Republic

Use LLM capabilities like regular Python components, with auditable execution traces by default.

Republic is not a bigger framework. It is a small set of composable primitives:

• LLM: One entry point for chat, tools, stream, and embeddings.
• StructuredOutput: Key interfaces return value + error.
• Tape: Append-only records with anchor/handoff/context/query.
• ToolExecutor: Tool calls can be automatic or manual.

30-Second Preview

from republic import LLM

llm = LLM(model="openrouter:openrouter/free", api_key="<API_KEY>")
out = llm.chat("Explain tape-first in one sentence.", max_tokens=48)

if out.error:
    print(out.error.kind, out.error.message)
else:
    print(out.value)

What You Get

• Smaller API surface with stronger control.
• Visible tool execution paths without hidden magic.
• Run/tape-level behavior tracing for debugging and audits.
• Both text streaming and event streaming for CLI, web, and workers.

---

This project is derived from lightning-ai/litai and inspired by pydantic/pydantic-ai; we hope you like them too.