Research Workspace

The Quant Research workspace is the primary authoring surface for quant projects.

It is not the old PyPScript editor. Quant projects should use Research as their main coding surface.

Supported files

Today, the workspace intentionally supports only three files:

• strategy.py
• quant.config.json
• README.md

That restriction is intentional. It keeps the contract explicit and avoids pretending PyP already supports arbitrary multi-file Python packages.

What each file is for

strategy.py

This is the executable strategy source.

It contains:

• imports for supported packages
• helper functions
• train(data, config)
• predict(model, market_data, config)

quant.config.json

This is the environment and runtime declaration for the strategy.

It contains:

• market targeting
• parameters
• training package declarations
• inference compatibility declarations
• artifact target selection

README.md

This is human-facing documentation for the project.

Use it for:

• notes about the signal idea
• assumptions
• feature definitions
• release context for teammates

Do not use it for execution-critical configuration.

Why the workspace is constrained

A real unrestricted Python project implies:

• packaging rules
• import resolution
• dependency graphs
• build context assembly
• much more complex deployment behavior

PyP does not expose that yet inside Quant Mode. The current Research workspace is a controlled environment optimized for a reliable training-to-artifact pipeline.

Expected workflow

1. Start in strategy.py
2. Configure market and artifact behavior in quant.config.json
3. Keep notes in README.md
4. Save the workspace
5. Move to Jobs to run training

Best practice

Keep the strategy contract explicit:

• do feature preparation in named helpers
• keep train() deterministic
• keep predict() narrow and deployment-safe
• document assumptions in README.md

If you are trying to add many support modules, you are probably exceeding the current Research contract and should simplify the design before training.