An open-source hybrid scientific research agent.
JSON tools orchestrate; persistent Python/R kernels do the science.
Launched by the Peking UniversityโYuanKong Intelligence AI Joint Research Laboratory.
็ฑๅไบฌๅคงๅญฆโๅ
็ฉบAI่ๅๅฎ้ชๅฎคๆจๅบใ
English ยท ็ฎไฝไธญๆ
Tip
Why "two cuts"? No pricey frontier-model key needed โ OpenAI4S runs on Doubao (่ฑๅ
) via the cheapest "Small" plan on Volcengine Ark (็ซๅฑฑๆน่): ยฅ9.9 / month (โ US$1.4). Pick the ark provider in the UI and you get a Claude-Science-class agent for less than a cup of coffee.
Volcengine Ark ยท Agent Plan (Personal) โ the entry Small tier is ยฅ9.9 / month.
OpenAI4S deliberately has two action planes. Provider-native JSON tool
calls handle deterministic orchestration, permissions, metadata, external
services, and human approval. Python/R Code-as-Action handles computation,
exploration, analysis, simulation, and long-running scientific work in
persistent kernels. Python cells can synchronously call the in-kernel host
API while they run; R is an independent persistent analysis channel.
This is not a choice between tools and code: each does the job it is good at.
Tool-only and conversational work can finish through the Engine-owned,
strictly structured finalize_response action. Scientific cells keep the
important host.submit_output(...) completion contract, including structured
artifacts and metrics. host.submit_output is the only completion signal that
can fire inside a Cell; a later sole finalize_response may still close the
Engine after earlier Cells have run.
| JSON control plane | Python/R science plane | |
|---|---|---|
| Best for | workflow, permissions, metadata, services | computation, analysis, simulation |
| Action unit | One ordered native-tool batch | One complete code cell |
| Composition | auditable schemas and resource policy | for, if, libraries; Python also has mid-cell Host RPC |
| State | append-only Action Ledger | kernel memory + versioned artifacts |
| Completion | Engine-owned finalize_response | Python: host.submit_output(...); R: no in-cell completion |
| Extending | named Tool subclass | import a library or load a Skill |
# ReAct: ~14 round-trips (read โ โฆ โ filter โ sort โ plot). OpenAI4S: one code cell.
hits = [f for f in files if pattern in host.read_file(f)]
top3 = sorted(hits, key=os.path.getsize, reverse=True)[:3]
frames = [pd.read_csv(f) for f in top3] # a 100k-row DataFrame stays in the kernel...
host.save_artifact(plot(frames)) # ...only "<DataFrame 100000ร20>" hits context | ||
2026-10-18๐v0.3.1โ local project folders and portable history โ link source folders, browse saved conversation history, import scientific datasets with checksum verification, and work with 611 bundled Skills. Linux and Windows/WSL2 packages are on GitHub, Python packages on PyPI, and the public container image on GHCR. macOS uses PyPI while signing and notarization are deferred. The database moves from schema 32 to 35; read Upgrading to 0.3.1 before the first start.2026-09๐งญv0.3.0โ the new workbench, and the first Windows package โ the Preact/TypeScript workbench replaces theapp.jsmonolith as the default UI, and the release adds the first Windows/WSL2 zip beside the Linux tarball. Every local daemon now requires its access token: theOPENAI4S_REQUIRE_TOKEN=0loopback opt-out is gone. Interactive HTML reports run on a separate sandboxed origin, Notebook cells keep the exact artifact versions they produced, long-context compaction actually lands and survives a restart, delegated sub-agents inherit their environment and leave durable, exportable cell records, and Auto Mode admits budget atomically and stops a run that makes no progress. The single-cell RNA analysis Skill brings the total to 604. The Apple Silicon macOS image is an ad-hoc-signed preview, not a notarized one. The database schema moves from 27 to 32, so read Upgrading from 0.2.x before the first start.2026-08-24๐v0.2.0โ the multi-platform release โ one release, two desktop packages: the Apple Silicon.dmgand a relocatable Linuxx86_64tarball carrying the same embedded Python and science stack (the Windows/WSL2 zip is built and under stabilization โ it ships in a coming release). Underneath: Auto Mode with a Guardian review boundary, honest completion-evidence reconciliation (a crashed cell can no longer render as a clean success), the MCP Streamable HTTP transport with the Volcengine DataPro connector and Doubao web search, Anthropic Messages SSE streaming, the pinned 561-recipe bioSkills collection (603 Skills in all, installable anywhere vianpx), a trajectory-ledger view in the workbench, Docker/Kubernetes deployment,openai4s --version, and the interrupt-signal train that makes a running R cell reliably stoppable on every platform.2026-08-04๐ญmainโ on the way tov0.2.0โ read-only session sharing over an outbound relay tunnel (openai4s share/openai4s relay), seven normalized public-database connectors that carry where a record came from and when, a versioned/api/v1surface (keyset pagination, one error envelope, a resumable WebSocket cursor), environments as a transaction (openai4s env plan|apply|rollback), a redacteddoctor/diagnosticssupport bundle, consent-gated revocable telemetry, a retrosynthesis-planning Skill, and a 10-workflow / 20-case benchmark that runs against the real Store, kernels, and dispatcher. Linux and Windows desktop packages were built and tested here โ the Linux package ships inv0.2.0above, and the Windows package follows in a coming release.2026-07-15๐v0.1.0โ macOS app โ a one-click, no-toolchain Apple Silicon.dmgwith an embedded Python and the full default kernel science stack (rdkit ยท scanpy ยท the single-cell stack), plus PyPI packaging (pip install openai4s) and release automation. New here? โ Startup guide.2026-07-06๐ Open-sourced โ the pure-stdlib Code-as-Action engine, the scientific web app, 24 science Skills, and BYOC remote compute.
- ๐งฌ Hybrid action engine โ class-based native JSON tools orchestrate while persistent Python/R kernels execute science. CLI and Web adapters start foreground language slots lazily, so tool/finalize routing itself does not spawn one; individual tools may still manage dedicated workers.
- ๐ Ledger-first runtime โ action groups/events and terminal facts are append-only; execution attempts, generation lifecycle, usage, and completion records remain durable and reconstructable.
- ๐ Pure-stdlib core โ the engine and the web server are stdlib-only (
http.server+ hand-rolled WebSocket, no framework, no deps). The LLM client speaks OpenAI / Anthropic / Gemini overurllibalone. - ๐ One-line multi-provider โ
ark(doubao ยท glm ยท kimi ยท deepseek ยท minimax) plus officialchatgpt ยท claude ยท gemini, behind a singlehost.llm; switch from the UI. - ๐ฅ๏ธ Scientific workbench โ live streaming, versioned artifacts, provenance, an Action Timeline surface, and a read-only-by-default Notebook. An explicit developer flag enables multiline Python/R input against the shared kernels.
- ๐ Hardened local execution โ strict child-environment allowlists, durable approvals, one-shot generation-bound
host.bashcapabilities, and OS sandbox adapters (Seatbelt on macOS, bubblewrap on Linux) with visible degraded/fail-closed modes. - ๐ฌ 612 bundled Skills โ 51 curated OpenAI4S recipes for GPU/model science, research workflows, and platform operations, plus all 561 recipes from the pinned MIT-licensed GPTomics/bioSkills collection. Skills are recipes of code, not JSON schemas; the large third-party collection is searched on demand and occupies only one always-on prompt line. User-authored Skills stay under the data directory and cannot shadow bundled trust.
- โ๏ธ BYOC remote compute โ with a configured, reachable provider, dispatch GPU jobs via
ssh:<alias>or the bundled NVIDIA NIM integration. General remote compute remains a Prototype surface;host.folduses a strict no-fabrication policy. - ๐ Read-only session sharing โ publish a session as a snapshot anyone with the link can view and import, through a relay you run. The daemon never binds a public port; it dials out. Memories, permission state, and keys never leave, and residual secrets fail the publish closed. โ Web sharing
- ๐ Source-attributed retrieval โ seven normalized public-database connectors (UniProt ยท RCSB PDB ยท Ensembl ยท ChEMBL ยท PubChem ยท arXiv ยท OpenAlex). Retrieved records carry where they came from and when, without the API key that fetched them.
- ๐งฐ Operable, not just runnable โ a versioned
/api/v1(keyset pagination, one error envelope, correlation IDs, a resumable WebSocket cursor), a local credential required at startup, a redacteddoctor/diagnosticssupport bundle, and consent-gated telemetry that is off by default and destroys its identity when revoked.
A capability map of the current tree โ what is implemented and reachable, plane by plane.
| plane | what's implemented |
|---|---|
| Control & orchestration | class-based native Tools ยท append-only Action Ledger ยท plan/review with a durable state machine ยท context compaction that archives the raw slices it summarizes ยท concurrent sub-agent delegation (fanout 48, depth 4) a user can stop mid-flight ยท enforced Specialist allowlists a child cannot widen ยท MCP connectors ยท cross-session memory |
| Scientific execution | persistent Python and R kernels ยท synchronous mid-cell host RPC ยท object-level data lineage ยท versioned artifacts ยท environment provenance recorded per kernel generation, never borrowed from the daemon ยท background execution ยท 612 Skills (51 curated + 561 pinned bioSkills) ยท a FIFO execution coordinator with ABA-safe watchdog recovery |
| Data & retrieval | seven normalized public-database connectors (UniProt ยท RCSB PDB ยท Ensembl ยท ChEMBL ยท PubChem ยท arXiv ยท OpenAlex) whose records carry source and time ยท a nightly canary over three of them ยท Agent-Plan-keyed Doubao Search Custom as the primary web search ยท Tavily and keyless search as backups ยท managed DataPro professional-dataset search |
| Workbench | live streaming ยท Action Timeline ยท read-only-by-default Notebook ยท branch fork/activate/revert ยท verified recovery with an explicit Partial/Failed state ยท @file references pinned to the version they name ยท 2D chemistry/genome/sequence/MSA/LaTeX renderers ยท Markdown and .ipynb export |
| Sharing & portability | read-only session shares over an outbound relay you operate ยท quarantined portable Session packages ยท an optional Jupyter KernelSpec bridge onto the same kernels |
| Ops, safety & release | /api/v1 and a startup credential ยท Seatbelt/bubblewrap sandbox adapters with visible degraded and fail-closed modes ยท durable approvals that deny by default when unattended ยท redacted diagnostics ยท revocable telemetry ยท environments as a transaction ยท a 13-workflow/46-case benchmark against the real Store, kernels, and dispatcher ยท a staged release pipeline that verifies artifacts before anything becomes public |
A default-off semantic judgment layer (TypeSafe Jev) can add Skill suggestions, literature claim checks, text-feature engineering, and recording-only safety / task-mode shadows. It stays off until you enable it, needs your own TypeSafe key, and sends text to a service hosted in the United States. โ Experimental semantic judgment
Live API workflow โ from UniProt / RCSB to a 3D structure & report |
Real-data analysis โ human insulin INS (P01308): from UniProt / RCSB to a reproducible report |
Visual artifact editing โ โraise the confidence cutoff to 75โ in one line |
Annotation-driven chart editing โ lasso a region & recolor the legend |
Plan-mode research โ artemisinin & paclitaxel solubility prediction |
Protein engineering โ from sequence to ranked mutants & structural rationale |
Current release: 0.3.1. Choose an installation channel below. If you already have data from 0.3.0 or earlier, stop the daemon and back up the whole data directory before installing; see Upgrading.
| Channel | 0.3.1 download |
|---|---|
| Linux x86_64 | Standalone tarball |
| Windows x86_64 with WSL2 | Complete ZIP |
| Python, including macOS | PyPI: openai4s 0.3.1 |
| Containers (linux/amd64) | GHCR: openai4s, tag 0.3.1 |
| Download verification | SHA256SUMS ยท Build and validation evidence |
To run the released source:
git clone --branch v0.3.1 --depth 1 https://github.com/PKU-YuanGroup/OpenAI4S && cd OpenAI4S
./setup.sh # one-time: build the environment with uv
./start.sh # launch the web UI at http://127.0.0.1:8760/setup.sh creates the lightweight control .venv with uv. For the comprehensive Python + R scientific kernels, install a Conda-family manager (micromamba, mamba, or conda) and run ./setup.sh --with-kernel-envs instead. Existing kernel environments can be synchronized with ./setup.sh --update-kernel-envs; updates do not prune user-installed packages. start.sh launches the daemon + web UI. No API key is needed to boot โ set your model in the UI (Customize โ Models). One-shot without the UI: uv run openai4s run "Compute the mean of [4,8,15,16,23,42] and submit it." -v.
Note
0.3.1 has no macOS DMG while Developer ID signing and notarization are deferred. Use PyPI 0.3.1 on Apple Silicon or Intel, or the tagged source checkout above. The old preview images install older versions; see the historical notes in the startup guide.
Install from PyPI into a virtual environment of its own. This works on Apple Silicon and Intel, with Python 3.10 or newer:
python3 -m venv ~/.venvs/openai4s && source ~/.venvs/openai4s/bin/activate
python -m pip install --upgrade "openai4s[science]==0.3.1" # use "openai4s[science,chemistry]==0.3.1" for RDKit
openai4s serve # starts the daemon and opens the workbench with its access tokenData lives in ~/.openai4s. If you close the tab, openai4s url prints the authenticated workbench URL again. The R kernel needs a Conda-family manager (micromamba, mamba or conda); run openai4s setup once to build it.
First run โ point it at a model, then at search. No key ships, so once the workbench is open:
- Model API โ open Settings โ โ Models, pick a protocol (Ark-compatible for Doubao/GLM/Kimi/DeepSeek/MiniMax, or OpenAI- / Anthropic-compatible), paste your API Key, click Add, then Set active. Cheapest path: the
arkprotocol on Volcengine Ark's ยฅ9.9/mo plan. - Search API (optional, recommended) โ open Settings โ โ Network, keep Allow network access on, and paste your Ark Agent Plan Key into the primary Doubao Search Custom card โ Save credential. If the active model is an Ark Agent Plan, you can skip this step: OpenAI4S saves that key as the default for both Doubao Search and DataPro professional datasets, and they keep working after you switch to another model. Tavily and keyless engines remain backup options; the dedicated Doubao health check never reports a fallback result as Doubao.
Full 0.3.1 walkthrough (install โ model โ search โ R kernel): Startup guide.
Download the 0.3.1 Linux tarball and SHA256SUMS into the same directory. Verify it, unpack it anywhere, and run it. It embeds its own Python and the pre-baked science stack, as a relocatable directory:
sha256sum --ignore-missing -c SHA256SUMS # the Linux tarball must report OK
tar -xzf OpenAI4S-0.3.1-linux-x86_64.tar.gz
cd OpenAI4S-0.3.1-linux-x86_64
./OpenAI4S # starts the daemon and opens http://127.0.0.1:8760/
./install.sh # optional: `openai4s` on your PATH + an application-menu entryinstall.sh is per-user and needs no root โ it only writes into $HOME, and ./uninstall.sh undoes it while leaving your data in ~/.openai4s alone. Install bubblewrap (apt install bubblewrap) so cells run sandboxed; without it the default OPENAI4S_KERNEL_SANDBOX=auto reports a visibly degraded, unisolated kernel. Only x86_64 is published โ on arm64 Linux, use a virtual environment and the pinned PyPI command above.
Download the complete 0.3.1 Windows ZIP and SHA256SUMS, verify the ZIP as described in the Windows/WSL2 guide, then extract all files and double-click OpenAI4S.cmd. No split-file restoration is needed. The first run checks WSL2 and a working bubblewrap 0.8.0+ sandbox, verifies and installs the bundled Linux payload, creates ~/.local/bin/openai4s, starts the daemon there, and opens an authenticated local URL in your Windows browser. No application download, no pip, no toolchain. Ubuntu 24.04 is the supported baseline; mainland PyPI/Conda mirrors and an optional WSL-reachable proxy can be configured by the launcher.
0.3.1 validation: archive structure, launcher checks, and the embedded Linux payload passed verification; full Windows/WSL2 runtime acceptance was not performed for this version.
v0.3.0 is the first release that ships this package. Its acceptance evidence covers WSL2 on x86_64 with the tested Ubuntu 24.04 distribution. The last section of the WSL2 parity audit ("Fix verification โ 2026-09-07") leaves the following unverified:
- A side-by-side macOS run for the parity comparison. The macOS side of that comparison was read from source, not run.
- Windows on ARM, other distributions, and WSL network modes other than the tested one.
- Real provider sign-in and inference. The scientist flow ran with
OPENAI4S_NOTEBOOK_REPL=1and no live model. - Conda environment provisioning, so R and other named environments on Windows are unverified. R was installed in the test distribution only as a test prerequisite.
- Every domain recipe.
- A Windows reboot. Only a restart of the WSL distribution was tested; it reopened the saved results and a stored credential.
- Cold-start performance. The unmodified full browser smoke did not pass: it exceeded its 20-second queue-admission wait. WSL service connection timeouts were also seen under concurrent load.
Native Windows is not supported, and the program refuses to start a kernel there rather than warning and proceeding โ it spawns POSIX subprocesses, the R channel rides file descriptors 3 and 4 through a shell redirection, and the sandbox has no Windows backend. WSL2 reports as Linux, so this package runs the same build every other platform runs. If you do not have WSL2 yet, the launcher stops and tells you the exact command (wsl --install, from an Administrator PowerShell). Details: Supported platforms.
The public GHCR package can be pulled without signing in. Pin 0.3.1 for this release; latest follows future releases. The image is linux/amd64; ARM hosts require amd64 emulation.
docker pull --platform linux/amd64 ghcr.io/pku-yuangroup/openai4s:0.3.1
docker run -d --name openai4s --platform linux/amd64 \
-p 127.0.0.1:8760:8760 \
-v openai4s-data:/data \
-e OPENAI4S_SECRET_LLM_LLM_API_KEY="$OPENAI4S_LLM_API_KEY" \
ghcr.io/pku-yuangroup/openai4s:0.3.1
docker exec openai4s openai4s url # the URL, token includedThe image contains Debian-slim CPython, the wheel, and the science extra, and runs as an unprivileged user with one volume at /data. The command forwards the model key from your shell's OPENAI4S_LLM_API_KEY to the container's OPENAI4S_SECRET_LLM_LLM_API_KEY; configure the provider and model as described in the container guide. Credentials are read from the environment. To build from the checkout instead, use docker compose up -d --build. For Kubernetes, replace image: openai4s:local in deploy/kubernetes.yaml with image: ghcr.io/pku-yuangroup/openai4s:0.3.1 before applying it to amd64 nodes; the guide covers the Secret, persistent volume, and access URL.
Release images pass container_smoke.sh before publication. Binding 0.0.0.0 inside the container makes the access token mandatory and switches the DNS-rebind Host allowlist off, so keep the published port on loopback. An unprivileged container cannot give bubblewrap the namespaces it needs, so the kernel sandbox degrades visibly and the container becomes the boundary; the container guide explains these limits.
The 612 bundled Skills are recipes โ prose, code, and the operational knowledge to run them โ and nothing about them is OpenAI4S-specific. The npm release @pku-yuangroup/openai4s-skills@0.2.0 contains 603 Skills: 42 curated + 561 pinned bioSkills. Install that fixed release with:
npx @pku-yuangroup/openai4s-skills@0.2.0 install --all # v0.2.0: 42 curated Skills
npx @pku-yuangroup/openai4s-skills@0.2.0 install --collection bioskills # v0.2.0: 561 pinned bioinformatics recipes
npx @pku-yuangroup/openai4s-skills@0.2.0 install alphafold2 boltz --target claude
npx @pku-yuangroup/openai4s-skills@0.2.0 list
npx @pku-yuangroup/openai4s-skills@0.2.0 uninstall --allnpx @pku-yuangroup/openai4s-skills <command> selects the latest npm release. To use the current repository catalog instead (612 Skills: 51 curated + 561 bioSkills), run directly from GitHub; this form follows the default branch:
npx github:PKU-YuanGroup/OpenAI4S install --all # the 51 curated Skills
npx github:PKU-YuanGroup/OpenAI4S install --collection bioskills # the 561 pinned bioinformatics recipesBoth sources use the same target and overwrite rules. --target claude writes to ~/.claude/skills, --target openai4s to <data_dir>/user-skills, and --dir <path> anywhere you name; the resolved absolute path is printed before anything is written there, and --dry-run stops at that plan. Every installed file's SHA-256 goes into a manifest beside the Skills, so a reinstall refuses to overwrite a Skill you have edited or one it did not install, and an uninstall removes only files it wrote. Every curated Skill page and the collection root under skills/ carry an Install section with their own name already filled in, so you can install from whichever page you landed on.
If you already run OpenAI4S from this checkout, you already have all 612 โ a bundled Skill takes precedence over a same-named one in your data directory. The command exists for the other direction.
The canonical bilingual documentation is published at openai4s.org/docs. Its public source and issue tracker live in Nobody-Zhang/openai4s-docs; the links below point to the code-adjacent copies kept with this repository.
| doc | what's inside |
|---|---|
| Startup guide | macOS 0.3.1 walkthrough: PyPI installation on Apple Silicon and Intel, model setup, and Doubao Search; older previews remain as historical reference |
| Upgrading | Newest first. 0.3.1 (schema 32 โ 35): stop the daemon and copy the data directory before the first start, because going back is unsupported. From 0.2.x: back up the database before the 27 โ 32 schema migration, why going back to 0.2.x is unsupported, and the access token that is now always required |
| Architecture | the hybrid action router, Action Ledger, host RPC, and lazy kernels |
| Backend extension guide | where new Tool classes, host services, repositories, and session behaviour belong |
| Model backend bring-up | local/remote GPU selection, checkpoint staging, real-inference canary admission, and connector portability |
| Skills | 51 curated Skills + 561 pinned bioSkills + how to write your own |
| Skill development record | the empirical-analysis Skills developed here, their method choices, implementation and validation |
| Research experiments | Separate data and results for bounded DiD worked-example replications using the panel preprocessing Skill |
| Remote compute | BYOC GPU jobs, host.fold, auto-provisioning |
| Science connectors | the seven public databases, their filters, and retrieval provenance |
| Web app | UI features, Action Timeline, read-only Notebook, artifacts, and implementation status |
| Web sharing | read-only session shares, the trust model, and running your own relay |
| Jupyter adapter | optional standalone Python/R KernelSpecs, install commands, and compatibility limits |
| Configuration | model providers, env vars, conda envs, CLI |
| Docker / Kubernetes | the image, compose.yaml, the cluster manifests, and what a wildcard bind actually changes |
| Supported platforms | the per-OS support tiers and why native Windows refuses to start a kernel |
| Windows / WSL2 | Ubuntu 24.04 installation, sandbox checks, lifecycle commands, mainland mirrors, and localhost proxy behavior |
| Security | defense-in-depth safety layers & remote-access notes |
| Experimental semantic judgment | default-off TypeSafe Jev layer: how to enable and disable it, what each capability sends, kill switch, doctor/audit status |
- Ship the next-generation workbench foundation: branch activation and append-only Revert/Undo projections, verified recovery with explicit Partial/Failed state, dependency-level stale propagation, durable delegation, quarantined portable Session packages, checkpointed plan/review/memory state, and dedicated 2D chemistry/genome/sequence/MSA/LaTeX renderers. Arbitrary in-memory namespace objects are deliberately not serialized; recovery remains Partial unless a safe recipe can rebuild and verify them, and Fork is offered only on records that carry a proven checkpoint mapping, so older history returns 409.
- Read-only session sharing over an outbound relay you operate, with the daemon never binding a public port and residual secrets failing the publish closed.
- An executable benchmark of end-to-end scientific workflows โ 13
workflows / 46 cases run against the real Store, kernel managers, host
dispatcher, and compute manager, where a declared
failure/permission_denied/recovered/provenanceoutcome fails when the run succeeds. Publishing comparable public results is still ahead. - Environments as a transaction (
openai4s env plan|apply|rollback): a generation is built fresh, verified, and only then pointed at atomically, so an artifact's provenance can name an immutable one.
- A notarized macOS image and arm64 packages. The Linux package ships from
v0.2.0and the Windows/WSL2 package fromv0.3.0, while 0.3.1 omits the macOS DMG (older images remain unnotarized previews), and onlyx86_64is published for Linux and Windows. A Developer ID-signed, notarized.dmgplus arm64 Linux and Windows-on-ARM packages would let every supported platform install without a toolchain. - NVIDIA scientific computing suites โ bring BioNeMo (biomolecular foundation models) and Parabricks (GPU-accelerated genomics pipelines) in as first-class Skills and BYOC backends, beyond today's NVIDIA NIM integration.
- Local GPU model serving so structure/design Skills run without remote compute.
- More BYOC providers (Modal / SLURM) beyond SSH + NVIDIA NIM.
- Stronger Linux isolation beyond bubblewrap where available (for example seccomp), and wider packaged sandbox smoke coverage.
- Keyless
web_searchbeyond DuckDuckGo (rate-limit resilience).
OpenAI4S is a community effort to keep the Code-as-Action paradigm open.
Before opening a PR, please read .github/CONTRIBUTING.md โ it defines branch naming, the PR checklist (.github/pull_request_template.md), review & release policy, and the offline-test policy. Contributors can fork the repository and open PRs without write access. Nobody-Zhang reviews and decides when to merge; other reviewers are not automatically requested.
Requires Python โฅ 3.10 and uv.
git clone https://github.com/PKU-YuanGroup/OpenAI4S && cd OpenAI4S
./setup.sh # uv sync --locked --extra science + pre-commit hook
./setup.sh --with-kernel-envs # optional: full Python + R kernel stacks
uv run pytest # offline test suite (LLM mocked)
uv run pre-commit run --all-files # format + lint everythingStyle is enforced by pre-commit โ black, isort (--profile black), and ruff, pinned in .pre-commit-config.yaml. Runtime deps: the core is zero-dependency (pure stdlib); the optional science extra pins numpy>=1.24 ยท pandas>=2.0 ยท matplotlib>=3.7.
- New Skills โ a
SKILL.md(+ optionalkernel.py) underskills/โ recipes of code, not schemas. - New providers โ a wire adapter under
openai4s/llm/providers/plus its provider definition and registry entry, or a BYOC compute provider. - Engine & UI โ the core is pure stdlib and readable; the web app is framework-free.
Keep the core dependency-free, guard optional science imports behind try/except ImportError, and make sure uv run pytest and uv run pre-commit run --all-files pass before opening a PR.
- Claude Science (Anthropic) โ the closed reference architecture whose Code-as-Action design, persistent kernel, host-RPC protocol, and safety layers OpenAI4S independently reproduces in open source.
- CodeAct โ "Executable Code Actions Elicit Better LLM Agents" โ code as a unified action interface.
- ReAct โ "Synergizing Reasoning and Acting in Language Models" โ the
tool_usebaseline this project departs from. - The science Skills stand on ColabFold / AlphaFold, ESM, OpenFold, Boltz, Chai, ProteinMPNN, DiffDock, Evo2, Borzoi, scGPT, scVI-tools and open data services (NCBI, UniProt, RCSB PDB, EBI, OpenAlex, Crossref).
Released under the MIT License โ see LICENSE.
@misc{zhang2026openai4scodeactionscience,
title={OpenAI4S: Code as Action, Science as Sessions},
author={Gongbo Zhang and Hao Li and Yu Wang and Mujie Lin and Liuzhenghao Lv and Yicheng Mao and Yimi Wang and Jun Zhu and Minhan Tang and Zhengxiang Jiang and Yusong Wang and Jiayu Yao and Kunpeng Ning and Dawei Pang and Yonghong Tian and OpenAI4S Community and Yuyang Liu and Li Yuan},
year={2026},
eprint={2609.15096},
archivePrefix={arXiv},
primaryClass={cs.AI},
url={https://arxiv.org/abs/2609.15096},
}
Auto-generated daily from the GitHub contributors graph and a maintained public-recognition list by scripts/update_contributors.py.