So everyone's raving about the visualizations and "serendipitous discovery," but I'm looking at the API documentation and wondering if anyone's actually tried to build something real with it.
The promise is obvious: pull your collections, papers, and citations into your own Zotero, Obsidian, or custom lit review tool. Break out of their walled garden. But the docs feel like an afterthought. A couple of REST endpoints, some vague rate limiting, and authentication that looks straightforward until you hit a 500 error fetching your own data. I tried to sync a collection to a small local SQLite db for offline work, and the pagination on the `/papers` endpoint just stopped returning `next_cursor` after page 3. No error, just a dead end. Is that "serendipity" or just broken?
More importantly, the data model they expose is shallow. You get basic paper metadata, but the rich context of *why* a paper is connected—the shared references, the citation graph weights—seems locked inside their black-box visualization layer. If you're planning to build any kind of logic on top of their "similar work" recommendations, you're out of luck. You get a list of paper IDs, not the reasoning.
Here's the basic fetch I was trying to run before it fell over:
```python
import requests
headers = {"Authorization": "Bearer YOUR_TOKEN"}
params = {"collection_id": "abc123", "cursor": None, "limit": 50}
papers = []
while True:
resp = requests.get("https://api.researchrabbit.ai/v2/papers", headers=headers, params=params)
data = resp.json()
papers.extend(data.get("papers", []))
params["cursor"] = data.get("next_cursor")
if not params["cursor"]:
break
# This loop just exits prematurely on some collections.
```
Has anyone pushed this further and gotten consistent results? Or is the API just a checkbox feature for now, not a real platform to build on? I'm not looking for marketing promises—I'm looking for someone who's tried to integrate it into a weekly automated digest or a migration tool to another system.
Just my 2 cents
Just my 2 cents