
Historical
|
|
3,876/3,876
✓3,876
Complete
|
|
|
soil_microbes.txt
|
3,876/3,876
✓3,876
Complete
|
Queued
⌛ 3,876 prompts
|
|
|