Chi (χ)

Chi ("kai") is an open-source autoresearch harness: point a fleet of LLM coding agents (any vendor) at any problem with a programmatic evaluator — a build, a correctness check, and a score — and let them iterate unattended.

Status: v1 (Phases 0–1) — single coder agent, two adapters (headless vendor CLIs and a LiteLLM tool loop), enforced SQLite+JSONL run store, hard budget caps, two-layer steering, deterministic watchdog.

Quick start

uv tool install --editable .   # or: uv venv --python 3.12 && uv pip install -e ".[dev]"
chi                            # opens the full-terminal session (Textual UI)

The session has a Claude Code-style interface: scrolling transcript, bottom input with a slash-command dropdown, live status bar, and modal fuzzy pickers. chi --plain gives a minimal line-based REPL (also used automatically when stdout is not a terminal).

Inside the session:

/vendors            pick providers (fuzzy); store keys with `chi providers --set-key X`
/models             pick coder models — saved as your global defaults
/run fleet.yaml     start a run; iteration results stream in live
just type           plain text while a run is active becomes a steering directive
/stop  /status  /ledger  /champion  /quit

Everything also works non-interactively for scripts and CI:

chi providers --enable anthropic,deepseek
chi models --pick anthropic/claude-sonnet-5,claude
chi validate examples/fleet.yaml
chi ping --fleet examples/fleet.yaml
chi run examples/fleet.yaml
chi steer runs/<run_id> "stop micro-tuning; try itertools"

Inspect results:

uv run chi status runs/<run_id>
uv run chi ledger runs/<run_id> --negative
uv run chi champion runs/<run_id> --export best.py

Define your own problem

A problem is a directory with a problem.yaml (see problems/optimize_function/): entrypoint commands for correctness and benchmark, held-out seeds, and a score metric/direction. Correctness is a hard gate; candidates never see reference outputs.

Coder adapters

  • litellm_loop — tools-in-a-loop over any LiteLLM-routable model (Anthropic, OpenAI, DeepSeek, GLM, MiniMax, ...).

  • cli_subprocess — drives a headless vendor CLI per iteration, e.g.:

    coders:
  - id: c1
    model: claude
    adapter: cli_subprocess
    command: "claude -p --permission-mode acceptEdits {prompt_file}"

  • scripted — deterministic playback, used by the test suite.

Design

See docs/superpowers/specs/2026-07-25-chi-v1-design.md (v1 design, grounded in a multi-agent fleet postmortem) and docs/product-spec-v1.md (full product spec). License: Apache-2.0.

Key dates

PyPI data
Data sourced directly from PyPI's database.
  • Released:
Newer release available (0.2.0)

1 maintainer

PyPI data
Data sourced directly from PyPI's database.
Avatar for kdpisda from gravatar.com kdpisda

License

Apache-2.0

Requires

Python >=3.11

Provides Extra

dev

Report project as malware

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

getchi-0.1.0.tar.gz (386.5 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

getchi-0.1.0-py3-none-any.whl (108.0 kB view details)

Uploaded Python 3

File details

Details for the file getchi-0.1.0.tar.gz.

File metadata

  • Download URL: getchi-0.1.0.tar.gz
  • Upload date:
  • Size: 386.5 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for getchi-0.1.0.tar.gz
Algorithm Hash digest
SHA256 afeff72e9351ab7ba3e455364ac2f15c0df0f08a194eb55b8eb14af3b37baf46
MD5 46f1974f12fc7e87ed7a1368d83605c2
BLAKE2b-256 0fafcaa665c0dc001c1419c28aa535408ea5d88df99782555a3c39b09bd80d45

See more details on using hashes here.

Provenance

The following attestation bundles were made for getchi-0.1.0.tar.gz:

Publisher: publish.yml on kdpisda/chi

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file getchi-0.1.0-py3-none-any.whl.

File metadata

  • Download URL: getchi-0.1.0-py3-none-any.whl
  • Upload date:
  • Size: 108.0 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for getchi-0.1.0-py3-none-any.whl
Algorithm Hash digest
SHA256 6ffb5e4d6a25b692722473e8fa5104fa5f6cefda721b8e0e6c242de2ed2145f4
MD5 590df1a0713a2370db36d4f5a3dc47df
BLAKE2b-256 d92f3c9a1b57d534dcc4753e5bdb32906b27bf777d209359967b77e16394667a

See more details on using hashes here.

Provenance

The following attestation bundles were made for getchi-0.1.0-py3-none-any.whl:

Publisher: publish.yml on kdpisda/chi

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page