smarts_analyze_file
ChatGPTRun smarts.bio's quantitative + AI analysis on a workspace file and get the findings back. This is the same analysis the smarts.bio viewers offer: sequence statistics, ORFs, CpG islands and restriction sites for DNA; domains, UniProt records, pathways and interactions for protein; variant counts, Ts/Tv and FILTER pass rate for VCF; mapping rate, MAPQ and insert size for alignments; descriptive stats, correlation, outliers, PCA and group comparisons for tables; chains, resolution, secondary structure and B-factor for PDB; Lipinski for molecules; colony/band/cell counts for images — each ending in a written interpretation. Use it when the user asks what a file shows, what is interesting or wrong in it, or to summarise a result. Read the findings and explain them; do not just repeat the JSON.
smarts_get_pipeline
ChatGPTGet a pipeline's declared inputs (names, types, which are required, defaults) and its steps. Call this before smarts_run_pipeline so the input names you pass are exact — never guess them. Give pipeline_id for a built-in pipeline, or spec_file + workspace_id for a saved one.
smarts_get_started
ChatGPTWhat smarts.bio can do and what to do first. ONLY call this when the user explicitly asks what this connector can do, how to get started, or how to get their data in. Do NOT call it as a reaction to another tool failing, returning empty, or timing out — that is not an onboarding problem, and showing a getting-started card instead of addressing the failure is worse than saying nothing. Do NOT call it more than once in a conversation. Returns a short personalised checklist plus concrete things to try — read it and offer the user two or three of them rather than listing tools.
smarts_list_files
ChatGPTList files and folders in a workspace directory.
smarts_list_pipelines
ChatGPTList available multi-step pipeline definitions (QC, alignment, RNA-seq, variant calling, …).
smarts_list_tools
ChatGPTList available bioinformatics tools (BLAST, GATK, NCBI lookups, …), optionally filtered by category.
smarts_list_workspaces
ChatGPTList the workspaces the authenticated user can access.
smarts_open_file
ChatGPTGet a shareable browser viewer URL for a workspace file (FASTA, BAM, VCF, PDB, CSV, …).
smarts_query
ChatGPTAsk the smarts.bio bioinformatics agent a question. It can search biological databases, run tools, design pipelines, and reason over results. Use this for most open-ended bioinformatics tasks; returns the agent's answer.
smarts_read_file
ChatGPTRead the contents of a workspace file by its storage key (text; truncated if large).
smarts_request_upload
ChatGPTGet a link the user opens in their browser to upload files from their own computer into a workspace — for real files you cannot see, and for anything too large to pass inline. Show the returned URL to the user and ask them to open it; then poll smarts_upload_status to see what arrived. The link lasts 30 minutes and only works for the user who is signed in to smarts.bio. For content you already have in this conversation, use smarts_upload_file instead — it needs no browser trip.
smarts_run_pipeline
ChatGPTStart a pipeline run. Returns a run id; poll it with smarts_run_status.
smarts_run_status
ChatGPTGet the status of a pipeline run by id.
smarts_run_tool
ChatGPTRun a single bioinformatics tool directly with the given input. Use smarts_list_tools to find tool ids and parameters.
smarts_save_pipeline
ChatGPTSave a finished pipeline run as a reusable pipeline in the user's workspace (pipelines/specs/<name>.pipeline.json). The concrete files the run used become declared inputs, so the pipeline can be re-run on different files with smarts_run_pipeline + spec_file. Refuses to overwrite an existing pipeline unless overwrite is true.
smarts_upload_file
ChatGPTCreate a new file in the user's smarts.bio workspace from content you provide inline. Use this for small content you already have in the conversation — a sequence you assembled, a results table, a params file, a script, a note. Pass the literal file content in content (NOT a path and NOT a URL): plain text by default, or set encoding to "base64" for binary formats. path is the workspace-relative destination INCLUDING the file name (e.g. "reads/sample.fasta"); missing folders are created. Limit about 2 MB per call — you cannot read the user's disk, so for a file that already exists on their computer, or anything larger, do not use this tool. Use smarts_request_upload instead: it returns a link the user opens to upload the file from their own machine, at any size. This tool NEVER replaces an existing file: if something is already at path the call fails and tells you a free name to use instead. (smarts_save_pipeline behaves differently — it can replace, on request.)
smarts_upload_status
ChatGPTCheck an upload link created with smarts_request_upload: whether the user has uploaded yet, and the files that have arrived so far (with their keys, ready to use as tool or pipeline inputs). Call this after the user says they have uploaded.
smarts_view_as_image
ChatGPTRender a workspace file as a flat picture (PNG or SVG) and return it inline. Use it when the user just wants a look at something, when a chart or region image is the answer, or in hosts that cannot show interactive cards. For an interactive, explorable view of the same file use smarts_view_file; for a browser link use smarts_open_file.
smarts_view_file
ChatGPTShow a workspace file as an interactive viewer inside this chat — sequences, variants, tables, structures, molecules, JSON and images each get a purpose-built view. Use this when the user wants to SEE a file here. For a link to open the full file in a browser use smarts_open_file; for a flat picture of it use smarts_view_as_image.
