Build on AICE
One key for your organization, typed clients in Python and TypeScript, and a guide for every service, from your first request to a streaming agent chat.
import osfrom aice import Aiceaice = Aice( api_key=os.environ["AICE_API_KEY"], # one tenant key per organization user_id=current_user.id, # who this request acts for base_urls=SERVICE_URLS, # host + path for each service)recs = aice.reco.recommend("user-1", n=10)hits = aice.knowledge.search("payment terms")Services
Every package and version- Agent PlatformStreaming conversational agents with tools, memory and human approval.aice-agent-platform @aiceafrica/agent-platform
- RecommendationTrack what people do, train a model, and ask what to show them next.aice-recommendation @aiceafrica/recommendation
- KnowledgeUpload documents, organise them into scopes, and search them semantically.aice-knowledge @aiceafrica/knowledge
- CrawlerCrawl a site within a page budget, review it, then publish it to Knowledge.aice-crawler @aiceafrica/crawler
- Data BrainConnect databases and ask questions of them without handing over credentials.REST API for now
- KYCVerify identity documents, capture faces, and review the results.REST API for now
- SandboxRun untrusted code in an isolated session with its own files and quotas.aice-sandbox Python only
Keys and acting users
Why one key serves every service, and how each request names the user it acts for.
Build an agent chat
Stream replies to a browser through your own backend, with titles and follow-ups.
Errors and retries
One error hierarchy across every client, and which failures are safe to retry.