Integrate
Most of the research work in Ishvana runs through the browser or Lagan’s built-in pipelines. But occasionally you need something the browser doesn’t hand you well — a structured response from an API, a page of Wikipedia content you want directly referenced in your Legendry, a dataset from an academic source that you want enriched with entity detection. The Integrate tab is the workshop for those cases. It’s where you run external enrichment against sources that aren’t random web pages (Wikipedia, academic databases, structured references) and where you can hit arbitrary REST APIs to pull specific data into your research workflow. It’s the least-used tab in the Research module for most authors, and the most useful tab for the few authors whose projects genuinely need it.
The tab has two main features: enrichment from structured sources, and an API fetcher for one-off requests. They share the panel because they’re both “pull external data, turn it into something usable” workflows, even though they’re technically different operations.
Enrichment from structured sources
Section titled “Enrichment from structured sources”The enrichment tool takes a text passage (or the current page’s extracted content) and runs it against one or more structured enrichment sources. The sources are curated APIs that return clean factual data:
- Wikipedia. Queries the Wikipedia API directly for entities mentioned in your text. Returns each matched entity’s canonical Wikipedia content plus relationships between entities. Useful when your text mentions real-world people, places, or concepts and you want to pull their Wikipedia data into your project.
- Academic. Queries academic databases (ArXiv, CrossRef, OpenAlex, Google Scholar depending on availability) for papers, authors, and research relevant to your text. Returns structured references with abstracts, authors, and DOIs. Useful for historical research, scientific research, or any work where you need to back claims with academic sources.
For each source, you pick:
- Source toggle. On/off per source. You can enable just Wikipedia, just Academic, or both.
- Max items. How many results to return from each source. Range 1-20, default 5. More items means more context but more noise.
- Language. For Wikipedia specifically, which language’s Wikipedia to query. Default is English.
Click Enrich and Lagan runs the query. The results appear in the tab as a structured list — one card per matched entity, each with:
- Entity name.
- Source (Wikipedia / Academic).
- Confidence score. How confident the extraction is that this is a real match.
- Summary or abstract.
- Full content excerpt (expandable).
- Actions: view the source page, add to Lore, add to research bookmarks, copy citation.
How enrichment differs from the Research Pipeline
Section titled “How enrichment differs from the Research Pipeline”Both the Research Pipeline and Integrate can consult external sources. They do it differently.
The Research Pipeline runs a natural-language query and synthesizes a context summary. You ask “how did medieval cities manage water supply” and Lagan decides which sources to consult, runs them, synthesizes the output. You don’t see the raw sources — you see the synthesis.
Integrate enrichment runs structured entity extraction on a specific text passage and returns raw matched entries from each source. You paste a paragraph, click Enrich, and you see each entity that appeared in the paragraph matched against Wikipedia (or Academic) with the full matched content displayed.
Pipeline is query → synthesis. Integrate is text → entity matches from structured sources. Different shapes, different use cases.
Use Pipeline when you have a question. Use Integrate when you have a passage and want to cross-reference every real-world thing in it against a source of truth.
The API fetcher
Section titled “The API fetcher”Below the enrichment panel, a collapsible API fetcher section lets you make arbitrary HTTP requests to any REST API.
Fields:
- URL. The endpoint you’re hitting.
- Method. GET or POST (other methods are supported for edge cases but rarely needed).
- Headers. Key-value pairs for custom headers (authorization, content-type, etc.).
- Body (for POST). The request body, usually JSON.
- Send. Fires the request.
The response comes back as displayed content — formatted JSON if the response is JSON, plain text otherwise. You can copy the response, parse it, or feed it to Lagan for analysis.
The API fetcher is deliberately simple — it’s not a full REST client like Postman, it doesn’t save request history, it doesn’t have collections or environments. It exists for the specific case where a fiction writer needs to look up one specific thing from one specific API without having to learn how to make HTTP requests themselves.
Use cases for the API fetcher
Section titled “Use cases for the API fetcher”A few specific examples:
Historical dataset lookups. You’re writing historical fiction and you want to check a specific date against a historical events API. Paste the URL, fire, read the response.
Language lookups. You’re building a conlang and you want to check a word’s etymology against a linguistic API. Same workflow.
Map coordinates. You’re verifying a location’s real-world coordinates against a geographic API. Same.
Reference database lookups. You’re writing about mythology and you want to query a mythology database for a specific deity. Same.
In all four cases, the API is small, specific, and one-off. The fetcher handles it without requiring you to switch to a dedicated developer tool.
Saving enrichment results
Section titled “Saving enrichment results”Any enrichment result can be saved to your project in several ways:
- Add to Lore. Creates a Legendry entry from the enriched entity, with the full source content as the entry’s body and the source URL as metadata.
- Save as bookmark. Creates a Smart Bookmark with the enriched entity as the content. Lower commitment than a lore entry; stays searchable via ChromaDB.
- Copy citation. Copies a formatted citation (author, title, URL, date accessed) to your clipboard for use in footnotes or references.
- Copy content. Copies the raw content of the matched entity.
Most enrichment workflows end with “Add to Lore” for entries that should be part of canonical project data, or “Save as bookmark” for entries that are reference material you might want later.
What Integrate isn’t
Section titled “What Integrate isn’t”- Not a web scraper. The tab doesn’t harvest arbitrary pages. It queries structured sources with defined schemas.
- Not a translation service. Wikipedia returns content in the language you requested; it doesn’t translate.
- Not a full academic search. The Academic enrichment is limited to available APIs and doesn’t cover every discipline equally. For genuine academic research, you want a dedicated academic search tool.
- Not for high-volume work. Enrichment makes API calls. For bulk enrichment (hundreds of entities at once), you’d hit rate limits on the underlying sources. The tab is for one-off or small-batch work.