Research & Analysis🤖claude🌐universal
OmniScientist V2
> A local-first, CLI-first open-source personal research agent. > Chat in your terminal (like Codex / Claude Code) to do grounded, cited literature Q&A, > research ideation, scientific figures, paper writing & review — with an auditable > hypothesis→claim→evidence trail, a reproducible run ledger...
Quick Start
Install this Skill
npx skills add tsinghua-fib-lab/OmniScientist-V2# OmniScientist V2 > **A local-first, CLI-first open-source personal research agent.** > Chat in your terminal (like Codex / Claude Code) to do **grounded, cited** literature Q&A, > research ideation, scientific figures, paper writing & review — with an auditable > hypothesis→claim→evidence trail, a reproducible run ledger, and an honesty `--verify` pass. > Durable state is stored locally (no MySQL/Redis); configured model, connector, and IM providers > still receive the data required for each request. Skills use a portable SKILL.md baseline for > Claude Code, Codex & OpenClaw (and MCP), and the same agent is reachable from > **WeChat / Feishu / DingTalk**. This repository is the **official next-generation (V2) implementation** of the OmniScientist project introduced in [*OmniScientist: Toward a Co-evolving Ecosystem of Human and AI Scientists*](https://doi.org/10.48550/arXiv.2511.16931) (Shao et al., arXiv:2511.16931). V2 distills the earlier HelixForge research-agent OS into a local-first, single-machine agent while preserving auditable research objects, workflows, and provenance. See [`CITATION.cff`](CITATION.cff) for software and paper citation metadata. For the full design and how it compares to Claude Code/Codex and other open-source research agents, see [`cli/docs/research-agent-design.md`](cli/docs/research-agent-design.md). The current repository metadata reports package version `2.0.0rc8` (Git tag `v2.0.0rc8`). This is a temporary prerelease identity, not the stable release line: the first stable V2 release will be `2.0.0`, followed by compatible `2.0.x` patch releases. Once the stable release is published, normal installations and updates should not pin a version; the package manager resolves the latest stable release. Pin `==<version>` only for reproducibility, rollback, or release verification. Release artifacts are published to the public index by `cli/scripts/release.sh`. ## Install and initialize ### Requirements - macOS, Linux, or Windows 10/11 - Python 3.11 or newer - [`uv`](https://docs.astral.sh/uv/) or [`pipx`](https://pipx.pypa.io/) for an isolated installation - Node.js 20.9 or newer to run the bundled research-pptx renderer; npm is required only for the first `omni skills setup research-pptx` (pnpm is not a substitute) - A model endpoint is optional for installation: the bundled `mock` provider supports offline setup and deterministic evaluation ### Quick install PyPI is the stable package authority. Install one isolated copy, then launch Omni: ```bash uv tool install OmniScientist-V2 # or pip install OmniScientist-V2 # or pipx install OmniScientist-V2 omni ``` The browser surface is an optional package extra for direct package-manager installs: ```bash uv tool install "OmniScientist-V2[web]" # or: pip install "OmniScientist-V2[web]" omni web ``` On the Web home screen, **Scientist persona** lets you choose an installed SoulAgent persona and describe the current research task. The same controls are available under **Settings → Scientist personas** for search, activation, task-specific refresh, switching, and restoring standard Omni. Web and CLI read the same workspace-scoped persona catalog and active state; every change runs as a normal, auditable SoulAgent Task, and an in-progress change is resumed after a page refresh instead of being submitted twice. These unpinned commands install the latest stable release. Until `2.0.0` is published, use the checkout installer below or intentionally select a prerelease for testing. An exact stable version is optional and should be used only when it is intentionally required, for example `pip install 'OmniScientist-V2==2.0.0'`. The first `omni` launch runs setup, prepares the lockfile-pinned Node runtime, and starts or repairs the single Home Service. Later launches are fast local checks. If a package manager replaces the package directly (`uv tool upgrade OmniScientist-V2` or `pipx upgrade OmniScientist-V2`), the next `omni` detects the new installation fingerprint and completes runtime/config/service convergence without downloading the Python package again. Use one product command for normal updates: ```bash omni update ``` It delegates the Python package step to the owning package manager, then converges managed runtime state, retires legacy daemons, refreshes the supervisor launcher, restores the Home Service, and verifies readiness. SQLite schema changes remain data-preserving and run with backups when each store is first opened by the new version. `omni update` updates to the latest stable release. If the installation is managed directly by a package manager, use its ordinary unpinned upgrade command instead: ```bash uv tool upgrade OmniScientist-V2 pip install --upgrade OmniScientist-V2 pipx upgrade OmniScientist-V2 ``` ### Install from a checkout From a downloaded/cloned source tree, the repository installer deploys the local checkout snapshot, including uncommitted work. It creates an isolated `uv tool` and boot ... (truncated)
Repository
8stars
2forks
Details
Related Skills
More research & analysis skills you might like
pr-improver
Research & AnalysisRuns an autonomous review-and-fix improvement loop over the current branch's changes until a PR review comes back clean, scoped mechanically to the directories the branch touched. Reviews are performed by an installed PR-review skill (default: pr-review-toolkit's review-pr). Use to fix review findings on a branch before opening or updating a pull request ('clean up this branch', 'fix this PR until review passes', 'run review-and-fix on my changes'). NOT for a one-time review — run the PR-review skill directly.
🤖claude🌐universal
7.5k
643
View
code-improver
Research & AnalysisRuns an autonomous review-and-fix improvement loop over any code target — a skill, plugin, module, or directory — using a reviewer the user names: any installed skill or agent. Keeps a cross-round findings ledger, escalates when fixes stop converging, and guards scope mechanically. Use when asked to 'improve this code until review passes', 'run an improvement loop with <reviewer>', or to iterate review-and-fix with a specific reviewer. For skills prefer the skill-improver entry; for a branch prefer pr-improver.
🤖claude🌐universal
7.5k
643
View
skill-improver
Research & AnalysisRuns an autonomous review-and-fix improvement loop over a Claude Code skill until a review comes back clean, with a cross-round findings ledger, escalation when fixes stop converging, and a mechanical scope guard. Reviews are performed by the plugin-dev skill-reviewer agent. Use to fix skill quality issues, iteratively refine a skill, or resume a loop after an escalation ('fix my skill', 'improve this skill until it passes review', 'skill improvement loop'). NOT for a one-time review — use the plugin-dev skill-reviewer agent directly.
🤖claude🌐universal
7.5k
643
View
audit-augmentation
Research & AnalysisAugments Trailmark code graphs with external audit findings from SARIF static analysis results, weAudit annotation files, and version-gated Trailmark 0.4.x binary-analysis graph exports. Maps findings to graph nodes by file and line overlap, creates severity-based subgraphs, and enables cross-referencing findings with pre-analysis data (blast radius, taint, etc.). Use when projecting SARIF results onto a code graph, overlaying weAudit annotations, importing binary graph findings, cross-referencing Semgrep, CodeQL, or binary-analysis findings with call graph data, or visualizing audit findings in the context of code structure.
🌐universal
7.5k
643
View