Introduction¶

DeepSeek Harness (DSH) supports plugin-based extension, allowing developers to connect external data sources or tools into DSH’s conversational flow. The dsh-strawberry-data-hub plugin aims to bridge Strawberry Data Hub’s strawberry research data (covering genome, multi-omics, literature, and prediction tools) with DSH’s natural language capabilities.

With this plugin, developers can perform literature retrieval, gene resolution, sequence export, and multi-omics context queries directly in DSH, without switching to a separate web interface.

Plugin Overview¶

• Name: dsh-strawberry-data-hub
• Maintainer: 798645706
• Version: 0.1.0-beta.8 (Beta 2)
• License: MIT
• Core value: Converts strawberry-research-specific data interfaces into a toolset that can be called from DSH, supporting integration of gene, literature, sequence, and BerryPlot/BerryLocus prediction workflows.

Core Features¶

The plugin integrates 37 tools, covering the entire process of data retrieval, analysis, visualization preparation, and result validation. The main capabilities are grouped into the following categories:

• Literature and evidence retrieval

  • sdh_literature_search: Searches strawberry research literature and preserves DOIs, evidence identifiers, and research materials.
  • sdh_evidence_check: Checks evidence IDs, DOIs, original excerpts, and numeric literal matches for structured claims.
  • sdh_data_catalog: Browses catalogs of genomic and other data resources.

• Gene and genomic context

  • sdh_gene_resolve: Resolves species and assembly version from a gene ID, with follow-up support when multiple candidates are found.
  • sdh_gene_context: Queries functional annotations and available expression, epigenomic, and other contextual information.
  • sdh_pangenome_query / sdh_graph_variation_query: Queries pangenome and graph variation evidence.
  • sdh_synteny_search: Queries synteny after selecting a reference version.
  • sdh_family_search / sdh_tf_search / sdh_mirna_search: Queries gene families, transcription factors, and miRNAs.

• Sequence processing and analysis

  • sdh_sequence_fetch: Queries protein/CDS metadata and validates sequences.
  • sdh_sequence_export: Generates a FASTA file and provides an attachment and a local web download link.
  • sdh_sequence_analysis: Sequence toolbox, ORF prediction, batch sequences, and primer design.
  • sdh_crispr_query: Queries CRISPR data.

• Multi-omics and functional analysis

  • sdh_single_cell_query / sdh_transcriptome_query / sdh_epigenome_query: Queries single-cell, transcriptome, and epigenome data.
  • sdh_functional_enrichment: GO / KEGG enrichment with an explicit species and assembly background.
  • sdh_metabolite_search / sdh_coexpression_query: Queries metabolome features and co-expression results.

• BerryPlot and BerryLocus workflows

  • sdh_berryplot_prepare: Prepares a BerryPlot plotting plan and returns a draft to be confirmed.
  • sdh_berrylocus_prepare: Prepares a BerryLocus prediction draft with complete reference and SNV parameters.
  • sdh_task_confirm / sdh_task_status: Submits a task after separate confirmation in DSH and queries the status of the same task.
  • sdh_prediction_result: Retrieves completed BerryLocus prediction results.
  • sdh_plot_download: Validates and downloads BerryPlot figures and data.

Installation and Configuration¶

Prerequisites¶

• Node.js: Version 24 or newer.
• DeepSeek Harness: A model must be configured.
• DSH toolchain: CLI / dsh-tools version 0.1.5-rc.2.

Installation command¶

Run the following command in the DSH command line to install the plugin:

dsh plugin --profile strawberry add dsh-strawberry-data-hub@0.1.0-beta.8

Configuration file¶

Configure the service address and timeout settings in the corresponding cordis.patch.yml:

- id: strawberry-data-hub
  config:
    baseUrl: https://sci.hainanu.edu.cn/strawberry/
    timeoutMs: 45000
    researchOnly: true

• researchOnly: Optional; disabled by default. When enabled, only the 37 SDH tools described above and ask_user_question are allowed, suitable for a strawberry-research-only environment.
• timeoutMs: Range 100–120000 ms, used for read-only interfaces; task workflow requests use a fixed 90-second timeout.

Usage Examples¶

1. Literature and gene retrieval¶

After installation, you can ask DSH directly. For example, to retrieve the research background for a specific gene:

Search for studies on FvSTOP1 regulation of strawberry anthocyanins, and list the research materials and DOIs.

Query FvesChr6G00057790.1, species Fragaria vesca, assembly v6.0 (Horticulture Research. 2023), for annotation and expression information.

2. Sequence export¶

After obtaining gene information, you can request sequence export:

Export the protein and CDS of the gene above into a FASTA file and provide a download link.

The plugin returns validated sequences (including SHA-256 checksums) and generates a local download link.

3. BerryPlot plotting and BerryLocus prediction¶

When using BerryPlot or BerryLocus, follow the workflow: “prepare draft -> confirm task -> retrieve result”:

Please draw an expression heatmap for FxaC_10g00030, FxaC_10g00031, and FxaC_10g00070 in the Camarosa dataset. Show the scheme first, then submit after confirmation.

Execution steps:
1. The plugin returns a structured parameter draft and does not perform actual calculation.
2. After separately approving sdh_task_confirm in DSH, the task starts.
3. Query the task status, and after success download the figure files and data.

Notes¶

• Memory isolation: Task sessions and confirmation credentials are stored only in the memory of the current DSH process; DSH model keys are not forwarded to the server.
• Download link validity period: Download links are only for the local machine running DSH and are valid for 15 minutes; they become invalid after restart or conversation switch.
• No automatic retry: If throttling or permission denial is encountered, the operation stops and is not automatically retried; after an explicit denial, the user must decide whether to try again.
• Data sources: This plugin distributes code only (MIT license) and does not include the databases or literature corpora themselves. Data and literature are used under the licenses of their respective sources.
• Coordinate systems: Genomic intervals use 0-based half-open coordinates; SNV draft positions and CRISPR intervals use 1-based coordinates. They must not be mixed.
• Prediction results: Model scores are not equivalent to experimentally validated functional effects. Prediction results retain the scientificValidation flag.
• File preview: If plot-data.json is no larger than 24 KB, a validated full-content preview is returned; larger files are provided for download only.

Summary¶

dsh-strawberry-data-hub adds 37 strawberry-research-specific tools to DSH. Through natural language, it supports full-chain operations from literature retrieval, gene resolution, and sequence export to BerryPlot plotting. When using it, pay attention to the task confirmation workflow and the validity period of download links.