MattiaIppoliti/ciele

What it solves

Ciele provides a complete system for building and deploying AI assistants that answer questions based on an organization's specific knowledge (websites, files, and FAQs) rather than general knowledge. It bridges the gap between a simple chatbot and a production-ready support system by adding routing, human escalation, and quality tracking.

How it works

Users can ingest data via website crawling, file uploads, or manual FAQ curation. The system uses a flow engine to route user intents to specific actions, such as delivering fixed messages or performing generative searches within the ingested knowledge. It can be self-hosted (running offline if connected to an OpenAI-compatible API) or used via a managed cloud version. For production management, it includes a conversation inbox for transcripts and an answer-quality tracker.

Who it’s for

Organizations that need a customer-facing or internal AI assistant that is grounded in their own data, with the ability to escalate complex queries to human teams and monitor the quality of AI responses.

Highlights

• Knowledge Ingestion: Supports crawling websites, uploading files, and curating FAQ pairs with citations for every answer.
• Flow Engine: An authoritative router that uses intent classification to trigger specific sequences of actions.
• Human Escalation: Integrated help desks with channels, forms, and ticketing integrations to hand off conversations to humans.
• Operational Tools: Includes a conversation inbox, analytics, and an improvements tracker for answer quality.
• Multi-tenancy: Built-in support for organizations, members, and roles with database-level isolation.