Skip to main content

RAG app quickstart

End-to-end: create a Knowledge Base, upload a document, and ask a grounded question against it. All examples are curl; the same requests work through the OpenAI SDK with the Advania ALT base URL.

Prerequisites​

  • An Advania ALT API key in $ALT_API_KEY
  • One PDF or Markdown file to index (a few paragraphs is enough)

1. Create the Knowledge Base​

curl -sS https://api.alt.advania.ai/v1/knowledge-bases \
  -H "Authorization: Bearer $ALT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "name": "quickstart-kb" }'

Save the returned id as $KB_ID.

2. Upload a file to Data Hub​

curl -sS https://api.alt.advania.ai/v1/files \
  -H "Authorization: Bearer $ALT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "filename": "handbook.pdf",
    "content_type": "application/pdf",
    "size_bytes": 1048576
  }'

Save the returned id as $FILE_ID. Follow the returned upload target to actually transfer the file bytes — see the interactive OpenAPI for the exact upload shape.

3. Attach the file to the Knowledge Base​

Once the file is uploaded, attach it to $KB_ID. The platform runs the indexing pipeline asynchronously — expect a short wait before content is searchable.

4. Ask a grounded question​

curl -sS https://api.alt.advania.ai/v1/retrieve-and-generate \
  -H "Authorization: Bearer $ALT_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{
    \"knowledge_base_id\": \"$KB_ID\",
    \"messages\": [{ \"role\": \"user\", \"content\": \"What does the handbook say about onboarding?\" }]
  }"

The response includes the answer plus citations to the source passages — your app can render those alongside the text so users can verify.

Next​