liquid-agent / README.md
superhy199148's picture
Clarify external LLM research use cases and API ownership
fab203a verified
|
Raw History Blame Contribute Delete
8.36 kB
---
pretty_name: LIQUID-Agent software releases
license: other
license_name: liquid-agent-early-access
license_link: https://lcrc-ai.github.io/liquid-agent-wiki/docs/en/release-terms/
language:
- en
tags:
- software
- software-distribution
- liquid-biopsy
- research
- local-llm
- windows
- macos
viewer: false
extra_gated_heading: Request LIQUID-Agent early access
extra_gated_description: Provide your full name and institution, complete the linked questionnaire, and submit your request for manual review.
extra_gated_button_content: Submit access request
extra_gated_prompt: |
Access is for approved non-commercial research and evaluation of LIQUID-Agent v0.1.0.
Before requesting access, complete the [full early-access questionnaire](https://forms.cloud.microsoft/Pages/ResponsePage.aspx?id=as2-rtQxAUuVzoJ0r-hT2crr7c84XABNtm_gHP1xL7VUN0owN1JPSFlaRzlCWVlCME1IMTNKWFlENC4u).
Review and agree to the [Terms of Use](https://lcrc-ai.github.io/liquid-agent-wiki/docs/en/release-terms/). Requests are reviewed manually; submitting a request does not automatically grant download access.
Do not submit patient data, confidential research data, or API credentials in either form.
extra_gated_fields:
Full name: text
Institution: text
---
# LIQUID-Agent v0.1.0
**A local-first workspace for liquid-biopsy research.**
LIQUID-Agent helps researchers inspect supported datasets, review analysis plans, run local scientific tools, and explore the resulting figures, tables and reports. Its browser workbench and interactive terminal interface use the same local analysis kernel. Conversations, task memory and linked tasks support continuing investigation.
This Hugging Face dataset repository distributes **software installers and release documentation**. It contains neither a trained machine-learning model nor research datasets.
[Project homepage](https://lcrc-ai.github.io/liquid-agent-wiki/) · [User documentation](https://lcrc-ai.github.io/liquid-agent-wiki/docs/en/) · [Terms of Use](https://lcrc-ai.github.io/liquid-agent-wiki/docs/en/release-terms/)
## Request access
The release description is public; installer downloads require approval for the early-access research programme.
1. Complete the [full early-access questionnaire](https://forms.cloud.microsoft/Pages/ResponsePage.aspx?id=as2-rtQxAUuVzoJ0r-hT2crr7c84XABNtm_gHP1xL7VUN0owN1JPSFlaRzlCWVlCME1IMTNKWFlENC4u).
2. Use your individual Hugging Face account to request access to this repository. Provide your full name and institution.
3. Read the programme's [Terms of Use](https://lcrc-ai.github.io/liquid-agent-wiki/docs/en/release-terms/). Download the appropriate installer after access is approved.
Both the full questionnaire and the Hugging Face access request are required. Submitting them does not automatically grant access. Do not include patient records, confidential research data or API credentials in either application.
## Available installers
| Platform | Package | Compatibility | Installation guide |
| --- | --- | --- | --- |
| Windows | [LIQUID-Agent-0.1.0.exe](https://huggingface.co/datasets/LCRC-AI/liquid-agent/resolve/main/v0.1.0/LIQUID-Agent-0.1.0.exe?download=true) | Windows x64; tested on Windows 11 x64 only. | [Windows instructions](https://huggingface.co/datasets/LCRC-AI/liquid-agent/blob/main/v0.1.0/INSTALL-Windows.md) |
| macOS | [LIQUID-Agent-0.1.0-macos-arm64.dmg](https://huggingface.co/datasets/LCRC-AI/liquid-agent/resolve/main/v0.1.0/LIQUID-Agent-0.1.0-macos-arm64.dmg?download=true) | Apple Silicon ARM64, macOS 15 or newer. Intel Macs and older macOS versions are unsupported. | [macOS instructions](https://huggingface.co/datasets/LCRC-AI/liquid-agent/blob/main/v0.1.0/INSTALL-macOS.md) |
The macOS compatibility minimum does not mean every eligible operating-system version has been tested.
These native installers include the Web workspace, documentation and CLI. You do not need to install Python, Node.js, npm or conda separately. Follow the platform-specific installation guide distributed alongside your package.
After installation, reopen your terminal and run:
```text
liquid-agent
liquid-agent cli
```
The first command opens the local browser workspace; the second starts the interactive CLI.
## Local and external language models
**Local inference:** install a compatible local model and select its profile in LIQUID-Agent. Model weights are optional, separate downloads; they are not included in the core installer. Downloading weights does not select or load a model and does not switch providers. Downloads check supported platforms and available disk space; working memory is checked when you explicitly select a model for inference. Allow adequate storage, RAM and download bandwidth for your chosen model.
For highly sensitive research data, prefer a local model on your own computer in an offline workflow after the required software and weights have been installed. Alternatively, use an institution-approved private endpoint with the appropriate access controls and governance. Do not begin by trying sensitive datasets through a public external LLM service.
**External LLM APIs:** consider a cloud provider for exploratory scientific discussion, analysis planning, or interpreting non-sensitive summaries when the material is approved for sharing outside your institution. Choose this route when reliable internet access is available and the provider fits your research needs. Configure it using your own API key; no shared project API key or prepaid API credits are included. Provider access, service availability and any API charges are your responsibility.
Scientific processing runs locally. When an external model is selected, your conversational requests and controlled scientific summaries can be sent to that provider. The application's summary boundary is designed to avoid sending raw dataset contents from scientific tools, but ordinary text entered in chat and aggregate results may still contain sensitive information. Local scientific execution does not make every conversation offline.
Do not paste identifiers, patient records, confidential material or credentials into chat. Review your institution's permissions and the selected provider's data-handling terms before enabling remote inference. Read the [local-data and model privacy guide](https://lcrc-ai.github.io/liquid-agent-wiki/docs/en/guides/local-data-privacy/) for the boundary and its limitations.
## Installation and security checks
Obtain installers only from this official repository and compare their SHA-256 checksums with the values below.
The Windows EXE is unsigned. The macOS package uses ad-hoc signing and is not Apple Developer ID signed or notarized. Follow standard per-application operating-system approval procedures.
Antivirus software may flag or quarantine bundled application components. After verifying the official download and checksum, and if your institution's security policy permits, consider a narrowly targeted allowlist or exclusion for the LIQUID-Agent installation folder before installation to prevent essential files being quarantined. If a detection occurs, review the reported detection before restoring files or reinstalling. Keep antivirus protection enabled; do not use global exclusions or disable Gatekeeper.
## Early-access terms
This is a closed-source early-access distribution: LIQUID-Agent application source code is not supplied here. Bundled third-party components retain their respective licences.
Access is for approved **non-commercial research and evaluation**. Redistribution, commercial exploitation and modification require the permissions specified in the [Terms of Use](https://lcrc-ai.github.io/liquid-agent-wiki/docs/en/release-terms/).
LIQUID-Agent is a research prototype, not a medical device or clinical diagnostic service. Clinical, diagnostic and patient-management use is not permitted. Researchers must independently review outputs and remain responsible for data rights, research governance and scientific interpretation.
## SHA-256 checksums
These values identify the exact v0.1.0 installer files distributed here.
```text
83c4cae75cc4ceeb96446f80afee1a57cf8fb11d11955f76cc5e0e96f032bc0f LIQUID-Agent-0.1.0.exe
e3290313ade6936f40238ffd6b89290daca4f79800027a76bf1d52e97a1134ca LIQUID-Agent-0.1.0-macos-arm64.dmg
```