SRHarness¶
SRHarness is a research runtime for agentic symbolic regression. Language models inspect data, propose formulas, fit parameters, and validate candidates through a controlled tool set. Search steps, tool results, model usage, and candidate formulas are retained as auditable run records.
Open the live WebUI View source
Where to start¶
- Install: install SRHarness and optional dependencies
- Quick Start: complete your first run
- SRHarness: understand commands, runtime options, and output files
- SRHarness Agent Workflow: understand the Agent loop, tool contract, parsers, and R-C-L-K search
- SRHarness Core Abstractions: learn the extension interfaces for tools and tool-call Parsers
- Structured Data and
context.data: reference storage rules for variables, axes, graphs, and hypergraphs - Formula Evaluation and Custom Evaluators: understand formula evaluation, built-in Evaluators, and custom extensions
- SRHarness WebUI: complete an interactive search in the browser
- SRHarness Engine: learn the expression syntax, parameters, and graph structures
- API Reference: look up Python types and method signatures