Pylab

AAAO Ecosystem Bottleneck Lab

A working Gradio prototype for theory-informed entrepreneurial ecosystem bottleneck diagnosis.

Core workflow

1. Data registry — select/upload Tracxn, YNOS, PATSTAT, Startup India/DPIIT, grants, procurement, regulation, ESO, Prowess/MCA or custom datasets.
2. Common stage matrix — integrate venture-level Technology, Finance, ESO Support, Government Procurement and Regulatory Approval stages.
3. Policy/supporting documents — upload policy, mission, scheme, procurement or regulatory PDFs/DOCX/TXT.
4. Research papers — upload the papers that define constructs/framework conditions.
5. AAAO + Theory of Constraints diagnosis — compute domain-stage mismatches, observable coverage and evidence-weighted candidate constraints.
6. Report — generate a deterministic evidence-backed report and optionally an LLM synthesis.

The included demo is the 200-company Tracxn domain-stage classification used in the prototype discussion.

Run locally

python -m venv .venv

# Windows
.venv\\Scripts\\activate

# macOS/Linux
# source .venv/bin/activate

pip install -r requirements.txt
python aaao_app.py

Gradio will print a local browser address (normally http://127.0.0.1:7860).

Optional LLM synthesis

The platform works without an API key: bottleneck metrics and the deterministic report are calculated locally.

For AI synthesis, either set:

export OPENAI_API_KEY="..."

or paste a key into the password field in the app. The prototype does not write the key to disk.

Methodological safeguards

• Missing is not weak. Stage 0 / missing Regulatory, Procurement or ESO evidence means not observable in the selected datasets, not absence of that ecosystem function.
• Mismatch is not yet a binding constraint. A cross-sectional stage gap is reported as a candidate constraint. A stronger Theory-of-Constraints claim requires longitudinal evidence that the gap delays or prevents venture progression.
• Coverage affects priority. The platform uses an evidence-weighted priority score so a large mismatch supported by sparse data does not automatically outrank a smaller but much better-observed mismatch. Tracxn-only procurement/regulatory mentions are shown as contextual signals but do not count as direct transaction/administrative coverage.
• AAAO localizes the mechanism. A domain-level candidate is followed by the question: Which Actor controls/provides which Asset, through which Activity, constraining which Outcome?

Data model

Common relational grammar:

Startup / Entity + Date + Actor + Asset + Activity + Outcome + Domain + Stage + Source

Diagnostic matrix:

Venture × Venture-development state × Support domain

Venture-development states

1. Opportunity & Conception
2. Development & Validation
3. Commercialization & Market Entry
4. Survival & Traction
5. Growth & Expansion
6. Scale-up & Stability

Support domains

• Technology
• Finance
• Entrepreneurial Support Organizations (ESO)
• Government Procurement
• Regulatory Approval

Source integration behavior

Tracxn

The prototype automatically reads either:

• a Domain Stage Classification sheet, or
• a raw Companies Covered 1.1 export.

For raw exports, Technology and Finance are inferred from observable company-stage/revenue/employee variables; Regulatory and Procurement are assigned only when explicit textual evidence is present.

Other sources

If a file already contains stage variables, the platform auto-detects common names such as technology stage, finance stage, procurement stage, and regulatory stage. You can also type explicit column names in the custom mapping panel.

This prototype intentionally does not force arbitrary PATSTAT, procurement or regulatory schemas into stages automatically. Production deployment should add source-specific adapters and persistent entity IDs.

Files

• aaao_app.py — Gradio user interface
• core.py — integration, staging, document extraction and bottleneck analytics
• sample_data/sample_tracxn_domain_stage.xlsx — demo dataset
• requirements.txt — dependencies
• ARCHITECTURE.md — research/technical architecture and production extensions
• run_windows.bat — one-click Windows setup/run helper
• run_mac_linux.sh — setup/run helper for macOS/Linux