Motivation
Buzz has no built-in way to connect OpenViking (volcengine/OpenViking) as a context/memory/skills backend. Users running Buzz who want persistent, hierarchical agent memory must manually configure OpenViking, build their own adapters, and manage credentials and errors themselves. This hits anyone trying to give Buzz agents durable, semantic memory beyond NIP-AE's engram model — especially teams running multi-agent workflows (e.g. content/newsletter teams) that need agents to recall past decisions and mistakes reliably.
Proposed solution
Add an official OpenViking integration for Buzz that:
- Provides an OpenViking entry in Buzz's integrations or context-provider settings.
- Supports embedded/local and remote OpenViking deployments.
- Allows secure entry of endpoint and credentials.
- Detects installation/configuration status.
- Exposes OpenViking memory, knowledge, and skills to Buzz agents.
- Lets users scope OpenViking to specific agents/workspaces.
- Shows clear status and actionable errors for connection, auth, and compatibility issues.
If full native integration isn't feasible, provide and document an official MCP-compatible adapter that Buzz can use without unofficial wrappers.
Alternatives considered
- Manual MCP/protocol configuration by users, maintained individually per install.
- Running OpenViking separately with no Buzz-side integration or visibility.
- Using generic notes vaults (e.g. Obsidian) or standalone vector DBs instead of OpenViking.
- Unofficial community-maintained adapters, which increase duplication and maintenance burden.
Additional context
OpenViking provides unified agent memory, knowledge, and skills with hierarchical (L0/L1/L2) retrieval, replacing the fragmented vector-store model of traditional RAG. Buzz's own NIP-AE engram model covers encrypted agent memory but has no semantic/vector recall layer — OpenViking would fill that gap.
Related:
Motivation
Buzz has no built-in way to connect OpenViking (volcengine/OpenViking) as a context/memory/skills backend. Users running Buzz who want persistent, hierarchical agent memory must manually configure OpenViking, build their own adapters, and manage credentials and errors themselves. This hits anyone trying to give Buzz agents durable, semantic memory beyond NIP-AE's engram model — especially teams running multi-agent workflows (e.g. content/newsletter teams) that need agents to recall past decisions and mistakes reliably.
Proposed solution
Add an official OpenViking integration for Buzz that:
If full native integration isn't feasible, provide and document an official MCP-compatible adapter that Buzz can use without unofficial wrappers.
Alternatives considered
Additional context
OpenViking provides unified agent memory, knowledge, and skills with hierarchical (L0/L1/L2) retrieval, replacing the fragmented vector-store model of traditional RAG. Buzz's own NIP-AE engram model covers encrypted agent memory but has no semantic/vector recall layer — OpenViking would fill that gap.
Related: