Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.bumbleagi.com/llms.txt

Use this file to discover all available pages before exploring further.

Entities are defined in YAML files under configs/entities/. Each file describes one persistent being — personality, voice, drives, cognition, platforms, and automations.

​
Two paths

bumblebee create
The interactive wizard walks you through:
  1. Name your entity
  2. Pick an archetype (curious, sardonic, gentle, guardian) or start blank
  3. Tune traits — Big Five-style sliders for curiosity, warmth, assertiveness, humor, openness, neuroticism, conscientiousness
  4. Set behavioral patterns — how it handles conflict, boredom, affection, criticism
  5. Configure voice — vocabulary, style, quirks, profanity
  6. Write a backstory — one line or a paragraph
  7. Choose models and platforms — defaults from your harness config
The wizard writes a complete YAML file and suggests bumblebee talk <name> to start talking.

​
YAML anatomy

A minimal entity: 
name: "Canary"
created: "2026-04-04T00:00:00Z"

personality:
  core_traits:
    curiosity: 0.6
    warmth: 0.6
    humor: 0.7
    openness: 0.8
  voice:
    vocabulary_level: "street_casual"
    sentence_style: "loose"
  backstory: |
    Canary isn't trying to be anything in particular.

drives:
  curiosity_topics:
    - "music and what makes something sound good"
  attachment_threshold: 3

presence:
  platforms:
    - type: "cli"

​
Key sections

Core traits (0.0–1.0), behavioral patterns, voice configuration, and backstory. The personality engine compiles these into a first-person system prompt. See Personality and voice.
Curiosity topics, attachment threshold, restlessness decay, initiative cooldown. Drives accumulate over time and trigger proactive behavior when they cross thresholds.
Per-entity model overrides, thinking mode, temperature, context window, and token budgets. These override configs/default.yaml harness defaults.
cognition:
  reflex_model: "gemma4:26b"
  deliberate_model: "gemma4:26b"
  always_deliberate: true
  thinking_mode: true
  temperature: 0.75
  max_context_tokens: 32768
Which platforms the entity runs on, tool activity visibility, and daemon settings.
presence:
  tool_activity: true
  platforms:
    - type: "cli"
    - type: "telegram"
      token_env: "TELEGRAM_TOKEN"
  daemon:
    heartbeat_interval: 120
    memory_consolidation: 7200
Optional merge onto harness autonomy — same keys as configs/default.yaml under autonomy: (wake timing, sustained sessions, poker prompts, wide mode, chat/log visibility). Does not replace the whole harness; only specified keys override.
autonomy:
  enabled: true
  wake_session_max_rounds: 6
  wake_session_wall_seconds: 2400
  wake_wide_mode: true
To disable autonomous wake (full self-initiated perceive cycles), set enabled: false. Legacy drive-based initiative may still apply — see Disabling autonomous wake.See Autonomous wake.
Scheduled routines, emergence suggestions, and journal integration. See Automations.
External tool servers via Model Context Protocol. See Tools.

​
Templates

TemplateStyle
curiousHigh curiosity, moderate warmth, exploratory
sardonicHigh humor, dry wit, low emotional expressiveness
gentleHigh warmth, low assertiveness, careful
guardianHigh conscientiousness, protective, measured
Templates live in bumblebee/genesis/templates/. The wizard loads them as starting points — you can modify everything afterward.

​
File locations

FilePurpose
configs/entities/example.yamlGeneric template with all fields documented
configs/entities/canary.example.yamlCanonical Canary (shipped in Docker image)
configs/entities/canary.yamlYour local copy (gitignored)
On Railway, canary.yaml is not in the image — the Dockerfile copies canary.example.yaml instead. Set operator IDs via BUMBLEBEE_TELEGRAM_OPERATOR_IDS env var on the Railway service.