Skip to content

Interactive guide

Who this is for: researchers and lab scientists — not developers.

Last checked: automated tests on PraisonAIBio v0.2.0 (51 checks passed, repo validated, benchmarks green).

At a glance

Area Status Plain English
Find models ✅ Search and rank BioModels in minutes
Read model structure ✅ Summarise SBML and reaction maps
Simulate ⚠️ Works when BASICO is installed
Compare & trust ✅ Score curation and diff simulations
Export a study bundle ✅ Zip SBML, notes, and rerun commands
Safety rules ✅ Blocks risky saves in public mode
Ready-made workflows ✅ 32 YAML pipelines (8-step demo included)
Recipes ✅ 5 packaged studies
How-to skills ✅ 16 short guides
Connect to other apps ✅ 28 tools via MCP

One model to try

BIOMD0000000206 — yeast glycolysis. Used in examples, cookbooks, and benchmarks.

Which path should I take?

Best when: you know what you want (search, one model ID, trust check).

  1. Install → pip install -e src/praisonai-bio
  2. Pick a quick task — copy, paste, run
  3. Start with examples/small/01_search.py

Needs internet to BioModels.org. No OpenAI key.

Best when: your question is open-ended (“models for p53 under DNA damage?”).

  1. Install + set OPENAI_API_KEY
  2. Run examples/big/01_find_models.py
  3. Or: praisonai workflow run workflows/discovery/biomodels_discovery_pipeline.yaml

Best when: you want discovery → check → simulate → report in one go.

praisonai workflow run workflows/cookbooks/full_platform_pipeline.yaml

Eight steps: find → inspect → map → you approve → simulate → perturb → compare → export.

Best when: you repeat the same study type.

export PRAISONAI_RECIPE_PATH=recipes/bio
praisonai recipe run biomodels-discovery

See Recipes for all five.

Do I need BASICO?

Answer three quick questions:

Will you run time-course simulations on your machine?

Yes → install simulation extras: 

pip install -e "src/praisonai-bio[simulation]"
No → discovery, trust scores, SBML maps, and workflows still work.

Will you only search and read models?

Yes → no BASICO needed. Use small examples 01–04 and 10.

Will you use Jupyter notebooks?

Notebooks 01–02 work without BASICO. 03–04 need BASICO + network.

What does the trust score mean?

Curated vs non-curated

Curated = manually checked on BioModels. Prefer these for publications.
Non-curated = use with extra caution.

Trust scorecard (tool)

Run examples/small/03_trust_score.py — returns curation status, publication links, and flags.

Human approval step

Full workflows pause before simulation so you can reject bad assumptions. Look for approve steps in YAML.

Your first glycolysis study (5 steps)

graph LR
  A[Search] --> B[Trust check]
  B --> C[SBML map]
  C --> D[Simulate]
  D --> E[Export bundle]
  classDef step fill:#8B0000,color:#fff
  class A,B,C,D,E step
Step You do Command / workflow
1 Search python examples/small/01_search.py
2 Trust python examples/small/03_trust_score.py
3 Map python examples/small/10_sbml_graph.py
4 Simulate python examples/small/05_simulate.py (BASICO)
5 Export python examples/small/09_repro_export.py

Or one workflow: workflows/cookbooks/glycolysis_demo.yaml

Feature checklist (test results)

Search & discovery — ✅ Works
  • Search BioModels by keyword
  • Rank and filter curated models
  • Advanced field search

Try: Quick task 1

SBML & reaction maps — ✅ Works
  • Summarise model contents
  • Validate SBML structure
  • Build reaction graph (local or download PNG/SVG)

Try: examples/small/04_validate_sbml.py, 10_sbml_graph.py

Simulation — ⚠️ Needs BASICO
  • Baseline time-course
  • Parameter perturbation
  • SED-ML-aware runs (when SED-ML file exists)

Install: pip install -e "src/praisonai-bio[simulation]"
Try: examples/small/05_simulate.py

Compare & trust — ✅ Works
  • Compare two models (metadata + structure)
  • Compare two simulation runs (RMSE-style metrics)
  • Trust scorecards

Try: examples/small/07_compare_models.py, 08_compare_sims.py (08 works offline)

Export & reproducibility — ✅ Works
  • Repro bundle with checksums and rerun script
  • COMBINE archive when BioModels provides it

Try: examples/small/09_repro_export.py

Safety & policies — ✅ Works
Mode You can…
bio-public Search and read — no saving
bio-lab Simulate, perturb, export

Set: export PRAISONAI_POLICY_PACK=bio-public

Workflows & recipes — ✅ Works
  • 32 workflow files validated
  • 8-step full platform pipeline
  • 5 recipes (biomodels-discovery, research, simulate, perturb, report)

Workflows guide · Recipes

Skills (16 guides) — ✅ Works

Short how-tos for agents — search, simulate, orchestrate, repro, and more.
Full skills list

Benchmarks & quality — ✅ Works
  • 10 Talk2BioModels-style routing checks — pass
  • Simulation accuracy check — pass
  • 51 automated unit checks — pass

(For developers: python benchmarks/run_all.py)

Something failed?

Install errors

Use GitHub install until PyPI publish: 

git clone https://github.com/MervinPraison/PraisonAIBio.git
cd PraisonAIBio && pip install -e "src/praisonai-bio[dev]"

Network / BioModels

Check internet and https://www.biomodels.org. Optional: BIOMODELS_BASE_URL.

Simulation failed

Install BASICO: pip install -e "src/praisonai-bio[simulation]".
Confirm model ID (try BIOMD0000000206).

Agent / workflow errors

Run import praisonai_bio first. Set OPENAI_API_KEY for LLM steps.
Install wrapper: pip install praisonai.

Permission / save blocked

Switch policy: export PRAISONAI_POLICY_PACK=bio-lab for export and save tools.

Talk2BioModels & ClawBio

Tool Role
vs Talk2BioModels Same 11 core tools + discovery extras
vs ClawBio ClawBio = genomics; PraisonAIBio = mechanistic models — use together

Next steps