Write
Artifact
Builds slide decks, Word documents, spreadsheets, PDFs and HTML pages from a structured spec, then edits or rewrites them on request, each change a new version.
pptx · docx · xlsx · pdf · html · create · edit · rewrite
AI that generates DOCX, PPTX and XLSX from company documents
DocsGPT agents turn company documents into DOCX, PPTX, XLSX, PDF and HTML files: they analyse data in a sandbox, render the file and keep every version. File bytes stay on the server and never pass through the model.
pptx · docx · xlsx · pdf · html · versions · sandboxed python
Last updated:
Artifacts
An artifact is a file an agent produces: a presentation, a Word document, a spreadsheet, a PDF or an HTML page. Preview it beside the chat, download it, or ask for changes; every edit adds a version, and any earlier version can be restored.
In workflows, artifacts appear in the run view and pass from one step to the next by reference. The model works with text and references, never the file bytes, and per-user quotas keep storage in check.
pptx · docx · xlsx · pdf · html · versions · restore · per-user quotas
From documents to a versioned file
- 01Your filesattachments · knowledge
- 02Agentreads · plans
- 03Code Executorpython in the sandbox
- 04Artifactspec → pptx · docx · xlsx
- 05Versionsv1 · v2 · restore
Text version
Your attached files and knowledge go to the agent, which reads them and plans the output. It analyses data with Python in the sandbox through the Code Executor, and the Artifact tool renders the file from a structured spec. Every edit adds a version that can be restored. File bytes stay on the server; the model sees text, run summaries and references.
Sandboxed code execution
The Code Executor runs Python in a sandbox session for the conversation or workflow run, about 60 seconds per run, with longer jobs started in the background. Install packages from the code, and every file it writes becomes an artifact.
Attach a spreadsheet and the agent can open it with pandas and chart it; the chart is shown back to vision models so they can read what they drew. Secrets are stripped from the sandbox environment, its network access can be locked down with the egress policy, and you can require approval before each run.
Code Executor
- Runtime
- Python, a session per conversation or run
- Time
- ~60 s per run · longer jobs in the background
- Packages
- install from the code
- Backends
- self-hosted runner (default) · Daytona
- Safety
- secrets stripped · egress policy · approval per run
docsgpt — code executor
Three tools
Three tools do the work: one writes, one runs and one reads. They are opt-in per agent, and an agent or a workflow step can use them alone or together. Answers can also draw Mermaid diagrams right in the chat.
artifact · code executor · read document · mermaid in chat
Run
Code Executor
Runs Python in the sandbox to clean data, compute and chart. Every file the code writes becomes a downloadable artifact.
python · pandas · matplotlib · files → artifacts
Read
Read Document
Parses an uploaded or produced file into text, Markdown, structured JSON with tables, or chunks, so a workflow step can work on its contents.
pdf · docx · pptx · xlsx → text | markdown | json
How it runs privately
Choose the self-hosted sandbox runner and code execution, file rendering and artifact storage all run inside your network. Daytona is a managed cloud sandbox for multi-tenant workloads; it is not for air-gapped estates.
Four deployment postures, one code base. Compare deployment options →
Inside your network
- read document · parsing
- self-hosted sandbox runner
- artifact store · versions
- per-user quotas
managed · on-premises · air-gapped
Cloud
Arc53 operates the sandbox and the artifact store.
Air-gapped
Self-hosted sandbox runner only, inside your network. No Daytona, no external network.
Example
6+ h → under 5 min
IP application drafting: 6+ hours to under 5 minutes per application, with lawyer review as the final step.
Source: Balt Alnoor
Next: Code nodes in workflows · approval before a tool runs · how documents get in · legal drafting · deployment options · pricing
Frequently asked questions
Can AI generate a PowerPoint from my documents?
- Yes. An agent reads the source documents, works out the content, and the Artifact tool renders it as a PPTX you preview and download from the conversation. Ask for changes and you get a new version, with the earlier one kept. DOCX, XLSX, PDF and HTML work the same way.
Is code execution sandboxed?
- Yes. Code runs in a sandbox, either a runner you host yourself or Daytona's managed sandboxes, with secrets stripped from the environment and about 60 seconds per run. The sandbox's network access can be locked down with the egress policy, and each run can require a person's approval.
Do my files pass through the model?
- No. The model receives text, a summary of each run and references to artifacts; file bytes stay on the server and move between the sandbox and the artifact store.
Can I get previous versions of a generated file?
- Yes. Every edit adds a version (v1, v2, …), and any earlier version can be opened, downloaded or restored.
Can I run this air-gapped?
- Yes, with the self-hosted sandbox runner, which runs inside your network. Daytona is a cloud option and is not for air-gapped estates. See air-gapped deployment.
See DocsGPT on your documents
A 30-minute demo on your use case, or a free on-premises review with an engineer.