Your self-hosted Paperless workspace

Documents in.
Useful work out.

Tagvico files documents, researches your Paperless archive, turns letters into trackable actions, and keeps your tags clean while every sensitive change remains visible and under your control.

Explore the workflows Read the documentation

Your Paperless copilot

Ask Tagvico

Research your archive with visible sources and tool activity.

GPT-5.3 Codex SparkChatGPT subscription · live catalog

Which insurance renewals need my attention this month?

Ai TagvicoAnswer grounded in your Paperless archive

Two policies need attention. Your household policy renews on 18 August. The vehicle policy can still be cancelled until 27 August.

1 Household policy 2026 2 Vehicle renewal notice

Create a follow-up for the vehicle policy?

The suggestion is ready, but nothing is written until you approve it.

Review action Not now

Ask a follow-up about these documents…

GPT-5.3 Codex Spark · read-only research

Clear from the first login

Four jobs. One coherent workspace.

The v3.2 navigation is organized around what you want to accomplish, not internal system names.

Review filing

See AI metadata suggestions before they touch Paperless, then approve or reject them in context.

Ask your archive

Research documents conversationally with visible searches, source documents, and model activity.

Follow actions

Turn a letter into an owner, due date, and checklist instead of another forgotten PDF.

Maintain the system

Manage automations, processing history, providers, and tag quality from predictable settings.

Companion you can inspect

See the work behind every answer.

Companion shows which Paperless tools ran, what they searched, and which documents support the answer. You can choose any live model exposed by a provider you already configured.

Visible tool timelineFollow searches, reads, comparisons, and proposed writes as they happen.

Source-backed responsesOpen the Paperless documents behind a result instead of trusting an unexplained answer.

Your configured modelsThe picker contains only live models from providers enabled in Settings.

Choose a model Configured providers only

ChatGPT subscriptionLive account catalog · selected model Selected

GitHub CopilotLive plan catalog · account authentication Live

OpenRouterConfigured cloud models Live

OllamaLocal model catalog Live

Tag unification review 3 suggestions · nothing applied

Invoice Invoices → Invoice

High confidence · same meaning across 48 documents

ApproveKeep both

Health insurance Medical insurance → Health insurance

Review needed · similar usage, not always interchangeable

ApproveKeep both

Tax 2025 Taxes → Keep separate

Low confidence · one tag carries a specific year

Dismiss

Controlled tag cleanup

Unify duplicates without losing meaning.

Choose a configured model to inspect your complete tag vocabulary. Tagvico proposes likely duplicates, explains the evidence, and waits for your decision on every single merge.

Analyze firstThe model receives names and usage context, then returns proposals without changing Paperless.

Apply approved mergesA separate operation updates only the suggestions you explicitly accepted.

Predictable by design

Your archive remains the source of truth.

Tagvico runs next to Paperless-ngx and makes its behavior inspectable before it becomes automatic.

Self-hosted

The application, configuration, history, and approvals live on infrastructure you control.

Provider choice

Use a local model, an API provider, ChatGPT, or Copilot according to your own data boundary.

Review first

Start with explicit approval for metadata, actions, and taxonomy changes. Automate only when ready.

Auditable

Model selection, tool activity, sources, retries, and applied changes stay understandable.

Transparent, privacy-first signals

Usage without visitor profiles.

Landing traffic and product activity are different signals. Tagvico counts page requests without cookies or visitor IDs. Product activity includes only installations that explicitly enable anonymous analytics.

Loading

Landing page views in the last 30 days

Requests, not unique people

Loading

Opt-in installation reports this month

Unauthenticated; small totals are suppressed

Good to know

Before you connect your archive.

Tagvico is designed to sit beside an existing Paperless-ngx installation. You keep control of the runtime, model and write policy.

Does Tagvico need a trigger tag?

No. New documents can be processed automatically without a trigger tag. Trigger tags are optional when you want to limit processing to a manually selected subset.

Can it use local models?

Yes. Configure Ollama in Settings and Tagvico loads its live model catalog. Cloud APIs, ChatGPT subscription and GitHub Copilot can be configured separately.

Will it change Paperless without asking?

You choose the write policy. Review mode stages suggestions first. Automatic mode applies the enabled metadata fields directly, while tag unification always keeps analysis, document moves and source deletion separate.

Where do credentials live?

Secrets stay in the Tagvico installation and are never returned to the browser after saving. Settings only report whether a secret is configured.

Can I undo a bad result?

Yes. Activity keeps the original snapshot for processed documents, supports rescanning with current settings and can restore the first saved state.

Documentation that matches your install

The right guide is inside every Tagvico instance.

Open the built-in documentation from the same host as Tagvico. It follows your installed major version, while archived guides remain available for older deployments.

http://your-tagvico-host:8080/docs/

Open documentation

Current stable install

Start beside Paperless.

Pin the stable image, keep the data volume, and complete the guided setup in your browser. Version 3.2.6 checks Paperless permissions and the runtime's live model catalog before saving. It uses the same persistent v3 data volume. Pin the immutable tag so upgrades stay deliberate and rollback remains simple.

Installation guide Release notes

docker compose

services:
  tagvico-ai:
    image: ghcr.io/arturict/tagvico-ai:3.2.6
    restart: unless-stopped
    ports:
      - "${TAGVICO_AI_BIND_ADDRESS:-127.0.0.1}:8080:3000"
    environment:
      TAGVICO_AI_BIND_ADDRESS: "${TAGVICO_AI_BIND_ADDRESS:-127.0.0.1}"
      TAGVICO_TELEMETRY_ENDPOINT: "${TAGVICO_TELEMETRY_ENDPOINT:-https://telemetry.tagvico.arturf.ch/v1/heartbeat}"
    volumes:
      - tagvico_ai_data:/app/data

volumes:
  tagvico_ai_data:

Then open http://localhost:8080/setup on the Docker host. For a trusted LAN browser, explicitly bind to 0.0.0.0 and enable ALLOW_REMOTE_SETUP=yes only until setup completes. Keep port 8080 behind the host firewall. Back up the data volume before every upgrade.

Documents in.Useful work out.