Feature Request
Problem
KIWI should support configurable AI bots that can participate in channels and answer questions using selected knowledge graph context. The current chat experience is user-driven, but teams need reusable bot personas that administrators can configure centrally and attach to channel workflows.
Proposed Solution
Integrate Chat SDK (https://chat-sdk.dev/) as the foundation for channel-aware bots in KIWI.
The implementation should provide a composable bot configuration model that allows system administrators to define reusable bots with their own behavior, model settings, instructions, and graph access. Bots should be attachable to channels with a specific team context and one or more knowledge graphs. When a bot is mentioned in a channel, it should generate an answer in that channel using the configured graph scope.
Requirements
- Add an administration flow for creating, updating, and disabling bot configurations.
- Allow bot definitions to include identity, description, instructions, model/runtime configuration, and access constraints.
- Allow bots to be associated with a team and one or more knowledge graphs when added to a channel.
- Support adding and removing configured bots from channels.
- Trigger bot responses only when the bot is explicitly mentioned in a channel message.
- Ground bot answers in the configured graph scope and enforce existing team, graph, and authorization boundaries.
- Persist bot configuration, channel membership, and graph associations.
- Ensure bot responses integrate with the existing channel message flow and streaming UX where appropriate.
- Provide clear error handling when a bot lacks access to a graph, is disabled, or cannot generate a response.
Acceptance Criteria
- An administrator can configure reusable bots.
- A channel can include one or more configured bots.
- Each channel bot instance can be scoped to a team and multiple graphs.
- Mentioning a bot in a channel causes it to respond in that channel.
- Bot responses use only the graphs and permissions configured for that channel context.
- The implementation includes API, persistence, and frontend support.
- Relevant tests or verification steps are added for bot configuration, channel membership, mention handling, and permission enforcement.
Technical Notes
- Evaluate how Chat SDK should compose with the existing AI SDK-based model infrastructure.
- Reuse existing graph retrieval and authorization mechanisms where possible.
- Prefer a data model that separates reusable bot definitions from per-channel bot assignments.
- Keep the integration extensible for future bot capabilities beyond mention-based replies.
Feature Request
Problem
KIWI should support configurable AI bots that can participate in channels and answer questions using selected knowledge graph context. The current chat experience is user-driven, but teams need reusable bot personas that administrators can configure centrally and attach to channel workflows.
Proposed Solution
Integrate Chat SDK (https://chat-sdk.dev/) as the foundation for channel-aware bots in KIWI.
The implementation should provide a composable bot configuration model that allows system administrators to define reusable bots with their own behavior, model settings, instructions, and graph access. Bots should be attachable to channels with a specific team context and one or more knowledge graphs. When a bot is mentioned in a channel, it should generate an answer in that channel using the configured graph scope.
Requirements
Acceptance Criteria
Technical Notes