vector_store so existing clients keep working if you change the base URL and key. The ids are the same object.
Create collections at Knowledge bases (display_name up to 50 characters, optional description).
Dashboard tabs
Supported uploads include PDF, Office (
.doc, .docx, .ppt, .pptx, .xls, .xlsx), text/markup (.txt, .md, .html, .csv, .tsv, .json), and common source files. Max 512 MB per file. Status is in_progress, then completed or failed.
Connectors
From the knowledge base, you can connect:- Google Drive (Docs, Sheets, Slides exported then ingested)
- OneDrive
- Dropbox
- Notion (pages as text)
- Amazon S3 (and S3-compatible)
/knowledge-bases/connectors/oauth/callback.
API naming
Your API key needs the Vector stores capability. A knowledge-base upload ephemeral key can POST files only to that id.
Add documents
Send eitherfile or file_id, not both.
upload_paths on the SDK walks directories. CLI: synapsai upload-vector-store-files.
Search
query may be a string or list. Optional filters use OpenAI comparison/compound shape. ranking_options.score_threshold drops low scores. max_num_results is 1–50 (default 10). Hits include file_id, filename, score, attributes, and content.
HTTP routes
Base URLhttps://api.synapsai.cloud.
List pages are OpenAI-style (
data, has_more, limit 1–100). Write routes are limited separately from search — see Rate limits.
MCP
Mount:https://api.synapsai.cloud/mcp (Bearer API key). Tools: search_knowledge_base (kb_id, query) and retrieve_knowledge_base_content (kb_id, document_id).