FEVER Demonstration
What does a self organizing file system mean? Find out!
Search the corpus
Search multimedia files - this demo runs with NO GPUs on AWS!
Curate a dataset
Describe your dataset and desired file attributes. In this demo, click the media cards and hit "Build dataset". You can tag a dataset version and download a parquet or NDJSON at will, and pull the same, consistent sample later.
How you'd download the dataset:
Dupe and near-dupe checks
Catch fraud and duplicate user submissions, within and across accounts. Save storage costs, or clean your datasets with rapid similarity indexing.
Similarity threshold
Cross account fraud
Fraudsters may submit duplicate product listings, duplicate profile photos, across multiple accounts. What's the chance the same user has 25 accounts? You could see it in your own data:
Duplicate dataset curation
Query your media with agents or automations
FEVER bundles a REST API and MCP server so your app can get media intelligence straight out the gate!
curl
curl -X POST https://YOUR-APPLIANCE/v1/search \
-H "Authorization: Bearer $FEVER_API_KEY" \
-H "Content-Type: application/json" \
-d '{"text": "a red truck", "k": 24, "kind": "image"}'
Python (httpx)
import httpx
r = httpx.post(
"https://YOUR-APPLIANCE/v1/search",
headers={"Authorization": f"Bearer {FEVER_API_KEY}"},
json={"text": "someone says drizzle", "k": 24, "kind": "audio"},
timeout=30,
)
for hit in r.json():
print(hit["blob_id"], hit["score"], hit.get("span", {}).get("text"))
JavaScript (fetch)
const res = await fetch("https://YOUR-APPLIANCE/v1/search", {
method: "POST",
headers: {
Authorization: `Bearer ${process.env.FEVER_API_KEY}`,
"Content-Type": "application/json",
},
body: JSON.stringify({ text: "sunset over water", k: 24 }),
})
const hits = await res.json()
MCP server
{
"mcpServers": {
"fever-retrieval": {
"command": "python",
"args": ["-m", "fever_service.mcp_server"],
"env": {
"FEVER_MCP_SERVICE_URL": "https://YOUR-APPLIANCE",
"FEVER_MCP_API_KEY": "<api-key>",
"FEVER_MCP_CUSTOMER_ID": "cust-42"
}
}
}
}
6 tools: search_by_image, find_duplicates, search_media, get_media_info, list_media_frames, list_media_transcript