Featured AI automation workflow · Literature Library + LLM gateway
Retina Evidence Hub
Designed and built end to end · Eye Care International · Pilot moving to production
An AI workflow that turns new ophthalmology literature into curated, published evidence for Medical Affairs teams. Papers queued in a Literature Library review collection are fetched into the app, where an agent on the company's approved LLM gateway API retrieves each one, scores relevance, quality, impact and novelty against the selected prompt, and recommends include or exclude with a confidence and a newsletter-ready summary. Then it waits for a reviewer.
Approved papers are published to the Retina Evidence Hub collection, and Literature Library is tagged automatically (REH-Include or REH-Exclude and Published, with NEW removed), so both collections stay in step without re-keying.
From the Figma Make scaffold to the LLM gateway relay, the review screens, the Literature Library tag sync and the move onto the corporate design system, I developed the whole application in VS Code with Claude Code.
Visual Studio CodeClaude CodeFigma MakeCorporate design system
- 7agent steps, from retrieval to metadata, with a reviewer gate before publishing
- 7evaluation prompts, from standard evaluation to EC priorities screening
- 3LLM gateway environments with GPT and Claude models and a context-overflow fallback
- 2Literature Library collections, a review queue and the curated hub, kept in step by tags
- React 18
- TypeScript
- Vite
- Express relay
- Literature Library REST API
- LLM gateway: GPT · Claude
- Corporate design system
- VS Code
- Claude Code
The demo runs in your browser on invented sample data: no real data, people or systems are connected.
How it works
How a paper becomes evidence
- Deterministic
- AI
- Human
-
1 · Queue
A Literature Library review collection
Candidate papers from brand literature collections land in the REH-Review collection, tagged NEW.
-
2 · Retrieve
Paper details and full text
The agent pulls each paper's metadata, abstract and full text where available through the Literature Library API, and flags abstract-only papers.
-
3 · Evaluate
Scored against the team's priorities
LLM gateway scores relevance, quality, impact and novelty, 25 points each, with the selected prompt, for a score out of 100 and a confidence.
-
4 · Recommend
Include or exclude, with the reasoning
A newsletter-ready summary, an at-a-glance table, structured screening fields and suggested tags.
-
5 · Review gate
A reviewer decides
Nothing is published until a person approves, edits or overrides the recommendation.
-
6 · Publish and tag
Published, tagged, logged
Approved papers move to the Retina Evidence Hub collection, Literature Library is tagged, and every step is in the fetch log.
Architecture
How papers and AI move through the hub
A React 18 and TypeScript app between two Literature Library collections, with every AI call relayed same-origin to the company's approved LLM gateway. Drawn from the application's own engineering notes.
- QueueCandidate papers are added from Literature Library collections to the REH-Review collection, tagged NEW.
- FetchThe app pulls only NEW papers, with their abstracts and full text where available, and logs every call.
- EvaluateThe agent sends each paper and the chosen prompt to LLM gateway through the same-origin relay; GPT or Claude scores and summarizes it.
- ReviewA reviewer approves, edits or overrides each recommendation. Nothing publishes without that decision.
- PublishApproved papers move to the Retina Evidence Hub collection with their summary, score and metadata.
- TagLiterature Library gets REH-Include or REH-Exclude and Published tags, and NEW is removed, so the queue stays clean.
The workflow, screen by screen
Eleven screens, in order
From a Literature Library collection to published, tagged evidence. People's and authors' names, IDs and the team's screening rationale are blurred. Select a screen to enlarge it.
1 · Find the evidence in Literature LibraryBrand literature collections in Literature Library, such as the Brand C product literature output, are where candidate papers come from.
2 · Queue papers for reviewPapers go into the REH-Review collection straight from Literature Library.
3 · The review queue in Literature LibraryThe REH-Review collection and its decision tags: NEW, Published, REH-Include and REH-Exclude.
4 · Fetch new papersFetch from Literature Library pulls only papers tagged NEW into the review queue, with a fetch log and an abstract-only warning.
5 · Choose the evaluation promptSeven prompts, from standard evaluation to EC priorities screening, set what the agent scores against.
6 · The agent scores and recommendsThe agent pipeline gives each paper an include or exclude call, a score out of 100 from relevance, quality, impact and novelty, and a confidence.
7 · Read the AI evaluationA newsletter-ready summary and an at-a-glance table for the reviewer. One line about the team's priorities is blurred.
8 · Approve, publish and apply metadataAfter approval the agent finalizes the tags and writes the metadata back to Literature Library through its API.
9 · Workflow completeStructured screening fields, the suggested tags and the decision tags applied in Literature Library. The scoring rationale is blurred.
10 · Published in the hubPublished papers carry their decision tags and score.
11 · The curated collection in Literature LibraryThe Retina Evidence Hub collection, with evidence tags such as anti-VEGF refractory, biomarkers, BRVO and CRVO.
