[Skip to content](https://www.innguma.com/en/resources/software/library/#contenido)

Glossary · Software

# The Innguma Library: your organisation's watch memory

**In short** 

Innguma's Library is the repository where what your team validates in its watch is kept and built up, with its tags, comments and a link to its source, searchable as a history. Newsletters are sent from it, with summaries generated by AI. It is what turns scattered signals into accumulated knowledge: search, watch, history and context.

## Why does a watch need memory?

A single signal says little. A competitor’s investment announcement gains meaning when it is read alongside the patent it filed a year ago and the tender it won six months ago. Without a shared repository, that reading depends on someone remembering it, and it is lost when that person changes post.

Knowledge management is standardised as a management system: ISO 30401:2018 sets requirements for creating, maintaining, reviewing and improving a knowledge management system in any organisation, and in October 2026 ISO has it under review, with a new edition at draft stage (ISO/DIS 30401).[\[1\]](https://www.innguma.com/en/resources/software/library/#ref-1) UNE 166006:2018 includes storage among the requirements of the watch process, alongside search, processing and distribution.[\[2\]](https://www.innguma.com/en/resources/software/library/#ref-2)

## What is kept from each signal, and what it is for

-   **The validated signal.** Only what your team has reviewed and decided to keep goes in. Unlike a search engine, the Library does not collect everything published, but what has gone through filtering, AI classification and validation.
-   **Tags.** They group and retrieve by the topics that matter to your organisation, which do not always match the AI’s general categories: an internal project, a customer, a product line.
-   **Comments.** They explain why a signal is relevant and to whom. They are the part of the knowledge that is not in the source: your team’s interpretation.
-   **Source.** Each signal links to the publication it comes from, so that any statement can be checked years later.

## Free tags or a controlled vocabulary?

How tags are assigned decides whether the history can be retrieved three years from now. Documentation resolved this decades ago with thesauri: lists of preferred terms with their synonyms and relationships. The reference standard is ISO 25964-1:2011, on thesauri for information retrieval, whose second edition is being published in October 2026.[\[3\]](https://www.innguma.com/en/resources/software/library/#ref-3)

| Approach | Advantage | Risk | When it makes sense |
| --- | --- | --- | --- |
| Free tags | Quick; each analyst tags as they think | Synonyms and variants (‘H2’, ‘hydrogen’, ‘green hydrogen’) that split the history | Teams of one or two people, at the start |
| Closed list of tags | Reliable retrieval; reports comparable over time | Falls short if it is not reviewed | Several people validating, or several areas consulting |
| Closed list with preferred terms and synonyms | Brings variants under one term and keeps consistency between areas | Requires a person responsible for maintaining it | Corporate watches with years of history |

## Newsletters sent from what has been validated

Newsletters are sent from the Library, so that each recipient receives information that has already been reviewed, not a raw list of results. Innguma generates newsletter summaries with AI so that the reader grasps the essentials before opening each signal, and the original source is one click away.

## Querying the history from your AI assistant

An AI assistant answers better when it works on retrieved and reviewed documents than when it relies only on what it learnt in training. This is the idea of retrieval-augmented generation (RAG), set out by Lewis and others in 2020, whose own paper acknowledged that providing provenance for answers was still an open problem.[\[4\]](https://www.innguma.com/en/resources/software/library/#ref-4) A Library in which each signal has a source, a tag and a comment is exactly that kind of base.

Innguma offers connectivity through the Model Context Protocol (MCP), the open standard for connecting AI applications to external systems,[\[5\]](https://www.innguma.com/en/resources/software/library/#ref-5) currently compatible with ChatGPT, Claude, Microsoft Copilot and Gemini. With the capabilities that your organisation enables in its integration, from the assistant you can search the published content, manage published news items (edit or delete them) and publish summaries, insights, tags or additional fields. What is automated, and what goes through human review first, is decided by your team.

For example, you could ask ‘summarise what we have validated this quarter on solid-state batteries’, or ‘prepare a briefing with this competitor’s signals, with their links’. These are illustrative examples. Your AI provides analytical capacity. Innguma provides context.

## How to keep a Library useful

1.  **Validate with judgement, not by volume.** A Library with everything captured is as hard to use as having none.
2.  **Add before renaming.** Renaming a tag breaks the continuity of the history; adding a new one does not.
3.  **Write down the why.** In a year nobody will remember why a news item seemed important; the comment will.
4.  **Tidy up once a year.** Merge the tags that mean the same thing and retire those nobody uses.

## Sources and references

1.  \[1\] [ISO. ISO 30401:2018 Knowledge management systems — Requirements (under review). 2018](https://www.iso.org/contents/data/standard/06/86/68683.html)
2.  \[2\] [Utrilla, F. (UNE). Watch and intelligence systems in R&D management. UNE Magazine, No. 3. 2018](https://revista.une.org/3/sistemas-de-vigilancia-e-inteligencia-en-la-gestion-de-la-id.html)
3.  \[3\] [ISO. ISO 25964-1 Thesauri and interoperability with other vocabularies — Part 1, 2nd edition (in publication; will replace the 2011 edition). 2026](https://www.iso.org/standard/86713.html)
4.  \[4\] [Lewis, P. et al. Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. arXiv:2005.11401. 2020](https://arxiv.org/abs/2005.11401v4)
5.  \[5\] [Model Context Protocol. What is the Model Context Protocol (MCP)? Official documentation](https://modelcontextprotocol.io/)

## Related

- [Factory: where watch signals are processed, classified and validated](https://www.innguma.com/en/resources/software/factory/)
- [What is MCP (Model Context Protocol)?](https://www.innguma.com/en/resources/faq/what-is-mcp/)
- [Academy](https://academy.innguma.com/en/)
- [Can I always see the original source of a signal?](https://www.innguma.com/en/resources/faq/original-source/)

- [Request a demo](https://www.innguma.com/en/demo/)
- [Back to the glossary](https://www.innguma.com/en/resources/glossary/#g-plataforma)