finkritintel

The bridge layer of the finkrit stack. It exposes finkritq (the deterministic quant core) as framework-neutral tool contracts an agent can call, without depending on any particular LLM framework.

Deliberately pydantic-free and agent-framework agnostic, so you can wire these tools into pydantic-ai, LangChain, an MCP server, or your own runtime. finkrit's own agent layer (finagent) is just one consumer.

Install

pip install finkritintel      # pulls finkritq

What it gives you

• ToolContract — agent-facing metadata for a tool (name, description, tags), independent of how it is implemented.
• ToolBinding — a contract paired with an input and output schema and a concrete finkritq implementation. binding.execute(...) runs it.
• Capability — a named group of bindings (risk, performance, optimization) an agent exposes as one unit.

The integration.finkritq subpackage wires finkritq functions into ready-made bindings, with room for other data sources later.

Example

List the tools a capability offers, framework and LLM free:

from finkritintel.capability.risk import RISK_CAPABILITY

for binding in RISK_CAPABILITY.tools:
    print(binding.contract.name, "-", binding.contract.description)

Each binding carries the schema and the callable, so a framework author can turn it into a tool however that framework expects.

License

Apache-2.0. See LICENSE.