connecting…
Files
Wiki
Source chunks
Wiki chunks
⚠ Abliterated model — content filters removed
Main
Ingest
Admin
Kali: checking…
Sandbox: checking…
Workspace files
Loading…
Start a conversation with your knowledge base.
Generate wiki entry or briefing
Generated entry will appear here and be saved to your vault.
Upload files to knowledge base
Drop files here or click to browse
Supports: .md .txt .pdf .docx .pptx .html .json .csv .epub
Upload results will appear here.
YouTube URLs — one per line
Paste YouTube URLs above to download transcripts.
Clip a web page into your knowledge base
Paste a URL above to clip the page text and add it to your knowledge base.
Runtime Settings
0.3
10
200
30
✓ Saved
Web Search
5
✓ Saved Enable per-role via web_search: true in YAML
Role Management
Knowledge Base Actions
Action results will appear here.
Source Files
Wiki Entries
Source Chunks
Wiki Chunks
Test Retrieval — shows raw chunk matches without LLM
Results will appear here.
ChromaDB Collections
Your conversation history
No conversations yet.
User Management
No users yet.
Active Sessions
Loading…
LLM Provider

Switch between Ollama (local, free) and AWS Bedrock (managed cloud, billed per-token). Bedrock requires AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, and AWS_DEFAULT_REGION environment variables set before starting the server.

Ollama — Local Models

Select the active model from those available in your local Ollama instance.

Loading…
AWS Bedrock — Cloud Models
Persistent Memory

Facts the assistant remembers across all conversations. The LLM can also write memories using <REMEMBER key="…">…</REMEMBER> tags.

Telegram Bot Status
Checking…
Bot Configuration
Usage

/start — greeting
/new — clear conversation history
/memory — list stored memories
Just type any message to ask the knowledge base.

Scheduled Tasks

Tasks run on a cron schedule and save their output as markdown notes in the vault. Use standard 5-field cron syntax, e.g. 0 8 * * * for daily at 08:00 UTC.