Chidori — Deterministic Agents as Code

Chidori is a YAML-free AI agent framework where agents are written as Starlark scripts — a deterministic Python dialect that enables checkpointing, replay, and visual editing.

Every side effect — LLM calls, tool invocations, HTTP requests, human input — goes through a fixed set of host functions the runtime can log, cache, and replay. Save a session's call log to disk, replay it later for byte-identical output with zero LLM calls.

Why Chidori?

• Agents look like Python. Native control flow, variables, list comprehensions — no template DSL, no YAML.
• Deterministic execution. Starlark has no I/O of its own; every side effect is a logged, cacheable host function call.
• Zero-cost checkpointing. Save a session's call log, replay it for identical output at zero token cost.
• Event-driven. Run an agent as an HTTP server that reacts to webhooks, alerts, or custom events.
• Rust core, Python SDK. The runtime is a single binary. The Python SDK is pure-stdlib HTTP — no pip install, no native bindings.

A first agent

# agents/summarizer.star
config(model = "claude-sonnet")

def agent(document):
    summary = prompt("Summarize in 3 bullets:\n" + document)
    actions = prompt("Extract action items:\n" + summary)
    return {"summary": summary, "action_items": actions}

chidori run agents/summarizer.star --input document=@notes.txt