Quick start

Plant Omics Agent is a human-AI collaborative platform for plant omics analysis, built around two usage paths:

1. Manual Submission: pick a tool from the analysis tool catalog, fill in parameters, submit the task, inspect logs, and download results;
2. Agent-Assisted: describe what you need in natural language in the Agent Workbench; the Agent calls tools under tiered permissions with write confirmation, queries databases, reads task logs, and every submission action is subject to a confirmation gate and audit.

Both paths share the same tool, task, and result management — no matter which path a task comes from, you can track it, view its logs, and download its artifacts on the Tasks page.

Deployment and access

Self-hosted deployment (one-command Docker start)

git clone <repo-url> && cd plant_omics_agent
cp .env.runtime.secrets.example .env.runtime.secrets
# Edit .env.runtime.secrets: at minimum set DB_PASSWORD, JWT_SECRET_KEY, and one LLM API key
./scripts/auto_start.sh

The first startup takes about 2–5 minutes. No GPU required: bioinformatics tools run in local containers, and LLM calls go through cloud APIs.

Access entries

API documentation is available at http://localhost:38000/docs.

First login

1. Open the Web UI URL to reach the login page;
2. On first login the system guides you through setting the administrator account password;
3. After logging in you land on the platform home page.

After a successful login you land on the platform home page, whose top navigation switches between the four entries:

Home page tour

From the home page you can reach four core entries:

Path 1: Manual Submission in 5 steps

1. Upload data: open File Management and upload your input files (FASTQ/FASTA/expression matrix, etc.; drag-and-drop supported).
2. Pick a tool: open Analysis Tools, browse by domain or search, and click through to the tool detail page.
3. Fill parameters: command-line parameters are rendered as a form; required fields are marked with an asterisk, and file parameters let you pick from uploaded files.
4. Submit the task: click submit and the task enters the queue; the Tasks page shows status transitions and live logs.
5. Download results: when the task finishes, review the artifact list and preview or download online.

See Manual mode for details.

Path 2: Agent-Assisted in 6 steps

1. Open the Agent Workbench: launch it from the home page navigation.
2. Prepare data: upload the files you want analyzed to the platform (the Agent references workspace and task-artifact files).
3. Describe your need: for example, “Run differential expression on this expression matrix with FDR < 0.05, then do GO enrichment on the significant genes.”
4. Review the plan: the Agent searches available tools and proposes an approach; you can ask follow-up questions about why it chose them.
5. Confirm and submit: write operations such as submitting a task (L2) require your confirmation before the Agent executes them.
6. Read the results: the Agent reports task status and artifact locations, and can interpret the results for you on request.

[Screenshot coming: an Agent conversation that includes a task-submission confirmation]

See Agent-assisted mode for details.

Common startup issues

For more issues see the FAQ.

Next steps

• Usage paths: the two usage paths and three entry types, and how to choose.
• Native workflows: one-click direct submission for multi-step pipelines such as RNA-seq.
• Use cases: end-to-end analysis examples.