4.6 KiB
4.6 KiB
VoidBot Architecture Overview
This initial scaffold follows the phase-one shape from the spec:
Discord Gateway
-> Command Router
-> Permission Engine
-> Context Builder
-> Job Queue
-> Provider Adapter
-> Storage Layer
- PostgreSQL-backed jobs, audit events, and interaction memory
- persistent local archives for RAG
- Qdrant or local JSON vector stores for retrieval
Current state
apps/botowns Discord connectivity, routing, and command handling.apps/workerpolls owner jobs, runs the local Discord-safe Codex lane, serves bounded history-search tool requests from the RAG index, and prepares handoff bundles when needed.packages/corecontains permissions, context assembly, Postgres-backed state storage, style loading, and the rotating system message catalog.packages/providersisolates provider contracts from Discord code.packages/ragcontains the message archive, chunking pipeline, local embedding backends, pluggable vector-store backends, and log import state.packages/sandboxdefines policy profiles and a dry-run policy runner.
VoidBot's Verse-facing service contract lives in
docs/architecture/voidbot-verse-service-contract.md. The short version:
VoidBot owns Discord cognition, archive/source retrieval, typed Void self-state,
and repo Persona compatibility rails; Huginn owns Persona/.cc runtime inspection
stewardship; Bifrost owns governed work transport; Odin discovers
provider-owned CultMesh surfaces; Eve/CultUI is the presentation contract.
Current shortcuts and tradeoffs
- Postgres now owns jobs, audit events, and interaction memory, but the archive side of RAG is still file-backed under
.voidbot/. - Retrieval now uses a persistent vector backend with a local embedding backend by default (
qwen3-embedding:0.6bthrough Ollama). VECTOR_STORE_KIND=local_jsonkeeps the zero-service path available for local development, butVECTOR_STORE_KIND=qdrantis now supported for a proper vector database.- Discord history vectors and source-tree/lore vectors stay in separate physical partitions so source indexing no longer rewrites the entire history corpus.
- The bot live-ingests every message that matches the configured indexing rule and can also backfill from exported log files.
- Source and lore indexing can also be kicked off by local repo
pre-pushhooks, which launch a detached incremental resync of the pushed repo into the source archive and source-side vector store. - The owner Codex flow can perform repeated read-only history searches through a bounded tool loop instead of relying on a single retrieval snapshot.
- Stock Discord replies are loaded from
config/system-messages.jsonand rotated through shuffled variants to avoid repetitive canned responses. - Changing embedding models requires a full vector rebuild from the archived messages, which is handled by
npm run rag:rebuild. - Local and offsite backups now have a verifier plus a recurring watchdog, so backup freshness and remote retention stop relying on operator vibes.
owner_codexhas two lanes:local_exec_owner_onlytries a read-onlycodex execpass for direct Discord repliesmanual_packageremains available when you want explicit approval-gated packaging
- When the local Discord-safe lane is not enough, the worker posts a handoff notice and writes the full context bundle to
.voidbot/artifacts/<job-id>/. - The current swarm operator surface is published as an Eve/CultUI CultMesh
provider at
asgard.starfire.voidbot/swarm/eve/tui, backed by.voidbot/status/cultmesh/voidbot-swarm-state.cc; the legacycultmesh://voidbot.local/eve/providers/voidbot.swarmkey is only a route, and any generated browser page is only a debug lowering of that provider surface.
Upgrade path
- Keep the current
voidbot.swarmCultMesh publication discoverable through Odin. - Add
.ccwitnesses or native typed CultCache documents for remaining JSON-only status surfaces before treating them as Verse state. - Move Persona and repo Persona
.ccinspection to Huginn-owned tooling while VoidBot remains the Discord compatibility carrier. - Publish separate Eve/CultUI surfaces for Discord, archive, source, repo Persona, and swarm state instead of growing a bespoke dashboard authority.
- Add a true restore drill path instead of stopping at backup verification plus freshness monitoring.
- Add moderation, budgeting, and rate limits before enabling
openai_apifor member traffic. - Expand worker-side run records and admin tooling around the interaction memory/event stores.
- Expand sandbox execution from dry-run policy checks to real constrained runners.